This action might not be possible to undo. Are you sure you want to continue?
Assessment of Buildings
ATC58 50% Draft
Prepared by
APPLIED TECHNOLOGY COUNCIL
201 Redwood Shores Parkway, Suite 240
Redwood City, California 94065
www.ATCouncil.org
Prepared for
U.S. DEPARTMENT OF HOMELAND SECURITY (DHS)
FEDERAL EMERGENCY MANAGEMENT AGENCY
Michael Mahoney, Project Officer
Robert D. Hanson, Technical Monitor
Washington, D.C.
PROJECT MANAGEMENT COMMITTEE
Christopher Rojahn (Project Executive Director)
Ronald O. Hamburger (Project Technical Director)
John Gillengerten
Peter J. May
Jack P. Moehle
Maryann T. Phipps*
Jon A. Heintz**
William T. Holmes **
STEERING COMMITTEE
William T. Holmes (Chair)
Roger D. Borcherdt
Anne Bostrom
Bruce Burr
Kelly Cobeen
Anthony B. Court
Terry Dooley
Dan Gramer
Michael Griffin
R. Jay Love
David Mar
Steven McCabe
Brian J. Meacham
William J. Petak
* ATC Board Contact
** exofficio
STRUCTURAL PERFORMANCE
PRODUCTS TEAM
Andrew S. Whittaker (Team Leader)
Gregory Deierlein
John D. Hooper
YinNan Huang
Nicolas Luco
Andrew T. Merovich
NONSTRUCTURAL PERFORMANCE
PRODUCTS TEAM
Robert E. Bachman (Team Leader)
Philip J. Caldwell
Andre Filiatrault
Robert P. Kennedy
Helmut Krawinkler
Manos Maragakis
Eduardo Miranda
Keith Porter
RISK MANAGEMENT PRODUCTS TEAM
Craig D. Comartin (Team Leader)
Mary Comerio
Gregory Fenves
Mahmoud Hachem
Gee Hecksher
Judith MitraniReiser
Farzad Naeim
Hope Seligson
April 15, 2009
STRUCTURAL FRAGILITY
DEVELOPMENT CONSULTANTS
Charles Ekiert
Andre Filiatrault
Aysegul Gogus
Kerem Gulec
Dawn Lehman
Jingjuan Li
Laura Lowes
Eric Lumpkin
Hussein Okail
Charles Roeder
Benson Shing
Christopher Smith
Victor Victorsson
John Wallace
FRAGILITY REVIEW PANEL
Bruce Ellingwood
Robert Kennedy
Stephen Mahin
NONSTRUCTURAL FRAGILITY
DEVELOPMENT CONSULTANTS
Richard Behr
John Eidinger
Paul Kremer
Ali M. Memari
William O’Brien
John Osteraas
Xin Xu
RISK MANAGEMENT PRODUCTS
CONSULTANTS
Peter Morris
Scott Shell
VALIDATION/VERIFICATION TEAM
Jack Baker
David Bonneville
Charles Scawthorn
Notice
This document has been prepared by the ATC58 Project Team to assist interested parties in
obtaining an understanding of the methodology as it is being developed, and to facilitate
comment and feedback to the project team on its further development. The guidelines presented
in this document are incomplete at this time. The data and procedures are not necessarily
appropriate for use in actual projects at this time, and should not be used for that purpose. Reader
notes have been provided to describe the present status of development, and to identify portions
of the methodology that are not yet ready for implementation. The information contained herein
will be subject to further revision and enhancement as the methodology is completed in future
years.
April 15, 2009
ATC58 Contents i
Contents
List of Figures ............................................................................................... vii
List of Tables ............................................................................................... xiii
Glossary .................................................................................................... xvii
Chapter 1 Introduction ........................................................................ 11
1.1 Purpose ..................................................................... 11
1.2 The PerformanceBased Design Process ................. 11
1.3 Guideline Uses ......................................................... 13
1.4 Application ............................................................... 13
1.5 Performance Calculations ........................................ 14
1.6 Guideline Organization ............................................ 15
1.7 Limitations ............................................................... 16
Chapter 2 Performance Measures ....................................................... 21
2.1 Introduction .............................................................. 21
2.2 Factors Affecting Performance ................................ 21
2.3 Uncertainty in Performance Assessment ................. 21
2.4 Types of Performance Assessment .......................... 25
2.4.1 IntensityBased Assessments ...................... 25
2.4.2 ScenarioBased Assessments ...................... 27
2.4.3 TimeBased Assessments ............................ 28
2.5 Use of Loss Distributions in Decisionmaking ........ 29
2.5.1 Typical Building Performance .................... 29
2.5.2 Probable Maximum Loss .......................... 210
2.5.3 CostBenefit Analysis ............................... 212
Chapter 3 General Methodology ........................................................ 31
3.1 Introduction .............................................................. 31
3.2 Step 1 – Assemble Building Performance Model .... 31
3.3 Step 2 – Define Earthquake Hazards ....................... 32
3.4 Step 3 – Simulate Building Response ...................... 33
3.5 Step 4 – Develop Collapse Fragility ........................ 34
3.6 Step 5 – Damage and Loss Assessment ................... 36
3.6.1 Collapse Determination ............................... 37
3.6.2 Damage Assessment .................................... 37
3.6.3 Loss Calculation ........................................ 311
3.7 Loss Aggregation ................................................... 312
3.7.1 IntensityBased Assessments .................... 313
3.7.2 ScenarioBased Assessments .................... 313
3.7.3 TimeBased Assessments .......................... 313
Chapter 4 Implementation .................................................................. 41
4.1 Introduction .............................................................. 41
April 15, 2009
ii Contents ATC58
4.2 Building Performance Model ................................... 42
4.2.1 Occupancies ................................................. 42
4.2.2 Population Models ....................................... 43
4.2.3 Normative Quantities .................................. 45
4.3 Characterize Ground Motion .................................. 412
4.4 Structural Analysis ................................................. 412
4.5 Develop Collapse Fragility ..................................... 412
4.6 Form Realizations .................................................. 413
4.7 Damage Determination ........................................... 414
4.7.1 Damage Correlation .................................. 414
4.7.2 Fragility Specifications .............................. 415
4.7.3 Damage States ........................................... 415
4.7.4 Component Fragility Functions ................. 418
4.7.5 Consequence Functions ............................. 418
4.7.6 Damage Aggregation ................................. 420
4.8 Repair Costs ........................................................... 420
4.9 Occupancy Interruption Time ................................ 421
4.10 Casualties ............................................................... 422
4.11 Loss Distributions .................................................. 422
4.11.1 Intensity and ScenarioBased
Assessments ............................................... 422
4.11.2 TimeBased Assessments .......................... 424
Chapter 5 Ground Shaking Intensity .................................................. 51
5.1 Introduction .............................................................. 51
5.2 Building Location and Site Conditions .................... 51
5.2.1 Seismic Environment and Hazard ............... 51
5.2.2 Location ....................................................... 52
5.2.3 Site Soil and Topographic Conditions ......... 52
5.3 Spectral Adjustments for Soil Conditions ................ 53
5.4 USGSbased Ground Motion Calculator .................. 54
5.4.1 TimeBased Hazard Calculations ................ 54
5.4.2 IntensityBased Hazard Calculations .......... 57
5.5 Attenuation Relationships ........................................ 58
5.5.1 Introduction ................................................. 58
5.5.2 Functional Form .......................................... 59
5.5.3 Median Spectrum and Dispersion ............. 511
5.6 Hazard Characterization for Use with Nonlinear
ResponseHistory Analysis .................................... 512
5.6.1 Introduction ............................................... 512
5.6.2 Selection of Earthquake Ground
Motions ...................................................... 514
5.6.3 IntensityBased Assessment ...................... 514
5.6.4 TimeBased Assessments .......................... 515
5.6.5 ScenarioBased Assessments ..................... 518
5.7 Hazard Characterization for Use with Simplified
Analysis .................................................................. 519
5.7.1 Introduction ............................................... 519
5.7.2 IntensityBased Assessment ...................... 520
5.7.3 TimeBased Assessment ............................ 520
5.7.3 ScenarioBased Assessment ...................... 521
April 15, 2009
ATC58 Contents iii
Chapter 6 Response Analysis ............................................................. 61
6.1 Scope ........................................................................ 61
6.2 Nonlinear ResponseHistory Analysis ..................... 61
6.2.1 Introduction ................................................. 61
6.2.2 Mathematical Models of Components
and Elements ............................................... 61
6.2.3 Analysis Procedures .................................... 65
6.2.4 Residual Drift ............................................ 611
6.2.5 Response Data Input to PACT .................. 612
6.3 Simplified Analysis ................................................ 613
6.3.1 Introduction ............................................... 613
6.3.2 Mathematical Models of Components
and Elements ............................................. 614
6.3.3 Analysis Procedure ................................... 615
6.3.4 Response Data Input to PACT .................. 620
Chapter 7 Procedures for Collapse Assessment ................................. 71
7.1 Scope ........................................................................ 71
7.2 Collapse Analysis ..................................................... 71
7.2.1 Introduction ................................................. 71
7.3 Mathematical Models for Collapse Analysis ........... 72
7.3.1 Introduction ................................................. 72
7.3.2 Best Estimate Models .................................. 72
7.3.3 Simplified Model ........................................ 73
7.4 Analysis and Modeling Considerations ................... 75
7.4.1 Three and TwoDimensional Models ........ 75
7.4.2 Mathematical Models of Structural
Components ................................................ 75
7.4.3 Damping ...................................................... 76
7.5 Ground Motion Characterization for Collapse
Assessment ............................................................... 76
7.6 Development of Collapse Fragility Curves .............. 78
7.6.1 Introduction ................................................. 78
7.6.2 Truncation of IDA Curves for Non
Simulated Failure Modes and
Components ................................................ 78
7.6.3 Construction of a Collapse Fragility
Curve ........................................................... 79
7.6.4 Collapse Fragility Parameters ................... 710
7.6.5 Tools for Collapse Analysis of
Simplified, Planar Models ......................... 711
7.7 Collapse Descriptions for Casualty Assessments .. 713
Chapter 8. Example Applications ........................................................ 81
8.1 Introduction .............................................................. 81
8.2 Example Building #1 ............................................... 81
8.2.1 Introduction ................................................. 81
8.2.2 Seismic Hazard Characterization ................ 83
8.2.3 ResponseHistory Analysis and
Demand Parameter Matrices ..................... 813
April 15, 2009
iv Contents ATC58
8.2.4 Input of Data into the Performance
Assessment Calculation Tool (PACT) ...... 820
8.2.5 Loss Computations Using PACT .............. 825
Appendix A: Probability, Statistics & Distributions ............................... A1
A.1 Introduction ............................................................. A1
A.2 Statistical Distributions ........................................... A1
A.2.1 Finite Populations and Discrete
Outcomes .................................................... A1
A.2.2 Combined Probabilities .............................. A2
A.2.3 Mass Distributions ...................................... A3
A.2.4 Continuous Distributions ............................ A4
A.3 Common Forms of Distributions ............................. A6
A.3.1 Normal Distributions .................................. A6
A.3.2 Cumulative Probability Functions .............. A8
A.3.3 Lognormal Distributions ............................ A8
Appendix B: Fragility Group, Normative Quantity and Population
Models ............................................................................... B1
B.1 Fragility Group Classifications ............................... B1
B.2 Occupancy Default Assignment Tables .................. B1
B.3 Occupancy Default Normative Quantity Logic ....... B1
B.3.1 Commercial Office ..................................... B1
B.3.2 Residential .................................................. B1
B.4 Population Models................................................... B2
Appendix C: Default Structural Fragility Data ....................................... C1
C.1 Introduction ............................................................. C1
C.2 Vertical Seismic Framing Systems .......................... C1
C.3 Horizontal Seismic Framing Systems
(Diaphragms) ........................................................... C3
C.4 Gravity Framing Systems ........................................ C3
C.5 Default Fragility Data .............................................. C4
Appendix D: Nonstructural Fragility specifications ............................... D1
D.1 Default Nonstructural Fragility Data Tables ........... D1
Appendix E: Ground Shaking Hazards .................................................. E1
E.1 Scope ....................................................................... E1
E.2 Attenuation Relationships ....................................... E1
E.3 Fault Rupture Directivity and Maximum
Direction Shaking .................................................... E3
E.4 Probabilistic Seismic Hazard Assessment ............... E4
E.4.1 Introduction ................................................ E4
E.4.2 PSHA Calculations ..................................... E5
E.4.3 Inclusion of Rupture Directivity Effects .. E13
E.4.4 Deaggregation of Seismic Hazard Curves
and Epsilon ............................................... E13
E.4.5 Conditional Mean Spectrum and Spectral
Shape ........................................................ E16
E.5 Selection of Earthquake Ground Motions ............. E19
April 15, 2009
ATC58 Contents v
E.5.1 Introduction ............................................... E19
E.5.2 Earthquake Histories for Response
Analysis ..................................................... E19
E.6 Vertical Earthquake Shaking ................................. E19
E.6.1 Introduction ............................................... E19
E.6.2 Procedure for Site Classes A, B and C ...... E19
E.6.3 Procedure for Site Classes D and E ........... E20
E.7 SoilFoundationStructure Interaction ................... E20
E.7.1 General ...................................................... E20
E.7.2 Direct SoilFoundationStructure
Interaction Analysis .................................. E21
E.7.3 Simplified SoilFoundationStructure
Interaction Analysis .................................. E22
E.8 Alternate Procedure for Hazard Characterization
for ScenarioBased Assessment Using Nonlinear
ResponseHistory Analysis .................................... E25
Appendix F: Fragility Development ....................................................... F1
F.1 Introduction .............................................................. F1
F.1.1 Purpose ........................................................ F1
F.1.2 Fragility Function Definition ...................... F1
F.1.3 Derivation Methods..................................... F3
F.1.4 Documentation ............................................ F4
F.2 Fragility Parameter Derivation ................................. F5
F.2.1 Actual Demand Data ................................... F5
F.2.2 Bounding Demand Data .............................. F6
F.2.3 Capable Demand Data................................. F9
F.2.4 Derivation ................................................. F11
F.2.5 Expert Opinion .......................................... F11
F.2.6 Updating .................................................... F13
F.3 Assessing Fragility Function Quality ..................... F14
F.3.1 Competing Demand Parameters ................ F15
F.3.2 Dealing with Outliers using Pierce’s
Criterion .................................................... F15
F.3.3 Goodness of Fit Testing ............................ F16
F.3.4 Fragility Functions that Cross ................... F17
F.3.5 Assigning a Single Quality Level to a
Fragility Function ...................................... F18
Appendix G: Generation of Realizations for Loss Computations .......... G1
G.1 Loss Computations .................................................. G1
G.2 Realizations for Assessment Using Nonlinear
ResponseHistory Analysis ..................................... G1
G.2.1 Introduction ................................................ G1
G.2.2 Algorithm ................................................... G2
G.2.3 Sample Application of the Algorithm ........ G3
G.2.4 Matlab Code ............................................. G10
G.3 Realizations for Assessment Using Simplified
Nonlinear Analysis ................................................ G11
Appendix H: Residual Drift .................................................................... H1
April 15, 2009
vi Contents ATC58
H.1 Introduction ............................................................. H1
H.2 Prediction of Residual Story Drifts ......................... H2
H.3 Model to Calculate Residual Story Drifts ............... H5
H.4 Proposed Damage States for Residual Story
Drifts ....................................................................... H7
References .....................................................................................................I1
ATC581 Project Participants ..................................................................... J 1
April 15, 2009
ATC58 List of Figures vii
List of Figures
Figure 11 Performancebased design flow diagram ........................... 12
Figure 21 Example cumulative probability loss distributions for a
hypothetical building at four ground motion intensities .... 26
Figure 22 Example complimentary cumulative probability distributions
for loss at four ground motion intensities .......................... 27
Figure 23 Distribution of mean annual total repair cost ..................... 28
Figure 24 Cumulative loss function identifying a median (50th
percentile) loss ................................................................. 211
Figure 25 Annualized loss before and after proposed retrofit .......... 212
Figure 31 General Seismic Performance Assessment Procedure ....... 31
Figure 32 Representative collapse fragility for a hypothetical
building structure ......................................................... 35
Figure 33 Loss calculation flow chart ........................................... 37
Figure 34 Example family of fragility curves for special steel
moment frames ................................................................. 310
Figure 35 Sample consequence function for cost of repair .............. 312
Figure 36 Seismic hazard curve and timebased loss calculations ... 314
Figure 41 Graph illustrating the percent of peak occupancy, by
hour and day present .......................................................... 45
Figure 42 Representative Fragility Specification ............................. 417
Figure 43 Sample repair/replacement losses for a scenario or
intensitybased assessment ............................................... 423
Figure 44 Direct losses deaggregated by performance group for a
scenario or intensitybased assessment ............................ 423
Figure 45 Sample loss calculations for a timebased assessment ..... 424
Figure 51 Representative seismic hazard curve for a site in
San Francisco with T =1 second, plotted in semilog
format ................................................................................. 55
Figure 61 Generalized forcedisplacement behavior of
structural components ........................................................ 63
April 15, 2009
viii List of Figures ATC58
Figure 62 Definition of floor and story numbers and story height ... 616
Figure 71 Forcedisplacement relationships ....................................... 74
Figure 72 Results of Incremental Dynamic Analysis ......................... 76
Figure 73 Transforming results of IDA to a collapse fragility
curve ................................................................................. 710
Figure 74 Global forcedisplacement relationship in the SPO2IDA
space ................................................................................. 712
Figure 75 Sample SPO2IDA results for the example of
Figure 74 ......................................................................... 713
Figure 81 Photograph of the example building .................................. 82
Figure 82 Schematic structural framing plan for example building
(no scale) ............................................................................ 82
Figure 83 Elevation of typical steel moment frame for example
building (no scale) .............................................................. 83
Figure 84 Spectral accelerations for Bin I ground motions ................ 84
Figure 85 Spectral accelerations per the ChiouYoungs NGA
relationship for
W
M =7, r =1 km, strikeslip faulting
and
30
v =760 m/s, and varying as 0.47 T ........................ 86
Figure 86 Spectral accelerations for Bin S1 ground motions, 16
th
,
50
th
and 84
th
percentiles of spectral acceleration and the
11 target spectral ordinates for Bin S1 motions ................. 87
Figure 87 Sixteenth, 50
th
and 84
th
percentiles of spectral
acceleration for Bin S1 motions and demands predicted
by the ChiouYoungs NGA relationship ............................ 87
Figure 88 Screen capture from USGS ground motion calculator for
generating a onesecond hazard curve for the site of the
building .............................................................................. 88
Figure 89 Onesecond seismic hazard curve ...................................... 88
Figure 810 Seismic hazard curves for the site of the building .............. 89
Figure 811 Characterizing seismic hazard for timebased
assessment ........................................................................ 811
Figure 812 Target spectral ordinates for Bins T1 through T8 and
spectra for scaled ground motions in Bins T1 and T8 ...... 811
Figure 813 PACT Input Hub .............................................................. 820
April 15, 2009
ATC58 List of Figures ix
Figure 814 PACT General Info Screen .............................................. 821
Figure 815 PACT Building Information Screen ................................ 822
Figure 816 PACT Performance Group Quantity Screen (s) ............... 823
Figure 817 PACT View Analysis Cases Screen................................. 825
Figure 818 PACT Intensity and Scenarios Based Loss Screen .......... 826
Figure 819 PACT TimeBased Assessment Loss Screen ................... 827
Figure A1 Probability mass function indicating the probability
of “n” numbers of “headsup” outcomes in four
successive coin tosses ....................................................... A4
Figure A2 Distribution of possible concrete cylinder strengths for
a hypothetical mix design ................................................. A5
Figure A3 Calculation of probability that a member of the
population will have a value within a defined range ........ A6
Figure A4 Probability density function plots of normal distributions
with mean values of 1.0 and coefficients of variation of
0.1, 0.25 and 0.5 ................................................................ A7
Figure A5 Cumulative probability plots of normal distributions
with coefficients of variation of 0.1, 0.25, and 0.5 ........... A8
Figure A6 Probability density function plots of lognormal
distributions with median values of 1.0 and dispersions
of 0.1, 0.25 and 0.5 ........................................................... A9
Figure A7 Cumulative probability plots of lognormal distributions
with median values of 1.0 and dispersions of 0.1, 0.25,
and 0.5 ............................................................................... A9
Figure C1 Fragility specification for post1994 welded steel
moment frame .................................................................. C11
Figure C2 Fragility specification for exterior wall with structural
sheathing and cement plaster ........................................... C12
Figure C3 Fragility specification for interior wall with wood studs
and gypsum board sheathing ............................................ C13
Figure D1 Fragility for interior partitions .......................................... D1
Figure D2 Fragility for unitized glazed curtainwall ........................... D2
Figure D3 Fragility for suspended acoustic ceiling systems .............. D3
Figure D4 Fragility for gypsum ceiling on wood joists ..................... D4
April 15, 2009
x List of Figures ATC58
Figure D5 Fragility for concrete tile roofs.......................................... D5
Figure D6 Fragility for hydraulic elevators ........................................ D6
Figure D7 Fragility for roofmounted mechanical equipment ........... D7
Figure D8 Fragility for miscellaneous housewares and art objects .... D8
Figure D9 Fragility for home entertainment equipment ..................... D9
Figure D10 Fragility for desktop computer equipment............................ D10
Figure D11 Fragility for servers and network equipment .................. D11
Figure D12 Fragility for tall filing cabinets ....................................... D12
Figure D13 Fragility for unanchored bookcases ................................ D13
Figure E1 Sitetosource distance definitions (Abrahamson and
Shedlock, 1997)................................................................. E3
Figure E2 Fault rupture directivity parameters (Somerville et al.,
1997) ................................................................................. E4
Figure E3 Steps in probabilistic seismic hazard assessment
(Kramer, 1996) .................................................................. E6
Figure E4 Source zone geometries (Kramer, 1996) ........................... E7
Figure E5 Variations in sitetosource distance for three source
zone geometries (Kramer, 1996) ....................................... E7
Figure E6 Conditional probability calculation (Kramer, 1996) ......... E9
Figure E7 Seismic hazard curve for Berkeley, California
(McGuire, 2004) .............................................................. E12
Figure E8 Sample deaggregation of a hazard curve (from
www.usgs.gov) ................................................................ E15
Figure E9 Sample geometricmean response spectra for
negative, zero and positive record sets with each
record in the sets scaled to a) (0.8 )
a
S s =0.5 g and
b) (0.3 )
a
S s =0.5 g (Baker and Cornell 2006) ................. E16
Figure E10 UHS for a 2% probability of exceedance in 50 years
and original and scaled CMS for a rock site in
San Francisco .................................................................. E18
Figure E11 Analysis for soil foundation structure interaction
(FEMA, 2005) ................................................................. E21
April 15, 2009
ATC58 List of Figures xi
Figure E12 Reductions in spectral demand due to kinematic
interaction ........................................................................ E24
Figure E13 Calculation of spectral accelerations given a lognormal
distribution ....................................................................... E26
Figure F4 Illustration of (a) fragility function, and (b) evaluating
individual damagestate probabilities ................................ F2
Figure F2 Form for soliciting expert judgment on component
fragility ............................................................................ F13
Figure G1 Generation of vectors of correlated demand parameters
(Yang 2006) ...................................................................... G3
Figure G2 Relationships between demand parameters ...................... G4
Figure G3 J oint probability density functions .................................... G5
Figure H1 Idealized incremental dynamic analysis presenting
transient and residual story drift ratios .............................. H3
Figure H2 Idealized response characteristics for elasticplastic (EP),
general inelastic (GI) and selfcentering (SC) systems ..... H3
Figure H3 Idealized model to estimate residual story drift from
peak transient drift ............................................................ H6
April 15, 2009
ATC58 List of Tables xiii
List of Tables
Table 41 Recommended Default Peak Population Models .............. 44
Table 42 Example Fragility Groups for Momentresisting Steel
Frame Office Structure ...................................................... 48
Table 51 Representations of Seismic Hazard for Intensity
Based assessments ........................................................... 510
Table 43 Example Performance Groups and Quantities for
3story, Momentresisting Steel Frame Office Structure .. 49
Table 51 Representations of Seismic Hazard for IntensityBased
Assessments ....................................................................... 58
Table 61 Default descriptions and values for
c
β
............................. 67
Table 62 Default descriptions and values for
q
β
............................. 68
Table 63 Default Dispersions for RecordtoRecord Variability,
Modeling Uncertainty and Ground Motion Variability ..... 69
Table 64 StoryDrift and FloorAcceleration Correction Factors .. 619
Table 71 Sample probabilities that the specified portion of a
building will be involved in the collapse ......................... 714
Table 81 Seed Ground Motions for ResponseHistory Analysis ...... 84
Table 82 Spectral Demand Per the ChiouYoungs Relationship ...... 85
Table 83 Mean Hazard Curve ......................................................... 810
Table 84 Spectral Accelerations and MAFE (
i
λ ) for Boundaries
on the Seismic Hazard Curve........................................... 812
Table 84 Spectral Accelerations and MAFE (
i
λ ) for Boundaries
on the Seismic Hazard Curve........................................... 813
Table 86 Demand Parameters for IntensityBased Assessment...... 814
Table 87 Demand Parameters for ScenarioBased Assessment ..... 815
Table 88 Demand Parameters for TimeBased Assessment ........... 816
Table B1 Default Fragility Groups .................................................... B3
April 15, 2009
xiv Contents ATC58
Table B2 Default Assignment Commercial Office ......................... B45
Table B3 Default Assignment Multifamily Residential ................ B45
Table B4 Normative Quantities for Commercial Office
Occupancies .................................................................... B46
Table B5 Normative Quantities for Residential Occupancy .......... B46
Table B6 Default Time of Day and Day of Week Population
Variations (relative to Expected Peak Population) for
Office and Residential Occupancies ............................... B47
Table B7 Monthly Population Variations (Relative to Expected
Peak Population) for Office and Residential
Occupancies .................................................................... B48
Table C1 Classification System for Vertical Seismic Framing
Systems ............................................................................. C5
Table C2 Classification System for Horizontal Seismic Framing
Systems (Diaphragms) ...................................................... C9
Table C3 Classification System for Gravity Framing Systems ...... C10
Table E1 Ground Motion Attenuation Relationships ....................... E2
Table E2 Values of
i
η for Generating a Distribution of ( )
ai
S T .... E26
Table F1 Values of z ....................................................................... F10
Table F2 Parameters for Applying Peirce's Criterion ..................... F16
Table F3 Critical Values for the Lilliefors Test .............................. F17
Table F4 Fragility Function Quality Level ...................................... F19
Table G1 Matrix of Demand Parameters, X ................................... G4
Table G2 Mean and Variance of X ................................................. G4
Table G3 Demand Parameters, Y .................................................... G6
Table G4 Matrix
Y
D for the Sample Problem ................................. G6
Table G5 Matrix
YY
R for the Sample Problem ................................ G7
Table G6 Matrix
Y
L for the Sample Problem .................................. G8
Table G7 Matrix of Simulated Demand Parameters
(first 10 vectors) ................................................................ G9
April 15, 2009
ATC58 Contents xv
Table G8 Ratio of Simulated to Original Logarithmic Means ......... G9
Table G9 Ratio Of Entries in Simulated and Original
YY
R
Matrices ............................................................................ G9
Table H1 Damage states for residual story drifts ............................. H8
Table H2 Sample transient story drift ratios,
/ h ∆
, associated
with the residual story drift damage states of Table H1 .. H9
April 15, 2009
ATC58 Glossary xvii
Glossary
Annualized loss – over a period of many years, the average annual value of
loss,
Casualties – loss of life, or serious injury to persons, typically requiring
hospitalization
Component – one of many parts, structural and nonstructural, that together
comprise a building
Consequence function  a relationship that indicates the conditional
probability of incurring loss as a function of building damage
Correlation  a mathematical relationship that defines the extent that the
value of one parameter is dependent on the value of one or more other
parameters
Damage function – for a specific damage state, a detailed description of the
significant effects of the damage in terms of what is damaged, the repair
actions that it necessitates, the effect on occupancy, and the effects that could
result in casualties
Damage State – an extent of damage associated with a particular building
component and damage function.
Demand – a parameter, such as floor (ground) acceleration, component
deformation, story displacement, floor (ground) velocity, or component force
(stress) that is correlated with the occurrence of damage to one or more
components
Discount rate – a factor used to indicate the timevalue of money in
economic analysis
Downtime – the amount of time, following an earthquake, that a building
cannot be used for its normal intended function
Earthquake Scenario  a specific earthquake event, defined by a magnitude
and geographic location
Fragility Function– a mathematical function that indicates the conditional
probability of incurring damage associated with a particular damage state as
a function of a demand parameter
Fragility Group – the set of building components that can be assigned the
same fragility specification
April 15, 2009
xviii Glossary ATC58
Fragility Specification – a detailed description of damage states, damage
functions and fragility functions associated with a particular type of
component or collection of components
Geomean – abbreviation of geometric mean. In characterization of ground
motion intensity, the square root of the product of the value of a ground
motion parameter in each of two orthogonal directions
Intensity – the severity of ground shaking as represented by a 5%damped,
elastic acceleration response spectrum
Intensitybased Assessment – an assessment of a building’s probable
performance given that the building is subjected to a specific intensity of
ground shaking
Net present value – the present value of one or more expenditures incurred
or benefits received in the future, discounted for the time value of money
Nonstructural Component – a building component that is not part of the
structural system
Performance – the consequences of a building’s response to earthquake
shaking expressed in terms of the probable number of casualties, downtime
and direct economic loss.
Performance group – the subset of components within a fragility group that
will experience the same demand and which will produce similar
consequences
Realization  one possible outcome of a particular earthquake scenario or
intensity including a unique set of demands, damage and consequences
Repair cost – the cost, in present dollars, necessary to restore a building to
preearthquake condition, or in the case of total loss, to replace the building
with a new structure of similar type.
Return on investment – the annual income that can be derived from an
investment divided by the value of the investment
Scenariobased Assessment – an assessment of a building’s probable
performance given that the building is subjected to a specific earthquake
scenario
Structural Component – a building component that is part of the intended
vertical or lateral force resisting system, or that provides measurable
resistance to earthquakeinduced building deformations
Timebased Assessment – an assessment of probable building performance
over a specified period of time, considering all earthquake scenarios that
could occur during that period of time, and the probability of occurrence of
each.
April 15, 2009
ATC58 1: Introduction 11
Chapter 1
Introduction
1.1 Purpose
This report describes a basic methodology and recommended procedures to
assess the probable earthquake performance of individual buildings based on
their unique site, structural, nonstructural and occupancy characteristics. It
was developed as a product under a contract between the Applied
Technology Council (ATC) and the Federal Emergency Management
Agency (FEMA) to develop nextgeneration “PerformanceBased Seismic
Design Guidelines” for buildings, termed the ATC58 Project. The program
for developing this and other products is defined in the FEMA 445 report,
NextGeneration PerformanceBased Seismic Design Guidelines, Program
Plan for New and Existing Buildings (FEMA, 2006). As currently
envisioned, future products may include publications that suggest appropriate
performance characteristics for buildings of differing occupancy and use;
procedures to design new buildings or upgrade existing buildings to obtain
desired performance; and publications intended to assist design professionals,
building regulators, developers, owners, tenants, lenders, insurers and other
stakeholders to take advantage of the benefits of performancebased design
approaches.
1.2 The PerformanceBased Design Process
Performancebased design is a process that explicitly considers building
performance in the design process. This is in contrast to the typical building
design process in which building components and systems are proportioned
and detailed to satisfy prescriptive criteria contained within the building code
without direct consideration of the building’s performance. In the
performancebased design process, the designers and other stakeholders
jointly identify the desired building performance characteristics at the outset
and these performance goals then guide the many design decisions that must
be made. Figure 11 illustrates the key steps in the performancebased
design process.
The process initiates with selection of one or more performance objectives.
Each performance objective is a statement of the acceptable risk of incurring
damage or loss for identified earthquake hazards. Building
developers/owners, design professionals, and building officials will typically
participate in the selection of performance objectives, and may also consider
April 15, 2009
12 1: Introduction ATC58
the needs and desires of a wider group of stakeholders. These can include
prospective tenants, lenders, insurers and others who have impact on a
building’s value, but who generally do not have an opportunity to participate
in the design process directly.
Figure 11 Performancebased design flow diagram
Next, design professionals must develop a preliminary design to a sufficient
level of detail to allow determination of the building’s performance
characteristics. For new buildings, this will include, as a minimum,
identification of: (1) the location and characteristics of the site; (2) building
size, configuration and occupancy; (3) type, location and character of
finishes and nonstructural systems; and (4) estimates of the strength, stiffness
and ductility of the structural system. In the case of existing buildings, these
characteristics are already defined, and it is only necessary to determine what
they are, and then define preliminary concepts for retrofit measures, if
needed.
Performance assessment (the shaded box in Figure 11) is the process used to
determine if a design is capable of achieving the desired performance
objectives. In this step, which is the subject of these Guidelines, the engineer
conducts a series of simulations (analyses) to predict the building’s response
when subjected to the earthquake hazards identified as part of the
performance objectives. Following performance assessment, the engineer
compares the predicted performance with the desired performance. If the
assessed performance matches or exceeds the stated performance objectives,
April 15, 2009
ATC58 1: Introduction 13
the design is adequate and the project can be completed. If the assessed
performance does not meet the performance objectives, the engineer revises
the design, or alters the performance objectives in an iterative process, until
the assessed performance meets the desired objectives.
1.3 Guideline Uses
These Guidelines can be used by engineers for performancebased seismic
design or retrofit of buildings, as described in Section 1.2. The methodology
and procedures presented herein can also be used in the following ways:
• Engineers can determine the probable performance of buildings (e.g.,
probable maximum loss, business continuity, life safety protection) in
support of real estate transactions and occupancy decisions.
• Building product suppliers can determine the seismic performance
capability of their products.
• Building code developers can determine the performance consequences
of building code requirements to guide the improvement of these
requirements
• Researchers can identify areas where additional building performance
research is needed.
• Educators can use this information as instructional materials in
earthquake engineering curricula.
• Software developers can develop applications that implement the
performance assessment methodology either coupled with, or
independent of, structural analyses
1.4 Application
These Guidelines present a general methodology for seismic performance
assessment of individual buildings and one set of recommended procedures
to implement this methodology. Nothing contained in these Guidelines is
intended to prevent or discourage the use of alternative procedures that
produce unbiased assessments of probable building performance, and that
appropriately consider and portray the uncertainties inherent in the
assessment of future building performance.
The methodology and recommended procedures presented herein can be
applied to the performance assessment of any building type, regardless of
age, construction type or occupancy. However, application requires basic
data on structural and nonstructural component fragilities and the
consequences of component damage in terms of potential casualties, repair
April 15, 2009
14 1: Introduction ATC58
costs, and downtime. The data required is dependent on structural system
type; the specific details of construction; the type, location and means of
installation of nonstructural components and systems; and building
occupancy and use. Sources of such data can include laboratory testing of
individual building components, analytical evaluation, statistical information
on the actual performance of buildings in past earthquakes, and expert
judgment.
At the present time, the availability of such data is limited. This report
includes a set of such data gathered by the project development team. This
data set is sufficient to allow performance assessment of buildings with the
following structural systems:
• light wood frame,
• momentresisting steel frame,
• momentresisting concrete frame,
• braced steel frame,
• concrete shear wall, and
• masonry shear wall construction;
and conforming to one of the following occupancies:
• commercial office,
• education (k12),
• general hospital,
• hotel/motel,
• multifamily residential,
• research,
• retail, and
• warehouse uses.
In order to perform assessments of buildings of other construction types or
occupancies, it may be necessary to obtain additional fragility and
consequence data. These Guidelines include recommended procedures for
deriving such data.
1.5 Performance Calculations
Companion software, termed the Performance Assessment Calculation Tool
(PACT) accompanies this report. PACT includes databases of typical
April 15, 2009
ATC58 1: Introduction 15
component and contents inventories, fragility and consequence data for
buildings of the construction types and occupancies described above. PACT
also automates the repetitive calculations necessary to assess probable
building performance as described later in this report. Individual users can
enhance, expand and modify the PACT databases using routines embedded
within the PACT software. In addition, for individual building assessments,
users can modify default values obtained from the databases with building
specific values. PACT is provided as a developmental tool and to facilitate
implementation of the recommended procedures. As currently planned,
PACT will not be maintained by either the Applied Technology Council or
the Federal Emergency Management Agency, however, its code is open
source and available for use as the basis of either individually maintained or
commercially developed software.
1.6 Guideline Organization
Chapter 2 introduces the building performance measures used in these
guidelines and the uncertainties associated with performance assessment. It
also describes some ways to use the information obtained from performance
assessments to guide decision making processes. Chapter 3 presents the
general performance assessment methodology. Chapters 4 through 8
introduce a set of recommended procedures to implement this methodology.
Chapter 4 provides an overview of these procedures. Chapter 5 presents
detailed information on representation of seismic hazards for use in these
procedures. Chapter 6 presents recommended methods of structural analysis
for performance assessment. Chapter 7 presents analytical procedures for
determining the collapse fragility of individual buildings. Chapter 8 presents
example applications of these procedures to representative buildings.
Appendices to this report present background information that may be useful
for some readers. Appendix A provides a basic tutorial on probability and
statistics and the types of probabilistic distributions used to represent
uncertainty in these performance assessment procedures; Appendix B lists
the types and default quantities of nonstructural components and contents
contained in the PACT databases for the several building occupancy types; as
well as the building population models for each of these occupancies;
Appendix C contains the structural system and Appendix D the nonstructural
component fragility specifications contained in the PACT databases;
Appendix E provides detailed information on seismic hazard evaluation and
attenuation relationships; Appendix F summarizes methods for development
of fragility functions for structural and nonstructural components; Appendix
G describes the mathematical procedure used by PACT to derive earthquake
April 15, 2009
16 1: Introduction ATC58
response realizations; and Appendix H provides background on assessment
of residual story drift.
A separate PACT user’s manual provides instructions on running the
electronic performance assessment calculation tools (PACT).
1.7 Limitations
This report provides a general methodology and recommended procedures to
assess the probable performance of individual buildings when subjected to
future earthquake shaking. Specifically, the methodology assesses the
likelihood that building structural and nonstructural components and systems
will be damaged by earthquake shaking, and estimates the potential
casualties, repair costs, and interruption of beneficial building occupancy that
could occur as a result of such damage.
Earthquake shaking can cause other significant effects including loss of
offsite power, water and sewage, initiation of fires, inundation, and release of
hazardous materials. Similarly, earthquake effects other than ground shaking
including ground fault rupture, landslide, liquefaction, seiches and tsunamis,
and lateral spreading can significantly affect building performance.
While these effects can have significant impact on earthquake losses and the
general methodology presented in this report could be used to assess these
effects, assessment of these losses is beyond the scope of the recommended
procedures contained in this report. When conducting seismic performance
assessments of buildings using these procedures engineers should, as a
minimum perform qualitative evaluation of these other effects, and, if these
effects appear significant, report this.
Assessment of the probable performance of a building in future earthquakes
inherently entails significant uncertainty. The methodology and procedures
presented herein use state of art methodologies to assess future building
earthquake performance with explicit consideration of these uncertainties.
Regardless, it is possible that the performance of individual buildings in
actual earthquakes may either be better or worse than indicated by
assessments conducted in accordance with the procedures presented herein.
Further the accuracy of performance assessments depends in large part on
data and calculations generated by individual users. Neither the Federal
Emergency Management Agency, the Applied Technology Council, their
employees, nor their consultants, present any warranty expressed or implied
as to the accuracy of performance assessments made using these procedures.
April 15, 2009
ATC58 2: Performance Measures 21
Chapter 2
Performance Measures
2.1 Introduction
The methodology presented in these Guidelines uses three fundamental
measures of earthquake performance:
• casualties – defined as deaths and serious injuries that would normally
require hospilization
• repair cost – defined as the cost of repairing or replacing damaged
buildings, and their contents
• downtime – defined as the period of time during which a building is not
useable for its intended purpose or function as a result of earthquake
induced damage
It is impossible to precisely quantify the losses that a building will
experience in future earthquakes, before events actually occur. The
methodology, therefore, expresses performance using probabilistic measures.
2.2 Factors Affecting Performance
Building earthquake performance is dependent on:
• the intensity of ground shaking and other seismic hazards at the building
site;
• the manner in which the building responds to the ground shaking and
other hazards, and the amount of force, deformation, acceleration and
velocity experienced by the various structural and nonstructural
components, contents and occupants
• the vulnerabiltiy of the building and its systems to damage;
• the number of people, their location in the building, and the type,
location and amount of contents present when the earthquake occurs;
• the actions people take in response to the damage that occurs.
2.3 Uncertainty in Performance Assessment
Each of the individual factors that affect seismic performance is difficult, if
not impossible, to predict precisely. For example, it is not presently possible
to determine which fault the next earthquake will originate on, where along
the fault the surface the rupture will initiate, or what magnitude it will be, let
April 15, 2009
22 2: Performance Measures ATC58
alone the direction in which the fault rupture will propagate or the exact
character of the ground shaking that will result at a particular site. Similarly,
it is not possible to predict the day of the week or time of day at which the
earthquake will occur, which tenants and people will be present in the
building, what contents and furnishings they may have within the building,
what condition the building will be in, or the economic conditions that will
prevail at the time of the earthquake. The result of these uncertainties and
many others is that it is not possible to predict precisely the losses that will
occur, whether casualties, repair costs or downtime.
Although these uncertainties make it impossible to make a precise
assessment of building performance, in terms of a specific number of
casualties, repair cost or time of occupancy interruption, it is possible to
assess these performance measures in the form of probability distributions
that indicate the probability that losses of specified or larger magnitude will
be incurred. The methodology and procedures presented in this report
describe the means of determining these loss probability distributions.
Each step in the performance assessment process entails uncertainty. The
primary steps and the associated uncertainties are defined below.
Building Definition. In order to assess a building’s future performance it is
necessary to completely define the state the building will be in at the time the
earthquake occurs. This includes the configuration, strength, and detailing of
its structural elements; the location, make, model number, and means of
installation of each of the nonstructural components and systems; the
location, value and means of installation of all the building’s contents, and
the location of all people within the building.
Even if the drawings and specifications for a building are available, these
only define the specified, rather than actual building construction. Structural
drawings define the intended strength and ductility of the structural system.
Variability in material strength, workmanship, inspection, and condition can
result in a structure that is either more or less vulnerable to earthquake
effects. Similarly, nonstructural components will be described on the
drawings as to type and size, however, the make and model number of
individual items likely will not be described. Further, these components may
be anchored and braced to the structure or not. Even if they are initially
installed with appropriate bracing and anchorage, over the years, as tenants
make modifications to these systems, they may either improve or reduce the
effectiveness of these measures. As different tenants move in and out of a
building, they will change the configuration of ceilings, lighting systems,
partitions, and other features. Some tenants will have high value contents
April 15, 2009
ATC58 2: Performance Measures 23
and others will not. Some tenants will have persons present in a building 24
hours a day, seven days a week, while other buildings will be occupied only
a few hours a week.
Uncertainty in building definition can be reduced by obtaining a set of
updated drawings and specifications for the building, performing surveys to
assure that the inventory of components and their means of installation are
understood; performing materials testing to quantify the strength of structural
materials, and interviewing tenants to determine their patterns of occupancy
and the value of their contents.
Intensity of Earthquake Effects. The more intense the earthquake effects at a
site, the more damage and loss is likely to occur. These Guidelines focus
principally on earthquake performance associated with ground shaking,
though other earthquake effects including landslide and liquefaction could
also be considered. Earth scientists and geotechnical engineers use
attenuation relationships to predict the intensity of ground shaking at a site
from future earthquakes. These attenuation relationships predict the
amplitude of spectral response acceleration at various structural response
periods, as a function of earthquake magnitude, fault and rupture
characteristics, distance of the rupture surface from the site, and the geologic
and soil conditions along the path of travel of earthquake energy from the
fault to the site and at the site and other factors. Seismologists develop
attenuation relationships by performing statistical analysis on data sets of
ground motion recordings obtained in past earthquakes with known rupture
characteristics, at sites having known geotechnical conditions and distance
from the fault ruptures. The factors used in the statistical analysis do not
correlate perfectly with the recorded motions and as a result, even if the exact
magnitude, fault rupture type, distance, and soil conditions associated with a
specific site and earthquake are known, actual ground motions recorded at a
particular site may be somewhat more or less intense than predicted by the
statistical fit to the data. To account for this, modern attenuation
relationships incorporate an error term that indicates the possible variability
of actual motions from those predicted by the relationship. This form of
uncertainty cannot be reduced, except by improvement of the ground motion
prediction equations that reduce their inherent error. Appendix E provides
more information on this topic.
Structural Response. The amount of damage a structure will incur when it is
subject to a particular ground motion is, in part, a function of the building’s
response, that is, how much it displaces and how much stress and
deformation are induced in its various elements. This is a function of the
strength of the individual elements and their connections, the overall strength
April 15, 2009
24 2: Performance Measures ATC58
and stiffness of the structural system, its configuration, and detailing. It is
also quite dependent on the spectral shape of the shaking that affects the
structure and the exact character of the ground motion, as represented by an
accelerogram.
Engineers use structural analysis to predict the response of a structure to
ground shaking. To perform this analysis, it is necessary to build a
mathematical model that represents the structure’s important response
properties including stiffness, strength and hysteretic behavior. Since the
precise strength, stiffness and ductility of real structural elements in a
building are never known with a certainty, due to inherent variability in
materials and workmanship, the models used to predict structural response
typically incorporate errors that may either over predict or under predict the
building’s actual response.
Even with perfect models, which never exist, some methods of analysis are
better at predicting response than others. Nonlinear response history
analyses that use appropriate ground shaking records and models that
accurately represent the structure’s characteristics can provide excellent
estimates of response. Linear static procedures using coarse elastic spectra
as the basis for loading often provide inaccurate assessments of response.
Even the best models and most sophisticated analytical procedures cannot
produce accurate assessments of response if the characteristics of the ground
motion are not accurately represented in the loading. Thus, structural
analysis may either under or overpredict the true strength and deformation
demands experienced by structural elements.
Damage. These procedures predict the amount of damage that occurs to
structural and nonstructural elements using the value of peak demand
parameters, including acceleration, drift, plastic rotation and element force,
obtained from structural analysis. Laboratory data, and where such data s
unavailable, judgment is used to project the level of demand at which
damage of different types will occur. However, even if laboratory data is
available, test specimens seldom precisely represent the actual components in
a building or the boundary conditions that the building places on the
components; nor do loadings applied in the laboratory closely resemble the
actual shaking demands placed on building components during earthquakes.
As a result, even if the demands predicted by analysis are exact, we may over
or underpredict the amount of damage that occurs. When judgment, rather
than laboratory data is used to determine the demand levels at which damage
occurs, even greater is inaccuracy is likely.
April 15, 2009
ATC58 2: Performance Measures 25
Loss. The amount of loss that occurs as a result of damage is dependent on a
number of factors. These include:
• Casualties – the number of people present at the time of the earthquake,
their locations in the building, whether debris falls towards or away from
them, and the actions they take as the building shakes
• Repair Costs – the means used to repair the building, the efficiency of the
contractor, the availability of labor and materials and the contractor’s
desire for profit
• Downtime – whether the building is red or yellow tagged, by post
earthquake inspectors; the owner’s efficiency in retaining consultants and
contractors; whether or not the building is occupied while repairs are
made
The methodology and procedures presented herein attempt to account for
each of these uncertainties and others, and rigorously portray their effect on
the probable value of losses.
2.4 Types of Performance Assessment
The methodology and procedures can be used to develop three types of
performance assessment: intensitybased, scenariobased and timebased.
2.4.1 IntensityBased Assessments
Intensitybased assessments provide a distribution of probable losses, given
that the building experiences a specified intensity of shaking. Ground
shaking intensity is represented by a 5% damped, elastic acceleration
response spectrum. Intensity could also include representation of permanent
ground displacements produced by fault rupture, landslide, liquefaction, and
compaction/settlement, although procedures for doing this are not included
herein. This type of assessment could be used to answer such questions as:
• What is the probability that repair cost will be greater than $1million, if
the building experiences ground motion represented by a smoothed
spectrum with a peak ground acceleration of 0.5 g?
• How long is the building likely to be closed for occupancy if it
experiences ground shaking matching the design spectrum contained in
the building code?
• What is the probability of incurring one or more casualties, if the
building experiences a ground motion with intensity corresponding to the
maximum considered earthquake spectrum described by the building
code?
April 15, 2009
26 2: Performance Measures ATC58
The results of intensitybased performance assessment are cumulative
probability distributions that express the probability that loss will exceed
various values, given that the building experiences a particular intensity of
shaking. Figure 21 presents four such cumulative probability distributions
for repair cost for a hypothetical building. Each of the four curves plotted in
the figure represents the distribution of possible loss for different intensity
levels, labeled I1 through I4, where I1 represents the lowest intensity and I4
the highest. In the figure, the probability that total repair costs exceed a
specified value of total repair cost (trc) is plotted as a function of total repair
cost. For shaking intensity I4, the figure shows a 50% probability that repair
cost will exceed $1.8million and a 10% probability that repair cost will
exceed $3.5 million. For intensity I1, there is a 50% probability that repair
cost will exceed $0.6million and only a 10% chance it will exceed $0.9
million.
0 1 2 3 4 5 6 7 8 9 10
Total repair cost (trc), Million dollars
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
P
(
T
R
C
≥
t
r
c
)
I1
I2
I3
I4
Figure 21 Example cumulative probability loss distributions for a
hypothetical building at four ground motion intensities
The results of intensitybased performance assessments can also be presented
as complementary distribution curves. Complementary distributions present
the probability that loss will be less than a given value. Figure 22 plots the
same data shown in Figure 21 except in complementary form.
April 15, 2009
ATC58 2: Performance Measures 27
0 1 2 3 4 5 6 7 8 9 10
Total repair cost (trc), Million dollars
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
P
(
T
R
C
≤
t
r
c
)
I1
I2
I3
I4
Figure 22 Example complementary cumulative probability distributions for
loss at four ground motion intensities
2.4.2 ScenarioBased Assessments
Scenariobased performance assessments provide distributions of possible
loss, given that the building experiences a specific earthquake scenario,
defined as a combination of earthquake magnitude and distance of the
building from the fault on which the earthquake occurs. This type of
assessment can be used to answer the following types of questions:
• What is the probable maximum loss (repair cost with a 10% chance of
exceedance) if there is a magnitude 6.5 earthquake on a fault located 10
miles from the building site?
• What is the likelihood of casualties in a building, if we experience a
repeat of a historic event (e.g. 181112 New Madrid, 1964 Anchorage,
1906 San Francisco)
Scenario assessments may be useful for decision makers with buildings
located close to one or more known active faults. Scenariobased assessments
are very similar to intensitybased assessments except that uncertainty in the
earthquake intensity, given the scenario, is considered. The results of
scenariobased assessments are single cumulative or complementary loss
distributions, such as one of the loss curves in either Figure 21 or Figure 2
2. Such curves show the probability that loss will either be more or less than
different values, given that the building is subjected to the effects of the
particular scenario earthquake.
April 15, 2009
28 2: Performance Measures ATC58
2.4.3 TimeBased Assessments
Timebased performance assessments provide a distribution of probable
earthquake loss, considering all earthquakes that may affect the building in a
given time period, and the probability of occurrence of each. Timebased
assessments can be used to answer the following types of questions:
• What is the probability that my building will experience repair costs with
a net present value exceeding $3,000,000 in the next 30 years?
• What is the chance, during any year, of incurring interruptions of
occupancy in my building that exceed 30 days?
• What is the probability of having at least one earthquakecaused casualty
in my building over a fiftyyear period?
For timebased assessments earthquakeintensity is described by a seismic
hazard curve, which plots the relationship between earthquake intensity, e,
and the mean annual probability of exceedance of e, ( ) e λ . To form time
based assessments, a series of intensitybased assessments are performed at
for a series of intensities of earthquake shaking that span the intensity range
of interest and are then integrated (summed) over the hazard curve to
construct an annualized loss curve of the type shown in Figure 23. This
curve shows for example, that for the hypothetical building, there is
approximately a 0.2% chance per year that an earthquake will produce repair
costs exceeding $1.5 million.
0 0.5 1 1.5 2 2.5 3 3.5 4
Total repair cost (trc), Million dollars
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
A
n
n
u
a
l
r
a
t
e
o
f
e
x
c
e
e
d
i
n
g
t
r
c
Figure 23 Distribution of mean annual total repair cost
Using such curves, it is possible to compute a mean annual loss, sometimes
called an average annualized loss, by integrating the area under the loss
April 15, 2009
ATC58 2: Performance Measures 29
curve. For this example, the annualized loss is $37,900. While it is not
actually expected that an earthquake producing $37,900 of loss will occur
each year, in theory, if the owner of the building could selfinsure, by placing
this amount of money in an interest bearing account each year, over a very
long period of time, he should be able to pay for any actual earthquake repair
costs using the money in this account. In essence, the annualized loss for
repair costs represents the premium that one should be willing to pay for an
insurance policy. In reality, insurance companies will not charge this amount
for insurance premiums. This is because they rarely actually calculate the
annualized loss for a building and even if they did, can spread their risks over
a large number of buildings, undercharging for some policies, and over
charging for others. Also, insurance companies must cover their
administrative costs and want to make a profit, and must operate in a
competitive environment that affects the prices they can charge.
Nevertheless, annualized loss can be a valuable tool to assess the value of
insurance as well as other uses. Section 2.5 illustrates additional uses for the
data provided by performance assessments.
2.5 Use of Loss Distributions in Decisionmaking
A diverse group of stakeholders and decisionmakers have impact on
selection of the appropriate seismic performance objectives for both new and
existing buildings. These include owners/developers, building officials,
tenants, lenders, insurers and design professionals. Different decision
makers may use very different decision processes to determine acceptable
building performance. The loss distributions generated by these performance
assessment procedures can be used in different ways to satisfy the needs of
these different decisionmakers.
2.5.1 Typical Building Performance
Many building officials, owners/developers and design professionals will
choose performance equivalent to that expected of a “typical building”
designed in conformance with the building code as the minimum acceptable
performance objective for a performancebased design. Life safety protection
is often the primary concern. In the past, life safety was often considered an
absolute goal, with zero tolerance for life endangerment, regardless of how
low the risk. In reality, no building can be designed with zero risk of
earthquakeinduced life loss. In order to use the procedures contained in
these Guidelines to obtain equivalent performance to that attainable by
typical codedesigned buildings, it is necessary to determine the risk of life
loss for such buildings. Recently, in an effort associated with development
of national seismic hazard maps for use in the building codes, the Building
April 15, 2009
210 2: Performance Measures ATC58
Seismic Safety Council’s Seismic Design Procedures Group set a lifesafety
related goal for ordinary buildings conforming to Occupancy Category II as
defined in ASCE 705. This goal, relates to the probability of building
collapse as opposed to the probability of life loss. The Seismic Design
Procedures Group suggests that a conditional probability of collapse of 10%
given that the building is subjected to Maximum Considered Earthquake
shaking is an acceptable goal. For most sites in the United States, Maximum
Considered Earthquake shaking has an annual frequency of exceedance of 1
in 2,500 or 4x10
4
per year. When integrated with the hazard curve, this
typically results in an annual probability of collapse of approximately 0.001
per year.
The procedures used in this guideline can be used to determine the annual
probability of collapse for an individual building design, which can then be
compared with the values suggested by the Building Seismic Safety Council
as a means of demonstrating compliance with the capability of typical
buildings. However, the procedures presented herein can also consider the
annual frequency of casualty generation, and in the process, provide more
collapse resistant design for buildings with large occupancies, as intended by
the Occupancy Group requirements of the building code. In order to do this,
it will be necessary to determine the risk of life loss in typical code
conforming buildings of different occupancies. This task is beyond the scope
of this document, but could be performed by individual users or building
code developers.
2.5.2 Probable Maximum Loss
Decisions regarding the degree of initial investment in seismic protection for
a new building and the appropriate level of retrofit for existing buildings can
be made with the aid of cumulative loss distributions or loss functions.
Figure 24 presents a sample loss function for a hypothetical building. It plots
the probability (yaxis) that the costs associated with postearthquake repair
and/or replacement will be less than a dollar amount as function of the dollar
amount (xaxis). This particular loss curve presents the probability of loss
for a specific intensity of shaking with a return period of 475 years. Scenario
loss curves present the probability of loss for a particular earthquake
magnitude and distance. Annualized loss curves present the frequency of
loss considering all earthquakes that might occur and the probability of each
such earthquake.
The shape of the loss function will vary as a function of earthquake intensity
and the vulnerability of the building to damage. Reducing the vulnerability of
April 15, 2009
ATC58 2: Performance Measures 211
a new or retrofitted building will shift the curve in Figure 24 to the left, such
that there is reduced risk of incurring loss.
0 1 2 3 4 5 6
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
DV (capital loss, $M)
P
(
l
o
s
s
e
s
<
D
V

I
M
=
4
7
5
y
r
.
)
0 1 2 3 4 5 6
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
DV (capital loss, $M)
P
(
l
o
s
s
e
s
<
D
V

I
M
=
4
7
5
y
r
.
)
Median loss
= $1.1M
Figure 24 Cumulative loss function identifying a median (50th percentile) loss
Once a loss function is established for a given building, intensity of
earthquake shaking, scenario or time period, and performance measure
(repair costs, downtime, casualties), it can be used to make quantitative
statements regarding loss and to make decisions regarding acceptable levels
of loss. Figure 24 identifies a median loss for the hypothetical building of
$1.1million, where there is a 50% chance that the loss will be higher than the
median value and 50% change that it will be lower
Lenders often use measures of seismic risk associated with mortgage loans
that are based on such loss distributions. Two common measures used by
lenders are the Scenario Expected Loss (SEL) and Scenario Upper Bound
Loss (SUL). The SEL is the mean loss obtained from an intensitybased
assessment, typically for 475year earthquake shaking. Appendix A provides
procedures for determining the mean loss from a loss curve represented as a
lognormal distribution. The SUL is obtained from an intensitybased loss
assessment as that loss having a 10% chance of exceedance (90% chance of
nonexceedance).
When making mortgage loans, lenders commonly consider the assessed
value of SEL and SUL presented in an engineering report. Many lenders will
not offer loans for properties with excessive values of SEL and SUL and
some will make such loans, but require that the owners carry insurance to
protect the lender’s investment. As a result, many owners of existing
buildings desire seismic upgrades that will achieve specific values for these
loss estimates that will facilitate obtaining mortgage loans at attractive rates
April 15, 2009
212 2: Performance Measures ATC58
and terms. Although new buildings are generally assumed to be capable of
obtaining acceptable SEL and SUL some developers of new buildings are
also interested in design intended to target specific SEL and SUL values.
The procedures contained in these Guidelines provide a direct means of
designing for such performance.
2.5.3 CostBenefit Analysis
In addition to its use in calculating insurance premiums, described in an
earlier section, annualized loss is a potentially powerful metric for use in
performancebased design. Figure 25 presents two loss curves for a
hypothetical existing building. One curve shows the loss distribution before
the building is retrofitted and the other after retrofit. Before retrofit, the
average annualized loss, calculated as the area under the loss distribution
curve, is $156,000 per year; after retrofit the annualized average loss is
$96,000 per year. If the building is retrofitted, the average annualized loss is
reduced by $60,000 per year, which is a net economic benefit to the owner.
Figure 25 Annualized loss before and after proposed retrofit
If the owner anticipated that he would retain the building for a period of 10
years, at a discount rate of 7%, the net present value to the owner, of the
reduction in annualized average loss of $60,000 per year would be $450,000.
As long as this net present value is greater than the retrofit cost, retrofit
would have net benefit. It should be noted that this neglects the fact that the
retrofitted building would have increased sales value, and that additional
economic benefit would result from reductions in annualized downtime and
enhanced occupant safety.
April 15, 2009
ATC58 2: Performance Measures 213
Another way of using this information is in the form of an internal rate of
return on the investment. Many corporations and investors will not make an
investment unless they can achieve a target rate of return on their money, and
will use this metric as a means of deciding between alternatives. Rate of
return is a simple expression of the interest rate that can be obtained from an
investment. For a long duration investment, the return on investment can be
approximated as the annualized benefit of the investment divided by the net
present value or cost. If the cost of retrofit is $400,000, the return on
investment in this illustration would be the annual benefit of the investment,
$60,000, divided by the cost, $400,000, or about 15%.
In addition to assisting with decisions to retrofit existing buildings, cost
benefit analyses of this type can be used to determine appropriate levels of
retrofit. These analyses can also be used to provide information on whether a
new building should be designed for enhanced performance.
April 15, 2009
ATC58 3: General Methodology 31
Chapter 3
General Methodology
3.1 Introduction
This chapter describes the general building seismic performance assessment
methodology. Figure 31 identifies the five basic steps in the methodology,
each of which, are described in the sections that follow.
Figure 31 General Seismic Performance Assessment Procedure
3.2 Step 1 – Assemble Building Performance Model
The building performance model is an organized collection of all that data
necessary to define the building assets at risk. This data includes definition
of the:
• structural systems and elements – in sufficient detail to quantify their
location within the building, and the demands they will experience
during response to earthquake shaking; their vulnerability to damage
caused by earthquakeinduced deformations and forces; the
consequences of this damage, in terms of generation of lifethreatening
debris, influence on postearthquake building occupancy and necessary
repair actions.
April 15, 2009
32 3: General Methodology ATC58
• nonstructural systems, components and contents– in sufficient detail to
quantify their location within the building and the demands they will
experience during building earthquake response; their method of
installation as it effects their vulnerability to damage; the consequences
of this damage, in terms of generation of lifethreatening debris,
influence on postearthquake building occupancy and necessary repair
actions
• Occupancy – including the distribution of people within the building and
the variability of this distribution with time of day and day of the year.
Design professionals already assemble much of this data as part of the basic
building design process. For many years, design professionals collected such
data in the form of construction documents, including drawings and
specifications, produced by the individual design disciplines, separately
developed by architects, structural, mechanical, electrical, and plumbing
engineers. More recently, design professionals have begun to use electronic
databases termed Building Information Models to collect and organize this
data. Present generation Building Information Models contain nearly all of
the data necessary for performance assessment except definition of the
vulnerability of building systems and components to damage and the
potential consequences of this damage. Eventually, Building Information
Modeling systems may incorporate this additional data needed for seismic
performance assessment. For the present, however, the performance analyst
must separately collect this data and organized it into data bases for use in
the methodology.
3.3 Step 2 – Define Earthquake Hazards
Earthquake hazards can include ground shaking, ground fault rupture,
liquefaction, lateral spreading, land sliding, and tsunami inundation. Each of
these effects can occur with different levels of severity, or intensity, ranging
from imperceptible to highly damaging. Even in the most seismically active
regions of the world, earthquakes are infrequent events. Low intensity
earthquake effects, having little ability to cause damage or loss may affect a
building occasionally, perhaps once every ten years or so, while intense
earthquakes effects having great potential to cause damage and loss occur
rarely, often with hundreds of years separating the occurrence of truly
damaging effects at a site. Earthquake hazard is a quantification of the
intensity of these earthquake effects and the probability that effects of given
intensity will occur at a particular site.
The procedures presented in this report only consider ground shaking
hazards. However, the methodology and procedures can be extended to
April 15, 2009
ATC58 3: General Methodology 33
include consideration of the other hazards including permanent ground
deformation, inundation, etc.
In this methodology, ground shaking hazards are specified in different ways,
depending on the type of assessment desired and the type of structural
analysis used to quantify the building’s earthquake response.
For intensitybased assessments, the methodology takes as a given, that
earthquake shaking of a particular intensity will occur at the building site.
Depending on the type of structural analysis that will be used, ground
shaking intensity is characterized in the form of an elastic acceleration
response spectrum, discrete spectral response accelerations extracted from
this spectrum, or suites of ground motion records that have been selected and
scaled to this spectrum.
For scenariobased assessments, shaking hazards are defined by a median
acceleration response spectrum and perioddependant dispersions obtained
from an attenuation relationship, or by a suite of ground motion records that
have been selected and scaled to represent this spectrum and dispersion.
For timebased assessments ground shaking hazards are characterized by a
series of mean seismic hazard curves (or a median hazard curve and
dispersion), together with representative spectra and scaled ground motion
records. Each hazard curve is a plot of the probability of exceedance of a
particular ground motion intensity parameter, such as spectral response
acceleration at a particular period as a function of intensity.
3.4 Step 3 – Simulate Building Response
In the future, engineers may have access to modeling and analysis tools that
will be able to directly predict the damage resulting from a structure’s
response to a ground motion. At present, however, this capability does not
exist. Therefore, at present, engineers must construct structural models and
conduct structural analyses that predict the value of key building response
parameters (demands) that are used in later steps of the methodology to
predict damage to structural and nonstructural components and systems. Key
response parameters used for this purpose include peak floor accelerations,
peak story drifts, residual drift, and individual peak member forces and
deformations.
Since a building’s structural characteristics and the characteristics of
earthquake shaking are both highly uncertain, it is impossible to calculate
precise (deterministic) values of these demands. Therefore, in this
methodology, we predict a statistical distribution of the likely values of these
April 15, 2009
34 3: General Methodology ATC58
demands, considering the possible variation in earthquake intensity, ground
motion characteristics, and the inherent structural modeling uncertainty
(associated with uncertainty in the structure’s properties and the extent to
which the idealized analysis model captures these properties). The effect of
these uncertainties can be assessed in a variety of ways, including Monte
Carlo and similar statistical simulation methods where the ground motion
characteristics, the strength and hysteretic properties of the structural
elements, damping, and other important parameters are varied within
plausible bounds to determine their effects on the demands.
Alternatively, the methodology permits the calculation of median or best
estimate (mean) estimates of strength and hysteretic properties to predict
median values of the demands. These median estimates of demand are then
converted into distributions of demand based on the assessed variability of
each of the random parameters, for example damping or yield strength of an
element, and the likely effect of this variability on structural response. This
alternate method, albeit indirect and lacking the precision of analysis using
appropriate distributions of strength and hysteretic properties, is
computationally efficient and better suited to the analysis tools available at
this time.
3.5 Step 4 – Develop Collapse Fragility
Collapse fragilities are mathematical functions that indicate the probability
that one or more modes of structural collapse may occur, as a function of a
ground motion intensity parameter. Figure 32 plots a representative collapse
fragility for a hypothetical structure in which the probability of collapse is
plotted as a function of spectral acceleration at the fundamental period of the
building’s structural response. This figure portrays, for example, that if the
building experiences ground shaking with a spectral response acceleration at
the building’s fundamental mode of 0.5g there is a 14% change of collapse
and at 1.0g, there is a 70% chance of collapse.
April 15, 2009
ATC58 3: General Methodology 35
Figure 32 Representative collapse fragility for a hypothetical building
structure
The collapse fragility shown in Figure 32 takes the form of a lognormal
distribution. Such distributions are completely described by two values, a
median and dispersion. The median is that value of the predictive parameter,
in this case spectral response acceleration, at which there is a 50%
probability of an event’s occurrence, in this case structural collapse. The
median for the distribution shown in Figure 32 has a value of 0.8g.
Dispersion is a measure of the uncertainty as to how the probability of
occurrence varies with the value of the predictive parameter. In this figure,
the dispersion has a value of 0.45. Appendix A presents more information on
this and other forms of statistical distributions.
Collapse fragilities like that shown in Figure 32 can be developed in several
different ways. The most direct method involves the use of suites of non
linear structural analyses using ground motion records of varying intensity.
Alternatives include the use of statistics on the occurrence of collapse in
buildings of similar type, in past earthquakes and expert judgment.
Generally, the use of nonlinear analysis should result in a smaller dispersion
then the use of pastearthquake data, which in turn, should result in a smaller
dispersion than expert judgment. Chapter 7 of this document presents
analytical procedures that can be used to develop collapse fragilities.
In general, a structure may experience several modes of collapse, including
collapse that involves a portion of a single story or stories, an entire story or
stories, or the entire structure. The collapse fragility does not by itself
indicate the type of collapse that occurs, only the occurrence of collapse. In
addition to the collapse fragility, this methodology also requires the
development of descriptions of the possible collapse modes, and the
April 15, 2009
36 3: General Methodology ATC58
conditional probability of occurrence of each of these modes, given that
collapse occurs. While it is theoretically possible to develop data on the
possible modes of collapse and the probability of occurrence of the various
collapse modes using structural analysis, this will typically be done based on
a combination of analysis and expert judgment.
3.6 Step 5 – Damage and Loss Assessment
The fifth step consists of calculation of the distribution of damage to
structural and nonstructural building components using the demands obtained
from the structural analysis together with data on the building’s composition
and configuration obtained in Step 1. Each structural and nonstructural
component in a building will have a unique probability of sustaining damage
of different types in an earthquake, based on its construction characteristics,
location in the building and the demands imposed on it as a result of the
building’s response to earthquake shaking.
Given present limitations in modeling and analysis, two methods can
practically be used to develop these distributions of damage. The first of
these consists of performing a large number of analyses, using a large suite
of ground motions representing the particular intensity of motion. In this
approach, each analysis will produce a vector of peak response quantities
including floor accelerations, story drifts, and element strength and
deformation demands that can be used to assess the types of damage that will
occur. Alternatively, the statistical distribution of demands obtained using
the procedures described in Section 3.2.4 can be used to construct a series of
possible response states for the particular intensity of motion, termed
realizations. Each realization represents one possible set of peak response
quantities that is consistent with the particular intensity of motion. Appendix
G presents one set of mathematical procedures that can be used to generate
these realizations. Regardless of whether suites of analyses or statistical
procedures are used to generate demand sets, or realizations, thousands of
realizations are needed to predict the distribution of damage for a building
for a particular intensity of motion. The discussion below describes the
method of determining a building damage state for a single set of demands or
realization. These procedures must be repeated many times to determine the
distribution of building damage over many realizations. Figure 33 shows
the process used to assess damage, and then loss in each realization.
April 15, 2009
ATC58 3: General Methodology 37
Figure 33 Loss calculation flow chart
3.6.1 Collapse Determination
For each realization, an initial decision is made as to whether the building
has collapsed. This is done using the collapse fragility functions described in
the previous section. For each realization, a random number generation is
performed to determine, using the collapse fragility if collapse has occurred.
For example, for evaluation of damage to the structure with the collapse
fragility shown in Figure 32, for an intensity of motion having a spectral
acceleration at the building’s fundamental period of 0.5g, we find from
Figure 32 that there is a 15% probability that the building will collapse. For
a given realization, we would generate a random number in the range of 1 to
100. If the number were 15 or less, we would judge that collapse had
occurred, and this would define the building’s damage state. If, on the other
hand, a random number greater than 15 were selected, this would indicate
that collapse had not occurred, and we would generate a set of demands,
using analysis, or statistical procedures like those in Appendix G.
3.6.2 Damage Assessment
The type and extent of damage that a component will incur, given that it
experiences a particular demand, is also uncertain. We use component
specific fragility functions that indicate the conditional probability of
incurring damage given a value of demand to determine a damage state for
each set of demands.
April 15, 2009
38 3: General Methodology ATC58
For each component that is vulnerable to damage, we must define a set of
one or more possible damage states. Each damage state is defined by a
unique set of consequences consisting of a particular repair action that is
required, a potential impact on occupancy or on casualties, or a combination
of these. Not all damage states will have meaning for all of these potential
consequences. Damage states that are meaningful for one performance
measure (e.g., repair method) may not be useful for another performance
measure (e.g. casualties). Each damage state will represent a unique
consequence in terms of one or more of these performance measures. For
example, for a hypothetical exterior cladding component, meaningful
damage states could include:
• No damage (and no consequences with regard to repair, downtime or
casualties)
• Cracking of sealant joints, permitting moisture and/or air intrusion. Such
damage will have no consequences with regard to building appearance,
safety for occupancy or downtime, however, over the long term, will
present building maintenance issues, and therefore will require repair
consisting of replacing sealant joints. This will results in small repair
costs and no downtime or casualties.
• Damage consisting of visible cracking of the panels. To repair this
unsightly damage, the cladding must be removed from the building and
replaced. This damage will likely have more severe cost consequences
and may affect occupancy while the repairs are made but again, will not
pose a safety hazard during the earthquake.
• Damage consisting of panel connection failure, and pieces of the
cladding falling off the building. This damage will likely have similar
repair consequences as that described previously (replacement of the
panels) but will also have potential life safety impacts (when the
cladding falls from the building) and will have severe occupancy impacts
as the building may be deemed unfit for occupancy until the cladding is
repaired.
The probabilities of incurring such damage states as a function of demand are
defined by fragility curves, similar to the collapse fragility introduced in
Section 3.2.5. Fragility curves are cumulative probability distributions that
indicate the probability that a component will be damaged to a given damage
state or a more severe one, as a function of a particular demand.
A fragility curve is required for each damage state. For illustrative purposes,
fragility curves are introduced here using the repair cost performance
measure, though similar concepts apply to the other performance measures as
April 15, 2009
ATC58 3: General Methodology 39
well. Each damage state for repair cost is defined in terms the degree or
scope of repair that is required. Real damage generally occurs as a
continuum and not as a series of discrete states. For example, consider
damage to a steel beam measured using the amplitude of flange local
buckling. The amplitude of buckling is a continuous function of beam
deformation.
The cost for repair of this damage, however, is not a continuous function of
flange buckling amplitude. From the perspective of repair cost, it may not
matter if the buckling amplitude is 1/4 inch or 3/8inch, since the repairs in
either case will be very similar, and the costs nearly identical. Conversely,
modest increases in the level of damage can trigger large increments in
construction activity and cost. For example, buckling amplitude of 1/16inch
may not require repairs, while an amplitude of 1/8inch may trigger heat
straightening of the beam flange, which is a repair activity requiring
substantial work and cost.
Figure 33 presents a sample family of fragility curves for a hypothetical
special steel moment frame beamcolumn connection. For this example, three
damage states: DS
1
, DS
2
, and DS
3
, are used, where the damage states are
defined using discrete and well separated (in terms of cost) states of repair:
• DS
1
: Flange and web local buckling in the beam requiring heat
straightening of the buckled region.
• DS
2
: damage and lateraltorsional distortion of the beam in the hinge
region requiring partial replacement of the beam flange and web in the
hinge region, and corresponding construction work to other structural
and nonstructural components.
• DS
3
: Lowcycle fatigue fracture of the beam flanges in the hinge region
requiring replacement of a large length of beam in the distorted/fractured
region, and corresponding construction work to other structural and
nonstructural components.
April 15, 2009
310 3: General Methodology ATC58
0 1 2 3 4 5 6 7 8
Story drift (% story height)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
P
(
D
S
≥
D
S
i
)
DS1
DS2
DS3
Figure 34 Example family of fragility curves for special steel moment
frames
In this methodology, component fragility curves are represented by
lognormal distributions similar to those previously discussed in Section 3.5
for collapse fragility in Section 3.5. Each damage state fragility is defined by
a median value of the demand θ
DS
, and a dispersion β
DS
. The median value,
θ
DS
is that value of demand at which there is a 50% chance that a component
will enter this damage state. If a large number of components are subjected
to demand θ
DS
, and the performance of these components is uncorrelated,
half of the components will have entered the damage state and half will not
have. The dispersion, β
DS
is a measure of the scatter in demand values at
which similar components will enter the damage state.
The dispersion is associated solely with the onset of the associated damage as
a function of building response (i.e., demand) and is independent of the
uncertainty associated with the intensity of shaking or the prediction of
demand. As the value of β
DS
, increases the shape of the curve becomes
flatter, indicating a wider range of demand values over which there is
significant probability for the damage state to initiate. The dispersion reflects
variability in construction and material quality, the extent that the occurrence
of damage is totally dependent on a single demand parameter, and the
relative amount of knowledge or data on the response of the component.
For the family of fragility curves presented in Figure 34 above, for which
medians and dispersion for the three damage states are (3% drift, 0.35) for
DS
1
, (4% drift, 0.35) forDS
2
and (5% drift ,0.35) for DS
3
, the following
interpretations are possible:
April 15, 2009
ATC58 3: General Methodology 311
• At a story drift ratio of 4% the probability of no damage (neither Damage
State DS
1
, DS
2
, nor DS
3
have occurred) is 20% (=1.000.80), the
probability of damage in state DS
1
(assumed to be representative of
damage between curves DS
1
and DS
2
) is 30% (=0.800.50), the
probability of damage in state DS
2
is 24% (=0.500.26) and the
probability of damage in state DS
3
is 26% (=0.260).
• At a story drift ratio of 4%, the probability of being in damage state DS
1
or greater is 80% (equal to the sum of the probabilities of being in states
DS
1
, DS
2
, and DS
3
(30%+24%+26%).
Component fragility functions can be developed by laboratory testing, by
collecting data on the behavior of similar components in real earthquakes, by
analysis, by judgment, and by a combination of these methods. Appendix F
presents quantitative procedures for developing fragility functions and a
companion publication, FEMA 461 Interim Testing Protocols for
Determining the Seismic Performance Characteristics of Structural and
Nonstructural Components (FEMA, 2007), presents recommended
procedures for performance of laboratory tests intended to develop fragility
functions.
For each realization in which building collapse is not predicted, the demands
are used together with the component fragility functions for all components
that make up the building to determine the damage state for each component.
Taken together, the damage states for all of the components that comprise the
building define a building damage state.
The building damage state includes definition of the damage state for each
building component. For the purposes of repair cost calculations, the
building damage state includes a detailed description of the condition of the
building in terms of the required repairs. This description could be given to a
contractor to form the basis for an estimate of the costs and time necessary to
repair the building and replace its contents that have been damaged. The
damage state description can also be used to assess whether building
occupancy is possible, if casualties have occurred during the earthquake, or if
the building can be used for its intended purpose prior to repair. For the
purpose of casualty assessment, the damage state includes a description of
those areas of the building where falling debris or collapse has occurred, so
that the number of casualties can be assessed.
3.6.3 Loss Calculation
Losses are computed for each realization using the building damage states
developed in the previous step and a series of consequence functions.
April 15, 2009
312 3: General Methodology ATC58
Consequence functions are distributions of the likely consequences of a
building being damaged to a particular state in terms of repair cost, casualties
and downtime. Families of consequence functions are developed for each
performance measure (i.e., casualties, downtime, and repair cost). These
families will generally differ across types of buildings. The general functions
are complex and uncertain and must be simplified using approximate
procedures for practical implementation. Figure 35 presents a sample
consequence function for repair cost. For small quantities the unit cost is
constant at a maximum value. Beyond a certain quantity the cost diminishes
as the contractor can take advantage of economies of scale until a minimum
unit cost for large quantity repairs is reached. Since costs are subject to
uncertainty from market conditions, contractor bidding strategy, and other
factors, unit costs are assigned a median value (solid line in the figure) and
dispersion,
c
β . As with fragilities, a lognormal distribution of consequences
is assumed.
Contractors’ overhead and profit depend on the total amount of work and the
type of tradesmen and subcontractors required. The consequence functions
for repair cost use the building damage state to determine the need for
shoring, staging, finish protection, cleaning, and other general condition
costs; the costs associated with contractor overhead and profit and indirect
project costs including design services, fees and permits as well as the costs
of the actual labor and materials associated with the individual repairs
required.
Unit Cost, $
Quantity
Uncertainty, β
c
Max. cost
Min. cost
Min. quantity Min. quantity
Unit Cost, $
Quantity
Uncertainty, β
c
Max. cost
Min. cost
Min. quantity Min. quantity
Figure 35 Sample consequence function for cost of repair
3.7 Loss Aggregation
Each simulation of response or realization enables the development of a
building damage state and the calculation of a single value of the loss. This
section describes how these individual loss estimates are assembled into a
April 15, 2009
ATC58 3: General Methodology 313
distribution, similar to those shown in Figure 21 and 23. The sections
below describe the way in which this process is performed for intensity
based assessments, scenariobased assessments and timebased assessments,
respectively.
3.7.1 IntensityBased Assessments
The loss distribution is developed by repeating the realizationspecific
damage and loss calculation described in Section 3.2.6 many times and
sorting the losses in ascending or descending order to enable the calculation
of the probability that the total loss will be less than a specific value for the
given intensity of shaking. For example, if loss calculations are performed
for 1,000 realizations, and the realizations are assembled in ascending order
of repair costs computed, the repair cost with a 90% chance of exceedance
will the repair cost calculated for the realization with the 100
th
lowest cost.
3.7.2 ScenarioBased Assessments
Scenariobased assessments are performed using the same general procedure
described in the previous section for intensitybased assessments except that
uncertainty in intensity, given the scenario must be accounted for. This can
be performed by varying the intensity at which realizations are generated,
considering the uncertainty in ground motion intensity associated with the
selected attenuation relationship used to define intensity, or alternatively, by
adding the attenuation relationship uncertainty into the random selection
process used to generate attenuation demands, as described in Appendix G.
3.7.3 TimeBased Assessments
The product of a timebased assessment is a curve of the type shown in
Figure 23, which plots the total repair cost, probable number of casualties,
or days of downtime versus the annual rate of exceeding these losses. The
curve shown in Figure 23 can be constructed using the results of a series of
intensitybased assessments and an appropriate seismic hazard curve for the
site. Figure 36 shows a representative seismic hazard curve, where the
annual frequency of exceeding an earthquake intensity, ( ) e λ , is plotted
versus the earthquake intensity, e, where the typical earthquake intensity is
spectral acceleration (at the first mode period of the building).
April 15, 2009
314 3: General Methodology ATC58
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Earthquake intensity, e
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
A
n
n
u
a
l
f
r
e
q
u
e
n
c
y
o
f
e
x
c
e
e
d
a
n
c
e
,
λ
(
e
)
dλ(e)
de

e
I1
e
I1
e
I2
e
I3
e
I4
∆e
1
∆e
2
∆e
3
∆e
4
∆λ
1
Figure 36 Seismic hazard curve and timebased loss calculations
Hazard curves like that shown in Figure 36 are used together with Equation
(31) to calculate the annual probability that the loss L will exceed a value l:
( ) ( ) ( ) P L l P L l E e d e
λ
λ > = > =
∫
(31)
where the term
( )
P L l E e > = is the product of an intensity based assessment
for intensity e . Equation (31) will generally be solved by numerical
integration.
In order to perform this integration, the spectral range of interest is divided
into n intervals,
i
e ∆ . The spectral range of interest will typically range from
very low intensity that results in no damage, to very large intensity that
produces a high probability of collapse. The midpoint intensity in each
interval is
Ii
e , and the annual frequency of earthquake intensity in the range
i
e ∆ is
j
λ ∆ where the parameters
i
e ∆ ,
Ii
e and
j
λ ∆ are defined in Figure 35
for the sample hazard curve using n = 4. This small value for n is chosen for
clarity. In real timebased assessments, a much larger number, n, of
intensities will typically be necessary.
An intensitybased assessment is performed at each of the n midpoint
intensities,
1 I
e through
In
e , The number, n of intensities required to
implement this process will vary from structure to structure, and will depend
on the steepness of the hazard curve and the ability of the structure to survive
a wide range of ground shaking intensities. Earthquake intensity at intensity
1 I
e is assumed to represent all shaking in the interval
1
e ∆ . The product of the
n intensitybased assessments is n loss curves of the type shown in Figure 2
1. The annual probability of shaking of intensity
Ij
e ,
j
λ ∆ , is calculated
April 15, 2009
ATC58 3: General Methodology 315
directly from the seismic hazard curve. A sample calculation is shown in
Figure 35 for interval
1
e ∆ for which
1
0.054 λ ∆ = . Figure 23 is constructed
by: (1) multiplying each loss curve by the annual frequency of shaking in the
interval of earthquake intensity used to construct the loss curve; and (2)
summing the annual frequencies for a given value of the loss.
April 15, 2009
ATC58 4: Implementation 41
Chapter 4
Implementation
4.1 Introduction
This chapter presents an overview of a recommended set of procedures that
can be used to implement the general methodology described in Chapter 3.
These are not the only acceptable implementation procedures. Other
procedures that appropriately account for the uncertainties inherent in
assessment of ground motions, structural response to these ground motions,
damage to structural and nonstructural systems and components given this
response, and the consequences of this damage are acceptable.
These procedures address three types of performance assessments:
Intensitybased assessments – that indicate the probability of incurring
casualties, repair costs and occupancy interruption as a resulting from a
specific ground motion intensity, defined by a linear acceleration
response spectrum
Scenariobased assessments – that indicate the probability of incurring
casualties, repair costs and occupancy interruption as a result of a
specific earthquake event, defined by a magnitude and site distance
Timebased assessments – that indicate the probability of incurring
casualties, repair costs and occupancy interruption over a period of time,
considering all earthquakes that could occur in that time, and the
probability that they will occur.
The procedures employed for each type of assessment have differences, as
defined in the sections below and Chapters 5, 6 and 7. As described in
Chapter 3, and illustrated in Figure 31, the basic steps contained in these
procedures include the following steps:
1. Assemble the building performance model. This is a systematic and
quantitative description of the building components and systems at risk,
of damage as a result of building exposure to earthquake effects, the
damage states that can occur to these systems and components, and the
timedependent distribution of people exposed to injury in the building.
Section 4.2 provides guidelines on assembling building performance
models.
April 15, 2009
42 4: Implementation ATC58
2. Characterize the earthquake ground shaking hazard for which building
performance is to be assessed following the guidelines of Section 4.3 and
Chapter 5.
3. Perform structural analysis of the building for shaking as defined in step
2 following the guidelines of Section 4.4 and chapter 6.
4. Develop a collapsefragility and a collapse modeidentification for the
building following the guidelines of Section 4.5 and chapter 7.
5. Based on the results of the structural analysis, form a series of
realizations. Each realization is one possible outcome of the building’s
response to earthquake shaking, developed as described in Section 4.6.
6. For each realization, determine:
o the damage sustained by the building as described in Section 4.7
o the repair/replacement costs per Section 4.8, the occupancy
interruption time per Section 4.9 and the casualties, per Section 4.10
7. For each of the individual performance measures (repair/replacement
costs, down time, and casualties), order the realizations from least effects
to most severe effects and form a loss distribution, per Section 4.11.
As noted in Chapter 1, and described below, companion software, the
Performance Assessment Calculation Tool (PACT) performs many of these
tasks.
4.2 Building Performance Model
The building performance model is a complete and organized description of:
the building assets (structural and nonstructural components, contents
and people),
the location of each of these assets within the building
the types of damage that each can incur
The building performance model is subcategorized into three separate groups
termed occupancies, fragility groups and performance groups. The
subsections below describe each of these.
4.2.1 Occupancies
As used in this methodology, occupancy is used to categorize the primary
use of a building, or portion of a building. The primary use of occupancy is
to define the building’s population model, which is the number of people
present, at different times of the day and different days of the year for use in
April 15, 2009
ATC58 4: Implementation 43
assessing potential casualties. A secondary use of occupancies is to define
the typical quantities (termed normative quantities) of nonstructural building
components and contents, such as desks, file cabinets, partitions, etc. present
in the building. Appendix B presents default population models and
normative quantities for the following occupancies:
Commercial office
Education (K12), with subcategories of Elementary, Middle and High
School classroom buildings
Healthcare – general hospital
Hospitality
Multiunit Residential
Research Laboratories
Retail (shopping malls)
Warehouse
PACT contains the same population model and normative quantity data
shown in Appendix B to facilitate assignment of this data to buildings. It is
possible to assign different occupancies to different portions of a building, as
well as to change the data, based on buildingspecific information. It is also
possible to create new occupancies with unique population models and
normative quantities. Section 4.2.2 describes the format of the population
models and Section 4.2.3, the logic for assignment of normative quantities.
4.2.2 Population Models
Building population models defines the number of people per 1,000 square
feet of building floor space. This is defined in terms of the peak population,
that is the maximum number of people likely to be present, and the fraction
of this peak population that is likely to be present based on:
Time of day
Day of the week (weekday or weekend)
Month of the year.
From the model it is possible to determine the Equivalent Continuous
Occupancy (ECO). This is a time weighted average population. It represents
the number of persons on average, throughout the year. Using the peak
population, it is possible to generate a “worstcase” estimate of casualties,
assuming that the shaking occurs during peak occupancy. Using the ECO
April 15, 2009
44 4: Implementation ATC58
estimate it is possible to estimate the mean casualties, considering that
earthquake occurrence is random in time.
Table 41 below presents the peak populations for each of the above
occupancies and also indicates the time of day during which these peak
populations are expected to occur.
Table 41 Recommended Default Peak Population Models
Occupancy Description Peak Population
Model (Occupants
per 1000 sq. ft.)
Peak Population Model
 Time of Day
Commercial office 4.0 Daytime (3 pm)
Education (k12): Elementary
Schools
14.0 Daytime
Education (k12): Middle
Schools
14.0 Daytime
Education (k12): High
Schools
12.0 Daytime
Healthcare 5.0 Daytime (3 pm)
Hospitality 2.5 Nighttime (3 am)
Multiunit residential 3.1 Nighttime (3 am)
Research 3.0 Daytime (3 pm)
Retail 6.0 Daytime (5 pm)
Warehouse 1.0 Daytime (3 pm)
Building population varies with time of day, day of the week, and month of
the year. Population patterns reflect time of day (reflecting hours of
operations, lunch time fluctuations, etc.), day of week (weekdays vs.
weekends), and monthly population variations (due to building closures for
holidays, extended breaks, etc.). Figure 41 plots the timedependent
variation of population, as a percentage of peak population, for office
occupancy. Appendix B contains similar data for the other occupancies
shown above.
April 15, 2009
ATC58 4: Implementation 45
Figure 41 Graph illustrating the percent of peak occupancy, by hour and
day present
4.2.3 Normative Quantities
All structural and nonstructural components and contents that can be
damaged by building response to earthquake shaking must be categorized in
by fragility group and performance group.
A fragility group is a collection of components, or assemblies, all of which
have similar:
construction characteristics including details of manufacture and
installation
potential modes of damage
probability of incurring these damage modes, when subject to the effects
of earthquake shaking
potential consequences resulting from the damage, assuming that the
components are installed in the same locations
A fragility group can comprise individual components (e.g., pendant light
fixtures), or assemblies of components (e.g., interior fixed partitions
including the metal studs, gypsum board sheathing, wall coverings and any
embedded electrical wiring and plumbing). Fragility groups are organized
depending on whether they are part of the building’s core and shell, or part of
the building normally associated with individual occupancies and tenant
April 15, 2009
46 4: Implementation ATC58
improvements. Core and shell fragility groups include those associated with
the structural systems, exterior cladding, elevators and similar items.
Occupancyspecific fragility groups include ceilings, interior partitions,
HVAC systems, office furnishings, computers and similar items.
Each fragility group is identified by a unique identification code per the
National Institute of Standards and Technology Interagency Report
(NISTIR) 6389 Uniformat II elemental classification system (NIST, 1999).
Appendix B contains a listing of all fragility groups contained within the
default PACT databases. It is possible to define new fragility groups or to
modify data associated with existing fragility groups in order to meet the
needs of a specific building.
Components and assemblies that are of similar, but not identical construction
may have to be assigned to different fragility groups, in order to correctly
characterize their susceptibility to damage and the consequences of such
damage. For example, fixed interior partitions throughout a building could
all be assigned to the same fragility group, if they are all of the same height,
all installed in the same manner and all have the same wall coverings.
However, if some of the interior partitions were installed with slip tracks at
their top to permit the walls to move independently of the structure under
seismic shaking and some are provided with fixed top connections, each type
would have to be assigned to a different fragility group, as each would have
different susceptibility to damage under the same earthquake shaking.
Similarly, if some of the interior partitions had painted finishes and others
had ceramic tile finishes, they would have to be assigned to different
fragility groups because repairs would be more costly and more time
consuming for the ceramic tile covered walls then for the painted walls.
The following logic is recommended for the establishment of fragility
groups, when the default fragility groups provided in Appendix B are not
adequate to define the building assets at risk:
1. Identify the components that are likely to suffer damage and
contribute to potential losses. These are components that both can be
damaged and will have significant impact on building performance if
they are damaged. Components that are inherently rugged and not
subject to significant damage for credible levels of demand need not be
considered. For example, in a steel frame building the floor beams that
are not part of the building’s seismic force resisting system will not
typically be damaged by earthquake shaking and need not be considered.
Also, some nonstructural components (e.g., toilet fixtures) can often be
April 15, 2009
ATC58 4: Implementation 47
neglected as they are inherently rugged and generally not damaged by
earthquake shaking.
Some components may not be damaged by earthquake shaking, but can
be affected by the required repairs for other components. For example, a
conduit embedded in an interior partition, by itself is inherently rugged.
However, if the partition is sufficiently damaged to require complete
replacement, it will be necessary to replace the conduit, even though the
conduit was not damaged. In this case, the wall, and not the conduit,
would be assigned to the fragility group. The fact that the wall contains
the conduit, however, is acknowledged as discussed later.
2. Group components into logical sets considering normal design and
construction process and specification sections. One logical
performance group for many buildings is interior partitions. These
partitions are constructed out of a series of individual subcomponents
including coldformed steel framing, gypsum wallboard, fasteners,
drywall tape, plaster, paint, etc. Since these items have both a design and
construction relationship and tend to be damaged as an assembly, they
should be included in a common performance group. Despite the fact
that these subcomponents are grouped together as an assembly, it is
important to recognize that if the design or construction of one is
changed, for example, the manner in which the studs attach to the floors
above and below, this change will likely affect the damageability of all
these subcomponents. Similarly, repair of damage for one component is
likely to involve others in the group.
3. Group components into fragility groups such that all components in
the group are damaged by the same demands. In these procedures,
the probability that any building component (or collection) of
components will be damaged must be tied to a single demand parameter
such as peak story drift, peak floor acceleration, peak plastic rotation
demand, or similar quantity. All components within a fragility group
must be assumed to be dependent on the same general demand parameter
as an indicator of the extent of damage they will experience. It is
important to note, however, that not all components with in a fragility
group need to actually experience the same demands within the building.
For example, if the partitions at the third story of a building are of the
same type as those in the second story, they can be assigned to the same
fragility group even though the demands (story drift) will likely be
different in the third story than in the second story.
4. Group components such that fragilities, damage states and
consequence functions are logical for monitoring and repair. All
April 15, 2009
48 4: Implementation ATC58
components placed in a single fragility group should have similar
damage states and similar consequences of damage. Interior partitions
and exterior cladding should be placed in different fragility groups
because the damage states for each will likely be different, they will be
damaged at different levels of drift, and the consequences of the damage
to interior partitions are different than the consequences of damage to
exterior cladding. In these procedures, these damage characteristics are
categorized in the form of fragility specifications, discussed in section
4.7, below.
Performance groups are a subcategorization of fragility groups. A
performance group is that set of members of a fragility group that are
assumed to be subject to identical demands. Consider a multistory,
rectangular, momentresisting steel frame office structure. Table 42 lists the
minimum set of fragility groups that might be used to define the performance
model for such a building, including their NISTIR Classification per
Appendix B, their type (shell or occupancy), their description and the
demand parameter that is predictive of the type of damage each incurs.
Table 42 Example Fragility Groups for Momentresisting Steel Frame
Office Structure.
NISTIR
Classification
Type Description Demand Parameter
B1035.000 Shell Momentconnected beam
column assembly
Story drift parallel to
frame
B2020.001 Shell Exterior Cladding Story drift parallel to
wall
D3063.000 Shell Mechanical Equipment Floor acceleration
D1011.002 Shell Hydraulic Elevators Peak ground
acceleration
C1011.009 Occupancy Interior partitions Story drift parallel to
wall
E2022.011 Occupancy Desktop Equipment Floor acceleration
C3032.001 Occupancy Ceiling system Floor acceleration
E2022.000 Occupancy Bookcases & file cabinets Floor acceleration
Although many of these fragility groups may be essentially identical at each
level of the building, components and assemblies at each level will typically
experience different drift and acceleration demands, and therefore must be
assigned to different performance groups. Further, some components and
assemblies are sensitive to demands in one direction of shaking, rather than
another. Examples of this type of assembly include interior partitions and
April 15, 2009
ATC58 4: Implementation 49
cladding, each of which will be damaged by inplane, as opposed to outof
plane drift demand. Thus, separate performance groups must be formed for
partitions at and cladding at each level, oriented in a particular direction of
shaking. Table 43 presents a complete listing of the performance groups,
and quantities defined for a hypothetical threestory, steel framed office
building. The table lists the NISTIR number and description of each
performance group, the specific demand parameter used to predict damage to
that performance group and the normative quantity of components or
assemblies assigned to that performance group. Together, this information
defines the performance model for this building.
Table 43 Example Performance Groups and Quantities for 3story,
Momentresisting Steel Frame Office Structure.
No NISTIR
Classification
Type Description Quantity Demand
Parameter
1 B1035.000 Shell Roof EW
Moment
connected
beam column
assembly
8 each 3
rd
Story EW
drift
2 B1035.000 Shell Roof NS
Moment
connected
beam column
assembly
8 each 3
rd
Story NS
drift
3 B1035.000 Shell 3
rd
Florr EW
Moment
connected
beam column
assembly
8 each 2nd Story EW
drift
4 B1035.000 Shell 3
rd
Floor NS
Moment
connected
beam column
assembly
8 each 2
nd
Story NS
drift
5 B1035.000 Shell 1
st
Story EW
Moment
connected
beam column
assembly
8 each 1
st
Story EW
drift
6 B1035.000 Shell 1
st
Story NS
Moment
connected
beam column
assembly
8 each 1
st
Story NS
drift
7 B2020.001 Shell 3
rd
Story
Exterior
Cladding N & S
walls
2,600 sq.
ft.
3
rd
Story EW
drift
April 15, 2009
410 4: Implementation ATC58
No NISTIR
Classification
Type Description Quantity Demand
Parameter
8 B2020.001 Shell 3
rd
Story
Exterior
Cladding E & W
walls
3,200 sq
ft.
3
rd
Story NS
drift
9 B2020.001 Shell 2nd Story
Exterior
Cladding N & S
walls
2,600 sq.
ft.
2nd Story EW
drift
10 B2020.001 Shell 2nd Story
Exterior
Cladding E & W
walls
3,200 sq
ft.
2nd Story NS
drift
11 B2020.001 Shell 1st Story
Exterior
Cladding N & S
walls
3,000 sq.
ft.
1st Story EW
drift
12 B2020.001 Shell 1st Story
Exterior
Cladding E & W
walls
4,800 sq
ft.
1st Story NS
drift
13 D3063.000 Shell Mechanical
Equipment
5 Ea Roof
acceleration
14 D1011.002 Shell Hydraulic
Elevators
3 Ea Peak ground
acceleration
15 C1011.009 Occupancy 3
rd
Story EW
Interior
partitions
6,000 sq ft EW 3
rd
Story
drift
16 C1011.009 Occupancy 3
rd
Story NS
Interior
partitions
9,600 sq ft NS 3
rd
Story
drift
17 C1011.009 Occupancy 2nd Story EW
Interior
partitions
6,000 sq ft EW 2nd Story
drift
18 C1011.009 Occupancy 2nd Story NS
Interior
partitions
9,600 sq ft NS 2nd Story
drift
19 C1011.009 Occupancy 1st Story EW
Interior
partitions
6,000 sq ft EW 1st Story
drift
20 C1011.009 Occupancy 1st Story NS
Interior
partitions
9,600 sq ft NS 1st Story
drift
21 E2022.011 Occupancy 3
rd
Story
Desktop
Equipment
50 ea 3
rd
Floor
acceleration
22 E2022.011 Occupancy 2nd Story
Desktop
50 ea 2nd Floor
April 15, 2009
ATC58 4: Implementation 411
No NISTIR
Classification
Type Description Quantity Demand
Parameter
Equipment acceleration
23 E2022.011 Occupancy 1st Story
Desktop
Equipment
15 ea 1st Floor
acceleration
24 C3032.001 Occupancy 3
rd
Story Ceiling
system
16,000 sq
ft
Roof
acceleration
25 C3032.001 Occupancy 2
nd
Story Ceiling
system
16,000 sq
ft
3
rd
Floor
acceleration
26 C3032.001 Occupancy 1
st
Story Ceiling
system
10,000 sq
ft
2
nd
Floor
acceleration
27 E2022.000 Occupancy 3
rd
Story
Bookcases & file
cabinets
100 ea Floor
acceleration
28 E2022.000 Occupancy 2
nd
Story
Bookcases & file
cabinets
100 ea Floor
acceleration
29 E2022.000 Occupancy 1
st
Story
Bookcases & file
cabinets
30 ea Floor
acceleration
For buildings containing one or more of the default occupancy categories
listed in Table 41 PACT will automatically develop the necessary
performance groups and normative quantities based on input of key building
descriptive data including:
Number of stories
Typical story height
Typical EW building dimension (assuming rectangular plan shape)
Typical NW building dimension (assuming rectangular plan shape)
Occupancy of each floor
Most buildings will deviate from the performance model constructed by
PACT due to irregularities in plan shape, differences in the height of some
stories, or peculiarities of the individual building occupancy. For an
individual building, it is possible to modify the default properties as part of
the PACT input, to more closely represent the actual building.
In addition to modifying the quantities of performance groups and
components/assemblies within each group, it is also necessary to associate
each performance group with a particular fragility group, having specific
damage characteristics. This is defined by the fragility group’s fragility
specification. Section 4.7 describes fragility specifications.
April 15, 2009
412 4: Implementation ATC58
4.3 Characterize Ground Motion
After assembly of the building performance model, the next step is to
determine the ground shaking hazards for which the building’s performance
will be assessed. The way in which ground shaking hazard is characterized
is a function of the type of assessment being performed and the type of
structural analysis that will be used. For intensitybased assessments, the
ground shaking hazard is characterized by an elastic acceleration response
spectrum, or pairs of spectrumcompatible ground motions. For scenario
based assessments, ground motion is characterized by a specified magnitude
and distance; a median response spectrum appropriate to that magnitude and
distance; a dispersion associated with uncertainty in this spectrum; and, if
response history analysis is performed, pairs of spectrum compatible ground
motions. For timebased assessments, ground motion is characterized by a
series of intensities that span the range of potential damaging events.
Chapter 5 provides more detailed guidelines on this step of the assessment
process.
4.4 Structural Analysis
After characterizing the earthquake hazards, structural analysis is used to
characterize the peak demands produced by the building’s response to these
hazards. Two types of structural analysis are available:
Nonlinear response history analysis
Simplified analysis
The simplified analysis, limited in application to regular, lowrise structures,
uses an equivalent lateral force formulation to assess median story drift and
floor acceleration demands. One analysis must be performed for each
direction of building response. Nonlinear response history analysis can be
used to develop estimates of earthquake demands for any building. Chapter
6 provides detailed guidance on the use of both types of analyses.
4.5 Develop Collapse Fragility
Building collapse is the most common cause of casualties. Collapse is less
significant with regard to assessment of potential repair costs and repair
times, as building will typically experience extreme amounts of damage that
make repair impractical at ground shaking intensities below those that
produce collapse. Therefore, if potential building performance associated
with casualties is to be assessed, it is necessary to develop a building
specific collapse fragility function that indicates the probability of building
collapse as a function of ground shaking intensity.
April 15, 2009
ATC58 4: Implementation 413
Two methods are available to develop buildingspecific collapse fragilities.
One of these, termed Incremental Dynamic Analysis involves the use of
repetitive suites of nonlinear dynamic analyses to characterize the intensities
at which collapse is likely to occur. This technique can be used for any
building. A second simplified approach uses nonlinear static (pushover
analysis) to approximately characterize the building’s postyield
characteristics and relate these to collapse probability. The pushover
approach is limited to lowrise regular buildings.
If, for a particular assessment, it is not desired to characterize casualties,
robust collapse fragility parameters can be used in PACT to preclude the
probability of collapse. Chapter 7 provides detailed guidance on
development of collapse fragilities.
4.6 Form Realizations
Each realization represents one possible outcome of the building’s response
to a particular ground shaking intensity or earthquake scenario. For
purposes of casualty calculation, this includes determination of the time of
day and day of the year, in which the realization is assumed to occur. The
event time is determined by a random number generation scheme that within
a range of 1 to 8,760, which is the total number of hours in a year. An hour
number of 1 to 24 indicates an occurrence on J anuary 1, an hour number of
25 to 48 indicates J anuary 2, and the specific number indicates the time of
day that the earthquake occurs. For purposes of determining whether a day
is a weekend or weekday, an arbitrary standard calendar is assumed.
The outcome can either include building collapse, which results in total loss
and large numbers of casualties, or if collapse does not occur, a vector of
peak response quantities, including story drift at each level in each direction
and floor acceleration at each level. If collapse is not predicted to occur for
the realization, the vector of peak response quantities are used as the setoff
demands to determine the damage that occurs to the components of each
performance group.
Section 3.2.5 describes the methods used to determine if collapse occurs for
a particular realization. If collapse is not predicted, then the demands
obtained from the structural analyses described in Chapter 6 are used to form
a joint lognormal probability distribution for the demands on the structure,
considering record to record variability in response as well as modeling and
analysis uncertainty. PACT automatically forms this joint lognormal
demand distribution and also will generate as many realization demand
vectors as desired using the algorithm presented in Appendix G. In order to
April 15, 2009
414 4: Implementation ATC58
form a smooth distribution of the performance measures, at least 2000
realizations should be used for each ground motion intensity or earthquake
scenario.
4.7 Damage Determination
For each realization it is necessary to determine a building damage state. A
building damage state is a description of the damage to each component or
assembly in each of the performance groups, together, with a listing of the
following data used to determine performance:
The repair actions that are associated with the damage to individual
components
The labor hours associated with the repair of damage
The need to replace a building component that has a long lead time, that
is will take 3 or more months to obtain
The square foot of floor space at each level that has been subject to
falling debris, or volumetric space compression and in which people will
be subject to injury
4.7.1 Damage Correlation
Damage is determined on a performance group by performance group basis.
For performance groups that contain multiple components or assemblies, it is
necessary to determine if the damage to the components is correlated or
uncorrelated. If the performance group is designated as correlated, then for a
given realization, all components in the performance group will be deemed to
be damaged to the same extent. If the performance group is uncorrelated,
then for a given realization, each component in the performance group can
have a different level of damage.
In order to properly capture the uncertainty in building performance, most
performance groups should be designated as uncorrelated. Performance
groups can be assumed to be correlated when there are two or more means of
installing the components in the group, the means of installation has
significant effect on the level of damage that occurs, it is likely that all
components are installed in the same manner, regardless of what this is, and
the performance analyst does not know which type of installation was or will
be used. In this case, multiple damage states can be assigned to represent the
two different conditions, and all components will be either assigned to one or
the other, on a random basis.
April 15, 2009
ATC58 4: Implementation 415
4.7.2 Fragility Specifications
Each fragility group is associated with a unique fragility specification that
defines, for all components associated with that fragility group:
The components and/or assemblies that the specification applies to
The damage states that can occur
The probability of incurring these damage states, as a function of demand
The consequences of incurring these damage states in terms of repair
effort and time and potential injury to persons
Figure 42 illustrates the typical format for fragility specifications. Appendix
C contains a series of default fragility specifications for common structural
systems including:
Moment resisting steel frames (high ductility, intermediate ductility and
ordinary ductility)
Moment resisting concrete frames (high ductility, intermediate ductility
and ordinary ductility)
Steel Braced Frames (high ductility and intermediate ductility)
Concrete shear walls
Masonry shear walls
Light framed walls with structural sheathing
Appendix D provides default fragility specifications for the nonstructural
components and systems typically necessary to assess the performance of the
common occupancies indicated in Section 4.2.1. Individual users can also
specify fragility specifications for other types of structural systems and
nonstructural components using the format illustrated in Figure 42 and the
procedures of Appendix F. The subsections below describe the individual
fragility specification entries.
4.7.3 Damage States
Damage to building components and performance groups is generally a
continuum over which the scope and extent of damage increases with
increasing demand. In these procedures, rather than using a continuous range
of possible damage we use a relatively few, discreet states for each fragility
group, to characterize the different levels of damage that can occur.
The fragility specification for each fragility group includes a description of
each unique damage state that affects postearthquake functionality or
April 15, 2009
416 4: Implementation ATC58
occupancy, personnel safety, or repair cost and time. Each damage state is
described in terms of the physical and operational consequences. For
example, a fan unit might break its piping connections (repair implications)
resulting in an inoperable HVAC system (downtime implications) and posing
a possible falling hazard (casualty implications).
There are several possible relationships among damage states:
Sequential damage states occur one after another. These are damage
states that represent a progression of damage from none to higher and
higher levels along as demand increases. Usually the change from one to
another signifies a more serious consequence. For example, for a
concrete component whose behavior is controlled by flexure, the damage
states simply reflect more and larger cracks as demand increases. At first
these might be repaired with simple patching and painting. In
subsequent damage states the cracks may become large enough to require
injection with epoxy. Even higher damage states may reflect the need to
replace bent or fractured reinforcing bars and crushed concrete.
Mutually exclusive damage states are those for which two or more
alternatives are possible but not both. The implication for repair,
downtime, and casualties are different among the alternatives. For
example, the concrete component might reach a level of demand at
which either a.) the flexural cracks lengthen and widen (80% chance) or
b.) the behavior mode transitions to shear with the formation of a more
serious crack pattern (20% chance).
Simultaneous damage states can coexist with one another. The repair
efforts, downtime implications, or casualty consequences for these types
of damage states are generally independent of one another. An example
is an elevator that becomes inoperable for a given level of demand.
Damage can consist of one or more conditions each with its own
independent probability of occurring, given that the elevator has become
inoperable (e.g. damaged controls (30%), damaged vane and hoistway
switches (20%), damaged entrance and car door (30%), oil leak in
hydraulic line (10%). Other states are also possible, but not covered here
for brevity.
April 15, 2009
ATC58 4: Implementation 417
Fragility specification
RC wall panel (flexurally controlled)
Developer and date: Comartin (example for Guidelines), November 2008
Component category: B1044.000
Basic composition: Reinforced concrete and paint both sides
Normative quantity (unit): sf of wall area
Demand parameter: Interstory drift
No. of damage states (other than
DS0undamaged):
3
Type of damage states: Sequential
Sequential damage states
DS1 DS2 DS3
Description: Flexural cracks < 3/16"
Shear (diagonal) cracks < 1/16"
No significant spalling
No fracture or buckling of r/f
Not structurally significant
Flexural cracks > 1/4"
Shear (diagonal) cracks >
1/8"
Moderate spalli ng/ loose
cover
No fracture or buckling of r/f
Insignificant residual
drift/shortening
Repairable in place
Max. crack widths >3/8"
Significant spalling/ loose
cover
Fracture or buckling some r/f
Significant residual
drift/shortening
Repair in place impractical
Illustrations:
Figure 42 Representative Fragility Specification
Median demand (θ)
1
:
1.5%
3%
5%
Dispersion (β)
1
:
0.2
0.3
0.4
Probability
1
:
Correlation:
Consequence functions
Repair description:
Patch cracks each side
Remove loose concrete
Shore
Paint each side
Patch spalls with NS grout
Demo existing wall
Patch cracks each side
Replace
Paint each side
Patch and paint
Repair costs:
Max. consequence up to
lower
quantity
$4.00 per up to 800 sf
$10.00 per sf up to 800 sf
$50.00 per sf up to 200 sf
Min. consequence over
upper quantity
$2.00 per sf over 4000 sf $5.00 per sf over to 4000 sf
$30.00 per sf over 2000 sf
Dispersion (β): 0.2
0.3
0.3
Repair time:
Max. consequence up to
lower quantity
0.02hr per sf up to 800 sf 0.04hr per sf up to 800 sf
0.4hr per sf up to 200 sf
Min. consequence over
upper quantity
0.01hr per sf over 4000 sf
0.02hr per sf over to 4000 sf
0.2hr per sf over 2000 sf
Dispersion (β): 0.2
0.3
0.3
Casualty  affected
planar area (sf) per
normative unit:
n/a
n/a
n/a
Casualty rate in
affected planar area:
n/a
n/a
n/a
Post event tagging
flag:
green
red
red
T otal damage
quantity
(red tag
trigger )
n/a
40% of all walls
10% of all walls
April 15, 2009
418 4: Implementation ATC58
4.7.4 Component Fragility Functions
Section 3.2.6.2 outlines the basic concepts related to fragility functions.
These relationships specify the probability, as a function of a single demand
parameter, that a component in a performance group will be damaged to any
of its possible damage states. They are characterized by a median demand
for each damage state representing a 50% chance of occurrence and a
lognormal uncertainty, β, generating the chances of occurrence of the
damage state, given higher or lower demands. For simultaneous and
mutually exclusive damage states, the sum of the probabilities of no damage
and that of all possible damage states must be 1.0. This is true regardless of
demand. Individual probabilities of damage states may change with demand
but the performance group must always be in one and only one damage state
for a given demand including the possibility of no damage. This restriction
does not apply to simultaneous damage states. These damage states are
independent and their probabilities for any given demand may sum to a value
greater than 1.0.
4.7.5 Consequence Functions
Consequence functions are used to translate damage into losses. Since
damage can affect repair costs, downtime, and casualties, fragility
specifications include consequence functions that are used to determine the
losses that result from each damage state. With each damage state there is an
associated description of repair actions, downtime consequences and casualty
consequences. These are described below.
Repair actions
These are the required actions to return the components in a fragility group to
their preearthquake state are specified in detail. These specifications are
stepbystep lists of the work that a contractor would implement in the field.
For each damage state the unit costs for the repairs described for the
performance group are expressed with the following parameters (see Figure
33):
Lower quantity: the quantity of repair actions of a give type, below
which there is no cost discount to reflect efficiency in construction
operations (see Section 3.2.
Maximum cost: This is the cost to perform a unit damage state repair,
without any efficiency of scale being obtained.
Upper quantity: the quantity of repair work, above which no further
efficiencies of scale are attainable
April 15, 2009
ATC58 4: Implementation 419
Minimum cost: the minimum cost that can be attained for a unit repair
operation, considering all possible efficiencies of scale
Dispersion: the uncertainty associated with the unit cost.
Downtime
There are two primary sources of damagerelated downtime: the creation of
dangerous conditions that result in tagging of the building as unsafe to
occupy, and interruption associated with making habitable space available
for repair actions. Each damage state has associated with it, a flag that
indicates whether the onset of this damage state results in a condition that
would render a portion of the building unsafe for occupancy.
In order to estimate downtime associated with repair actions, each damage
state includes a time related consequence function that indicates the number
of labor hours associated with the required repair. Labor effort like repair
cost can benefit from efficiency of scale when a large quantity of a specific
repair action must be undertaken. These effects are described in the fragility
specification through the following parameters:
Lower quantity: the minimum quantity of a given repair action, below
which there is no time efficiency.
Maximum time: the maximum number of labor hours necessary to
implement a repair action without efficiency of scale.
Upper quantity: maximum quantity of a repair action above which there
is no further efficiency.
Minimum time: the minimum number of labor hours per unit repair
action, considering possible economies of scale.
Dispersion: the uncertainty associated with the labor effort associated
with a given repair action.
Finally, repair time can be heavily affected by long delivery times for critical
equipment and components. If a damage state requires replacement of a
component that is associated with a long delivery time, this is indicated in the
fragility specification.
Casualties
Although building collapse is the principle cause of earthquake casualties,
some damage states associated with individual components or assemblies can
have potential casualty consequences. These are generally associated with
falling debris, for example a parapet falling off a building, but can also be
associated with the release of toxic materials. In order to determine these
April 15, 2009
420 4: Implementation ATC58
consequences, the fragility specification must include description of the
number of individuals exposed to a potential hazard related to the population
of the planar area affected by the hazard (casualtyaffected planar area
(square foot) per normative unit). The area is expressed in terms of
normative units. For example, a toppled parapet might pose a safety hazard
in an area that is located within ten feet of its plan location. The normative
unit for parapets is lineal feet, so the casualty affected planar area is 10
square feet per lineal foot of parapet.
4.7.6 Damage Aggregation
For each realization in which collapse does not occur, each component (or
each performance group for correlated components) is tested to determine its
damage state. This is performed by using random number generation
techniques and the individual damage state fragility functions to determine
the damage state for each component. This is done automatically within the
PACT software.
Once the damage state for each component has been determined, PACT
aggregates the required repair actions of each type considering all
performance groups and determines the repair costs and labor hours using the
associated consequence functions. PACT also determines the number of
occupancy tagging flags that are triggered by the damage state, the presence
of any long leadtime repair actions, and the locations and quantities of
casualtyaffected areas in and around the building. Together, these
aggregated repair costs, labor hours, lead times, casualty impact areas and
tagging flags comprise the unique building damage state for the realization.
This process is repeated for each realization.
4.8 Repair Costs
As described in the previous section, for each realization, the building
damage state indicates the necessary repair actions of all types to restore the
building to preearthquake condition. These repair actions are expressed in
terms of a total repair cost. This repair cost is computed for a reference
standard location in the United States at a reference period of time,
neglecting uncertainty in contractor pricing strategies or construction cost
escalation. To adjust these costs for the buildingspecific location, it is
necessary to apply a construction cost index. These procedures use the
RSMeans Construction Cost Index. PACT will assume a construction cost
index of 1, unless a different value is input by the user.
Uncertainty associated with contractor pricing strategies and market
conditions is handled by a pricing uncertainty value expressed as a
April 15, 2009
ATC58 4: Implementation 421
percentage of construction cost. A random number generation scheme is
used to determine the number of standard deviations above or below the
median value that pricing should be set for each realization.
4.9 Occupancy Interruption Time
Actual time that a building will be unusable for beneficial occupancy is very
difficult to determine. Factors that affect this include:
Who is responsible for performing repairs – the owner or the tenants
Whether the party responsible for performing the repairs has the
necessary financial resources to pay for the needed repairs
The availability of design professionals to design repair operations and
the availability of contractors to perform these operations.
Whether or not the contractor has free access to the building to conduct
the repairs or if the building remains in service and the contractor must
limit their activity at any time to small portions of the building
Whether the building is tagged, and the length of time necessary to
convince the building official that it is safe to conduct repair operations
in the building
These uncertainties make the estimation of occupancy interruption time an
intractable problem. Therefore, these procedures provide the following
measures of occupancy interruption:
1. The length of time necessary to conduct repairs, assuming that the
building is made available to the contractor on an asneeded basis
throughout the repair process. This can be considered a lower bound on
the potential time that the building will not be available for occupancy,
free of construction operations.
2. The need to procure any long leadtime operations.
3. The probability that the building will be redtagged.
The first of these measures is calculated on a realization basis, by summing
the total labor hours obtained for all the repair operations indicated by the
building damage state, applying an uncertainty adjustment to these labor
hours and then dividing by a labor hour/1,000 square foot factor that reflects
the number of workmen that can reasonably work concurrently in a limited
construction site within an existing building. This calculation assumes that
construction work is performed on a standard 5day week, 8hour day
schedule.
April 15, 2009
422 4: Implementation ATC58
The need to procure longlead items is determined directly from the building
damage state by determining if any flags for longlead items are triggered.
The probability that the building will be redtagged is determined by
summing the “tag flags” indicated for the various performance groups and
determining if they exceed a critical level that indicates a high likelihood of
tagging.
4.10 Casualties
The building damage state for each realization includes a description of the
floor area at each level that is:
Subject to volumetric compression
Subject to falling debris or release of toxins
The number of persons at risk depends on the time of day and day of the
year, which is selected for each realization as described in Section 4.6.
The number of casualties for the realization is determined by factoring the
floor area associated with volumetric compression by the number of persons
per 1,000 square foot associated with the occupancy in that floor area for the
day of the year and time of day, and a structuralspecific system factor that
indicates the casualty rate per exposed occupant. The number of casualties
associated with falling debris or toxins is determined by summing the
affected areas, factoring these by the componentspecific casualty rate per
affected area and the occupancy specific number of persons.
Structural systemspecific casualty rates are contained in Appendix C.
4.11 Loss Distributions
As described in Chapter 3, performance is expressed in the form of loss
distributions for each of the performance measures including repair costs,
repair time, as described in the previous section, red tagging, and casualties.
These distributions are developed by ordering the realizations for each of
these performance measures, as previously described. The sections below
describe this process respectively for intensitybased, scenariobased and
timebased assessments.
4.11.1 Intensity and ScenarioBased Assessments
For each realization, losses are one possible outcome of a given shaking
intensity or scenario. By repeating the process numerous times a more
complete probabilistic perspective can be generated for the intensity or
scenario losses. The procedure is to order the losses of each type to generate
April 15, 2009
ATC58 4: Implementation 423
a cumulative distribution function. The probability that a realization loss of a
given type is equal to or less than any value is simply the number realizations
for which the loss is less divided by the total number of realizations. Figure
43 shows sample results of an intensity or scenariobased cumulative loss
distribution function for repair costs. In the figure, each dot represents a
scenario loss. Any value of total loss can be deaggregated into performance
group losses as shown in Figures 44, by summing up, on a realization basis,
the sources of the repair costs. Results for casualties and downtime are
similar. Casualties can be deaggregated into those due to collapse of the
modes and those caused by various noncollapse hazards associated with
performance groups.
Median loss $1.3M
Repair/replacement cost (thousands of $US)
P
r
o
b
a
b
i
l
i
t
y
t
h
a
t
c
o
s
t
i
s
≤
Figure 43 Sample repair/replacement losses for a scenario or intensitybased
assessment.
Figure 44 Direct losses deaggregated by performance group for a scenario or intensity
based assessment.
April 15, 2009
424 4: Implementation ATC58
4.11.2 TimeBased Assessments
A timebased assessment generates the annual probability of losses from all
intensities of shaking. This relationship is developed by integrating the
losses from a series of intensities over a hazard curve representing the
probability of experiencing varying intensities of shaking. PACT performs
this integration automatically. Figure 45 illustrates the results of this
process. The vertical axis in the plot is the annual probability that the loss on
the horizontal axis is exceeded. Note that the losses may be dollars of repair
costs, number of casualties, or days of downtime time. The area under the
loss curve is the average annualized loss. Integrating the global collapse
fragility function over the site hazard curve produces an estimate of the
annual probability of collapse.
Average annualized loss$35,000
Figure 45 Sample loss calculations for a timebased assessment.
April 15, 2009
ATC58 5: Ground Shaking Intensity 51
Chapter 5
Ground Shaking Intensity
5.1 Introduction
This section presents a set of acceptable procedures for characterizing
earthquake ground shaking either for simplified analysis or nonlinear
response history analysis. Appendix E presents background information on
characterization of ground shaking that may be useful for some readers.
The effects of horizontal shaking are considered by applying the earthquake
shaking effects simultaneously along each of two principal orthogonal
building axes. For simplified analysis, the bidirectional shaking is applied as
a pair of spectral response acceleration demands at the first mode period
along each axis. For responsehistory analysis, the bidirectional shaking is
applied as pairs of ground motion histories scaled to represent the desired
ground shaking intensity level. Adjustments for soil type are made using the
procedures of Section 5.3.
Section 5.4 presents the United States Geologic Survey (USGS) ground
motion calculator and describes how it can be used to generate acceleration
response spectra. Attenuation functions which form the basis for ground
shaking used in scenariobased assessment are introduced in Section 5.5.
Section 5.6 presents procedures for scaling ground motions for use with
nonlinear responsehistory analysis. Section 5.7 describes the
characterization of ground motion demands for use with simplified analysis.
5.2 Building Location and Site Conditions
5.2.1 Seismic Environment and Hazard
The characterization of earthquake shaking hazard is dependent on the
location of the site with respect to causative faults, the regional and site
specific geologic characteristics, local topographic conditions, and the
chosen type of assessment.
These performance assessment procedures address earthquake shaking
hazards only. However, other seismic hazards may exist at a building site
that could damage the building regardless of its ability to resist ground
shaking, including ground fault rupture, liquefaction, lateral spreading,
landslides, and inundation from offsite effects such as dam failure or
April 15, 2009
52 5: Ground Shaking Intensity ATC58
tsunami. Each of these hazards could contribute substantially to loss. As a
minimum, a qualitative assessment of the potential impact of these hazards
should be made and, if found to be significant, noted and addressed.
Earthquake shaking can be fully defined by two orthogonal horizontal
components and one vertical component. However, there is little evidence to
suggest that vertical shaking is a key contributor to damage and loss in most
buildings and so this chapter addresses only horizontal earthquake shaking.
Procedures to characterize vertical shaking effects, when deemed significant,
are presented in Appendix E.
Local topographic conditions (e.g., hills, valleys, canyons) can modify the
amplitude, frequency content and duration of earthquake shaking relative to
that expected at a flat, level site. Finiteelement and finitedifference methods
of site response analysis can be used to characterize the influence of
topographic effects on earthquake shaking but a treatment of topographic
effects is beyond the scope of these Guidelines.
5.2.2 Location
The building site’s exact location (longitude and latitude) must be identified
to characterize the seismic hazard for timebased assessments in order to
derive the seismic hazard curve. Three decimal fraction places are typically
sufficient to identify the latitude and longitude. For scenariobased
assessments, the distance from the building site to the causative fault must be
known. For intensitybased assessments, site location need not be defined.
5.2.3 Site Soil and Topographic Conditions
The properties of the soil (rock) at the building site must be defined for both
scenario and timebased assessments and should also generally be established
for intensitybased assessment to permit an appropriate spectral shape and
amplitude to be used to characterize the hazard. The soil should be
characterized by one of the following methods:
1. When attenuation relationships are used to determine the hazard,
characterize the site soil conditions with values of variables that are
consistent with the attenuation relationship(s), such as average shear
wave velocity in the upper 30 meters
2. When the USGS ground motion calculator (Section 5.4) is used to
determine the hazard, site soil conditions should be characterized by site
class per Section 5.3.
April 15, 2009
ATC58 5: Ground Shaking Intensity 53
5.3 Spectral Adjustments for Soil Conditions
The USGS maintains national seismic hazard maps and a ground motion
calculator that can be used to characterize site hazards. These maps provide
the geographic distribution of spectral response accelerations that are
appropriate for a reference site condition characterized by an average shear
wave velocity in the upper 30 m of the soil column of 760 m/sec. If the
USGS maps or ground motion calculator are used and this reference
condition is not representative of that at the building site, an adjustment must
be made for the local geology.
The procedures presented in ASCE 705 can be used to make spectral
adjustments for soil conditions. As an alternate to the ASCE 705
procedures, other acceptable procedures include 1dimensional site response
analysis and soilfoundationstructure interaction analysis, as described in
Appendix E.
Sidebar: Site Class. The procedures of ASCE 705 permit the use of Site
Class D when the site soil conditions are undefined because: (a) the
characteristics of the response spectrum derived using Site Class D
assumptions are generally more conservative for design than those derived
using Site Class A, B, or C; (b) Site Class D is most prevalent; and (c)
conditions associated with Site Classes E and F are relatively rare. Unlike
design for which conservatism is acceptable, unbiased estimates of loss are
sought for performance assessments. Therefore, the use of a default site class
is not recommended for performance assessment. Instead sufficient
investigation should be performed to adequately define the site class.
Under the ASCE 705 procedures, the soil at a building site is classified as
conforming to Site Class A, B, C, D, E or F, depending upon the soil’s
mechanical properties in the upper 30 m of the soil column immediately
below the building foundation. Once the site class has been identified
spectral demands computed for the reference site condition are adjusted to
account for amplification of ground motion that is associated with the
strength and stiffness of the soil column at the site. The correction factors
presented in Tables 11.41 (
a
F ) and 11.42 (
v
F ) of ASCE 705 can be used
for this purpose, where Table 11.41 presents correction factors for short
period (0.2 second) spectral response and Table 11.42 presents factors for
longperiod (1 second) response. Note that these correction factors are
average values for the short and long period segments of the spectrum.
Values of the 5% damped spectral acceleration at periods of 0.2 second and
1.0 second, for the assumed intensity of ground motion and the reference site
condition are required to compute the correction factors. A unique pair of
April 15, 2009
54 5: Ground Shaking Intensity ATC58
siteclass factors must be calculated for each ground motion intensity
considered.
To use the ASCE 705 tables to correct for site class, the Mapped Maximum
Considered Earthquake Spectral Response Acceleration Parameters (
s
S and
1
S per ASCE 705) are replaced with the values of spectral acceleration at
0.2 second (
0.2
S ) and 1.0 second (
1.0
S ) computed for the selected intensity of
motion. Values of the site coefficients,
a
F and
v
F , are then selected from the
tables, consistent with the site class and the shaking intensity. The adjusted
short and longperiod spectral parameters are computed using the equations:
0.2 0.2 a a
S F S = (51)
1.0 1.0 a v
S F S = (52)
These equations are also used to calculate a transition period
1.0 0.2
/
s a a
T S S =
between the constant acceleration and constant velocity segments of the
ASCE 705 spectrum.
5.4 USGSbased Ground Motion Calculator
A USGS ground motion calculator is available as a downloadable J ava
application at http://earthquake.usgs.gov/research/hazmaps/. This applet
permits the calculation of seismic hazards for sites in the conterminous 48
states. This section describes the use of this calculator for timebased and
intensitybased assessments.
5.4.1 TimeBased Hazard Calculations
The ground motion calculator can be used to develop probabilistic seismic
hazard curves that plot annual frequency of exceedance as a function of the
geometric mean (geomean) spectral demand at a given period. Although
several alternative hazard data sets are available, the most current USGS data
should be used for hazard computations.
Figure 51 presents a seismic hazard curve computed using the calculator for
a site in San Francisco. This hazard curve is plotted in semilog format
indicating the mean annual frequency of exceedance (yaxis) for spectral
response acceleration at a period of 1.0 second (xaxis) for the reference site
condition.
Sidebar: Uniform Hazard and Conditional Mean Spectra. The USGS
ground motion calculator can be used to generate spectra with selected
probabilities of exceedance in a given time period (e.g., 2% probability of
exceedance in 50 years). Such spectra are termed Uniform (or Equal) Hazard
Spectra (UHS) because each spectral ordinate has the same probability of
April 15, 2009
ATC58 5: Ground Shaking Intensity 55
exceedance. These Guidelines also makes reference to a Conditional Mean
Spectrum (CMS) that represents the geomean shaking demands associated
with a twocomponent earthquake record set. Later sections of this chapter
and Appendix E provide information on UHS and CMS.
The USGS ground motion calculator can be used to develop hazard curves
for spectral demands at periods of 0, 0.1, 0.2, 0.3, 0.5, 1.0 and 2.0 seconds.
For values of period, T , between these periods, linear interpolation can be
used to compute the spectral acceleration at a given mean annual frequency
of exceedance. For values of T greater than 2 seconds but less than or equal
to 5 seconds, the spectral acceleration can be computed as
2 ( 2sec)
( ) (2 5sec)
a
a
S T
S T T
T
=
= < s (53)
For site conditions other than those corresponding to the reference condition,
it is necessary to adjust the spectral acceleration on the xaxis of these hazard
curves using the procedures of Section 5.3.
A Uniform Hazard Spectrum can be constructed by generating seismic
hazard curves for periods in the range of interest and extracting spectral
demands for a given annual frequency of exceedance and plotting these
spectral demands versus period.
Figure 51 Representative seismic hazard curve for a site in San Francisco
with T = 1 second, plotted in semilog format
April 15, 2009
56 5: Ground Shaking Intensity ATC58
Sidebar: Constructing a Uniform Hazard Spectrum from Seismic
Hazard Curves. Consider the figure below that presents seismic hazard
curves for the reference site condition at periods of 0.2, 0.3, 0.5, 1.0 and 2
seconds for a site in San Francisco. Two horizontal lines are drawn at return
periods of 475 and 2475 years, which correspond to 10% and 2% probability
of exceedance in 50 years, respectively, and annual frequencies of
exceedance of 0.0021 (=1/475) and 0.0004 (=1/2475). The data points at
each annual frequency of exceedance in the upper figure (open red squares =
475year return period; open blue circles =2475year return period) are
mapped into the lower figure, which presents the Uniform Hazard Spectra for
these two return periods.
0 0.5 1 1.5 2 2.5
Spectral acceleration (g)
0.0001
0.001
0.01
0.1
1
A
n
n
u
a
l
f
r
e
q
u
e
n
c
y
o
f
e
x
c
e
e
d
a
n
c
e
Period
0.2 sec
0.3 sec
0.5 sec
1 sec
2 secs
2% PE in 50 years
10% PE in 50 years
Hazard curves at periods of 0.2, 0.3, 0.5, 1 and 2 seconds for a
representative site in San Francisco
0 0.5 1 1.5 2
Period (sec)
0
0.5
1
1.5
2
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
2% PE in 50 years
10% PE in 50 years
Uniform hazard spectra for 2% and 10% probability of exceedance in 50
years and the reference site condition
April 15, 2009
ATC58 5: Ground Shaking Intensity 57
The uniform hazard spectra in the figure above are associated with the
reference site condition. These spectra are adjusted in the figure below for
Site Class D using the procedures set forth in ASCE 705 and equations (51)
and (52). For the spectra presented above,
0.2
S and
1.0
S are (1.7 g and 0.84
g) for a 2% probability of exceedance in 50 years and (1.11 g and 0.50 g) for
a 10% probability of exceedance. The corresponding values of
a
F and
v
F
are (1.0, 1.5) for 2% probability of exceedance in 50 years and (1.06, 1.5) for
10% probability of exceedance in 50 years. Note that the spectra below have
a different shape from those that would be constructed by multiplying all
short period ordinates for the reference site condition by
a
F and all long
period ordinates by
v
F .
0 0.5 1 1.5 2
Period (sec)
0
0.5
1
1.5
2
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
Site Class D
2% PE in 50 years
10% PE in 50 years
Uniform hazard spectra for 2% and 10% probability of exceedance in 50
years and Site Class D, constructed using the procedures of ASCE 705.
5.4.2 IntensityBased Hazard Calculations
The USGS ground motion calculator can be used to generate an acceleration
response spectrum for intensitybased assessments. The calculator can
generate Uniform Hazard Spectra (for 2% and 10% probability of
exceedance in 50 years). The calculator enables adjustments to the spectra to
accommodate nonreference soil conditions at the site following the
procedures described in Section 5.3.
Table 51 presents the relationship between return period, annual frequency
of exceedance, and % probability of exceedance in 50 years. The annual
frequency of exceedance is the reciprocal of the return period. The %
probability of exceedance is calculated as
( ) 1
t
P SA sa e
ì ÷
> = ÷ (54)
April 15, 2009
58 5: Ground Shaking Intensity ATC58
where ( ) P SA sa > is the probability of exceedance of the Spectral
Acceleration in time period t (equal to 50 years in the table below) that is
expressed in the table below as a percentage, sa is the value of the spectral
acceleration and ì is the annual frequency of exceedance. One or more of
these return periods or percent probability of exceedance in 50 years could be
used to define a response spectrum for use in intensitybased assessment.
Table 51 Representations of Seismic Hazard for IntensityBased Assessments
Return
Period
(years)
Annual
Frequency of
Exceedance
% Prob. of
Exceedance
in 50 years
Return
Period
(years)
Annual
Frequency of
Exceedance
% Prob. of
Exceedance
in 50 years
10 0.10000 99.3 475 0.00211 10.0
20 0.05000 91.8 500 0.00200 9.52
30 0.03333 81.1 975 0.00103 5.00
40 0.02500 71.4 1000 0.00100 4.88
50 0.02000 63.2 1500 0.00066 3.28
72 0.01389 50.0 2475 0.00040 2.00
100 0.01000 39.4 4975 0.000201 1.00
200 0.00500 22.1 5000 0.000200 1.00
224 0.00446 20.0 10000 0.00010 0.50
ASCE 41 provides interpolation and extrapolation procedures to adjust
spectral response acceleration parameters mapped at 10% and 2% probability
of exceedance in 50 years for other probabilities of exceedance.
5.5 Attenuation Relationships
5.5.1 Introduction
For scenariobased assessment, ground shaking must be characterized for a
userspecified combination of earthquake moment magnitude and siteto
source distance (e.g., 7
W
M from a source 13 km from the building site).
Attenuation relationships are used, together with information on local and
regional geology and the soil profile at the site, to transform earthquake
magnitude and sitetosource distance into a median acceleration response
spectrum and perioddependant dispersion.
Information on bestestimate maximum moment magnitude for ruptures of
individual and multiple fault segments in the Western United States and the
Inner Mountain seismic belt can be found in the archival literature and at
http://earthquake.usgs.gov/. Such information can be useful for selecting an
April 15, 2009
ATC58 5: Ground Shaking Intensity 59
appropriate scenario, although in principle a scenario can be based on any
combination of magnitude and distance.
If the building site is within 15 km of the presumed zone of fault rupture and
the selected earthquake magnitude is 6.5
W
M or greater, fault directivity
effects should be considered in the spectral calculations by specifying that
the scenario event has: (a) forward directivity (rupture towards the site); (b)
reverse directivity (rupture away from the site); (c) null directivity; or (d)
unspecified directivity (random direction of rupture). Appendix E provides
guidance on faultrupture directivity and how to include it in hazard
calculations.
5.5.2 Functional Form
A simple form of a ground motion attenuation relationship is given by the
equation
1 2 3 4
ln ln Y c c M c R c R ¸ = + ÷ ÷ + , where lnY is the natural
logarithm of the median strongmotion parameter of interest (e.g., peak
ground acceleration, spectral horizontal acceleration at a particular period,
etc.), M is the earthquake magnitude, R is the sourcetosite distance or a
term characterizing this distance, and ¸ is the standard error term with a
mean of zero and a logarithmic standard deviation of
lnY
o (also referred to
as  ). The term
2
c M is consistent with the definition of earthquake
magnitude (the source) as a logarithmic measure of the amplitude of the
ground motion. The term
3
ln c R ÷ (the path) represents the geometric spread
of the seismic wave front as it propagates from the source. The value of
3
c
will vary with distance depending on the seismic wave type (body wave,
surface wave, etc). The term
4
c R ÷ treats the anelastic attenuation of seismic
waves caused by material damping (treating soil as a viscoelastic material)
and scattering (a result of reflections and refractions of seismic waves due to
the presence of heterogeneities and discontinuities in the earth’s crust,
causing multiple seismic waves to arrive at a site from different paths of
differing lengths). Typical attenuation relationships are more complicated
than the basic equation given above. Additional terms are used to account for
other effects including nearsource directivity, faulting mechanism (strike
slip, reverse and normal), site conditions (different relationships), and
hanging wall/footwall location of the site.
Attenuation relationships are derived using regression analysis on large
groundmotion data sets. Regression analysis is used to determine the best
estimate of the coefficients in the relationship, that is, the
1
c through
4
c in
the basic attenuation relationship given above. The value predicted by this
equation is the mean value of lnY , or the 50thpercentile or median value of
April 15, 2009
510 5: Ground Shaking Intensity ATC58
Y. The median value, by definition is exceeded by 50% of the observations
1
.
To capture randomness in the value of the ground motion parameter, the
probability of exceedance of any value of the parameter can be computed
using
1 ln
ln ln
Y
Y Y z
o o
o
÷
= + , where z
o
is the standard normal variable for an
exceedance probability of o . Although many attenuation relationships have
been developed, most apply to only a specific geographic region, a particular
site class, or type of rupture mechanism, based on the data set of ground
motion records used to develop the relationships. It is important to ensure
that an attenuation relationship selected for use in performance assessment is
appropriate to the particular building site and earthquake source.
Appendix E presents selected North American attenuation relationships. All
of the ground motion attenuation relationships described in the appendix use
moment magnitude
W
M to define earthquake magnitude. The Open Seismic
Hazard Analysis website, http://www.opensha.org/, provides a listing of
West Coast shallow crustal attenuation relationships (under the dialog box
Applications). The attenuation relationship plotter, which can be downloaded
from the website, can be used to generate median spectra and dispersions for
the listed relationships. The three Next Generation Attenuation relationships
(Boore and Atkinson, B_A; Campbell and Bozorgnia, C_B; Chiou and
Youngs, C_Y) developed recently for shallow crustal earthquakes on the
West Coast of the United States are available.
Sidebar: Attenuation relationships to compute spectral demands for
scenariobased assessments: Sample results of analysis using three of the
Next Generation Attenuation (NGA) relationships (labeled respectively:
B_A, C_B, C_Y) are presented in the figure below for Moment magnitude
W
M 7.25 and 5.0 earthquakes, sitetosource distance in the range of 1 to 100
km, and 0.2second spectral acceleration. Note that different attenuation
functions will return different values of spectral acceleration for a given
combination of earthquake magnitude and distance.
An acceleration response spectrum can be developed for the site and a given
combination of earthquake magnitude and distance by repeating the
calculation for many values of period across the period range of interest
(typically 0.01 to 5 second). Since different attenuation relationships will
provide somewhat different estimates of the median value of a spectral
response acceleration parameter, multiple attenuation relationships can be
used to generate a response spectrum, with appropriate weights given to each
attenuation relationship.
1
See Appendix A for information on the characteristics of lognormal
distributions.
April 15, 2009
ATC58 5: Ground Shaking Intensity 511
0 20 40 60 80 100
Distance (km)
0
0.5
1
1.5
2
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
T=0.2 sec
B_A, Mw=7.25
C_B, Mw=7.25
C_Y, Mw=7.25
B_A, Mw=5.0
C_B, Mw=5.0
C_Y, Mw=5.0
5.5.3 Median Spectrum and Dispersion
The attenuation relationships presented above return the median value Y of
the spectral quantity of interest, which is typically a geometric mean (or
geomean) spectral acceleration of two horizontal components
2
at a user
specified period. As noted previously, the dispersion (or scatter around the
median) is given by
lnY
o or  . Both the median spectral demand and the
dispersion are used for scenariobased assessments.
Some attenuation relationships use shearwave velocity in the upper 30 m as
an input variable; others are presented for one or more reference shear wave
velocities (e.g., 760 m/sec). If the selected attenuation relationship has an
input variable for shear wave velocity in the upper 30 m and the shear wave
velocity in the upper 30 m of the soil column has been computed per Chapter
20 of ASCE 7, use the sitespecific value to compute the median spectral
demand.
If the selected attenuation relationship uses a generic description of the
reference soil type (e.g., soft rock) and reports the reference shear wave
velocity, the median spectral demand can be factored by the ratio of the
appropriate site class coefficients for the sitespecific and reference shear
wave velocities
2
For a given pair of horizontal earthquake histories, the geometric mean
(
g
S of the spectral ordinates of the two components ( S
x
and
y
S ) is
generally used to characterize the pair of histories:
g x y
S S S = .
April 15, 2009
512 5: Ground Shaking Intensity ATC58
Sidebar: Median spectrum and dispersion that illustrate the variability
in earthquake shaking. The figure below illustrates the wide dispersion in
spectral demand. The figure shows median spectral demand for a
W
M 7.25
earthquake and a sitetosource distance of 5 km for the BooreAtkinson
NGA relationship together with 3
rd
, 16
th
, 84
th
and 97
th
percentiles (i.e.
probabilities of nonexceedance) of spectral demand.
0 1 2 3 4 5
Period (sec)
0
0.5
1
1.5
2
2.5
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
BooreAtkinson model,
Mw=7.25, r=5 km
97th
84th
Median (50th)
16th
3rd
5.6 Hazard Characterization for Nonlinear Response
History Analysis
5.6.1 Introduction
This section presents procedures to develop pairs of ground motion histories
suitable for intensity, scenario and timebased assessments. Appendix
Section E.8 presents an alternate procedure to develop pairs of ground
motions for scenariobased assessments, wherein the dispersion in ground
motion is considered explicitly.
For intensity and timebased assessments, seismic hazard can be represented
using either a Uniform Hazard Spectrum (UHS) or a Conditional Mean
Spectrum (CMS), both of which are described in Appendix E. Intensity
based assessment may use a spectrum that is neither a uniform hazard or
conditional mean spectrum. In the presentation below, building orthogonal
horizontal axes are denoted as X and Y and the corresponding fundamental
translational periods are denoted
1
X
T and
1
Y
T , respectively.
The ground motion scaling procedures associated with uniform hazard
spectra make use of a range of period (
max min
, T T ). Consistent with Chapter 16
of ASCE 705,
max
T may be taken as 1.5 times the larger of the fundamental
translational period of the building in either direction (i.e., the greater of
1
X
T and
1
Y
T ) and
min
T may be taken as
max
0.2T . If substantial response,
damage and loss accrue at periods less than
max
0.2T (e.g., in highrise
April 15, 2009
ATC58 5: Ground Shaking Intensity 513
buildings where multiple higher modes contribute significantly to the
acceleration and drift response, and buildings incorporating equipment
mounted at grade, which will experience ground shaking and not building
response) or
max
0.2T is similar in value to the smaller of
1
X
T and
1
Y
T , a period
smaller than
max
0.2T should be used to define the low end of the period range.
Scaling pairs of motions to a uniform hazard spectrum across a wide range of
periods will likely overestimate damage and loss for buildings with widely
separated modal periods, or responses in two or more modes along each axis
of the building. Response may also be overestimated for buildings that
undergo large inelastic deformations (with associated increases in period)
under high intensity ground motions.
The ground motion scaling procedures associated with conditional mean
spectra make use of an effective fundamental period, T . For buildings with
uncoupled first translational modes of response, the effective fundamental
period can be taken as the average of
1
X
T and
1
Y
T . For buildings with coupled
translational modes of response, the effective fundamental period should be
set equal to the fundamental period. The use of conditional mean spectra
assumes that response, damage and loss accrue primarily in the first
translational modes along each building axis. This assumption may be
unreasonable for tall buildings, where lower modes may dominate base
overturning moments and higher modes may dictate base shear forces,
buildings in which expensive and vulnerable equipment is mounted at grade,
where the excitation is ground motion (zeroperiod spectral response) rather
than floor motion, and mid to high rise buildings with a disproportionately
high percentage of vulnerable nonstructural components and content in the
lower stories where higher mode accelerations may dominate the acceleration
demands.
Reader Note: The conditional mean spectrumbased procedures should be
used with care if the ratio of
1
X
T and
1
Y
T falls outside a narrow range, which
is set equal to (0.8, 1.2) in this draft of the Guidelines.
If losses accrue in multiple modes along each axis of the building, multiple
conditional mean spectra should be developed using a procedure similar to
that presented in USNRC Regulatory Guide 1.165. Responsehistory analysis
would be performed using ground motions scaled to each conditional mean
spectrum and resultant losses would either be summed or weighted to
compute the total loss for the chosen intensity of motion. Procedures for
doing this are still being developed.
April 15, 2009
514 5: Ground Shaking Intensity ATC58
5.6.2 Selection of Earthquake Ground Motions
Pairs of earthquake ground motions should be selected from records having
faulting mechanism, earthquake magnitude, sitetosource distance and local
geology that are similar to those that dominate the seismic hazard. Bins of
ground motions are available from a number of sources, including
www.peer.berkeley.edu.
5.6.3 IntensityBased Assessment
Earthquake intensity is defined by a userspecified 5%damped, elastic
horizontal acceleration response spectrum, consistent with the characteristics
of the site on which the building is constructed. The spectrum may be a
uniform hazard spectrum, be constructed by anchoring a spectral shape to a
spectral acceleration at a userspecified period, or a conditional mean
spectrum.
For the first two representations of shaking intensity (i.e., uniform hazard
spectrum, anchored spectral shape), the earthquake hazard characterization
can involve the following steps:
1. Select a target response spectrum.
2. Select a suite of n pairs of ground motions; a minimum of 11 pairs of
motions is recommended.
3. Construct a geometric mean (or geomean) spectrum (see Section 5.5.3)
for each pair of ground motions.
4. Scale the suite of ground motion pairs such that within the period range
of
min
T
to
max
T the mean of the suite of geomean spectra approximates
the target spectrum of step 1.
Sidebar: When used in nonlinear response history analysis, each pair of
ground motions will result in a somewhat different prediction of the
magnitude of the response quantities used to assess performance. The
intent is to obtain an unbiased estimate of the structural response, given
the target spectrum, with limited error. Although the ASCE 705
procedures permit the use of suites of either 3 or 7 motions for analysis,
these motions will not provide sufficiently accurate estimates of median
response for use in loss assessment. A suite of 11 pairs of ground
motions is the minimum recommended by these Guidelines. Such a suite
will provide a 75% confidence that the predicted median response from
will be with + 20% of the true median value of response given the
spectrum (for an assumed dispersion of 0.5). Better estimates of the
median response can be achieved by using larger suites of motions.
April 15, 2009
ATC58 5: Ground Shaking Intensity 515
If earthquake intensity is represented by a conditional mean spectrum, which
will be constructed using a userspecified magnitude, distance, and c triple,
hazard characterization involves the following steps:
1. Construct the conditional mean spectrum per Appendix E.
2. Select a suite of n ground motion pairs (a minimum suite of 11 is
recommended), , 1,
i
p i n = with geomean spectra that are similar in shape
to the conditional mean spetrum over a userspecified range of period,
which must include
1
X
T and
1
Y
T and should be sufficiently broad to
account for period elongation due to inelastic response.
3. For each ground motion pair,
i
p , amplitude scale the geometric mean of
the two components to ( )
a
S T obtained from the conditional mean
spectrum.
5.6.4 TimeBased Assessments
Timebased performance assessments utilize seismic hazard curves, which
can be established for a building site using the USGS national seismic hazard
maps, described in Section 5.4.2, or by performing sitespecific probabilistic
seismic hazard analysis (PSHA), as described in Appendix E. This produces
a mean seismic hazard curve such as that shown in Figure 51. The mean
hazard curve includes an explicit consideration of ground motion uncertainty.
Appendix E provides summary information on the inclusion of uncertainty in
seismic hazard analysis.
The timebased assessment procedure requires selection of a range(s) of
spectral demands. The range should include intensities that produce building
response that causes negligible damage to complete loss. One acceptable
range is:
 Minimum
min
( ) 0.05
a a
S T S = = g for 1 T s second and 0.05/ T g
otherwise
 Maximum ( )
a
S T =
max
a
S =two times the spectral acceleration for an
annual frequency of exceedance =0.0004 (
4
4 10
÷
× ) but need not exceed
the median collapse capacity
ˆ
( )
a
S T per Chapter 7
where T is the effective fundamental period of the building, as defined
previously.
April 15, 2009
516 5: Ground Shaking Intensity ATC58
Reader Note: The range of spectral acceleration is preliminary and subject
to change. The utility of this approach will be reported in subsequent drafts
of these Guidelines. For new codecompliant buildings, collapse is not
anticipated for earthquakes with an annual frequency of exceedance of
0.00040 (return period of 2500 years). The upper bound value on spectral
acceleration was selected to drive codecompliant buildings to collapse (or
the brink thereof), based on studies conducted as part of the ATC63 project.
The writers recognize that older, nonductile buildings are much more likely
to collapse than modern codecompliant buildings and so shaking with an
annual frequency of exceedance of 0.01 might be sufficient to trigger
collapse, meaning that a number of the selected intensities of shaking might
trigger collapse. The user is also encouraged to change the range of spectral
accelerations, specifically the upper limit, so that no more than 2 of the 8
intensities of shaking trigger a collapse. The lower bound limit on spectral
acceleration should not result in damage to either structural or nonstructural
components of any vintage.
The chosen range(s) of spectral demand is split into m (suggested m=8)
intervals and the midpoint values of spectral acceleration in each of these m
intervals is used for ground motion scaling and response calculations. The
mean annual frequency of exceedance is computed for each midpoint value
of spectral acceleration and spectra are generated for each frequency of
exceedance.
Reader Note: The number of intervals required to produce stable
calculations of performance is a function of building type, occupancy and
location. On a preliminary basis, a number of 8 intervals suggested.
The sidebar illustrates calculation of the range of spectral demand; splitting
of the range into eight equal intervals,
1
e A through
8
e A ; calculation of the
midpoint values of spectral demand,
1
e through
8
e ; and the mean annual
frequency of spectral demand in each interval,
1
ì A through
8
ì A , which are
required for loss computations.
Sidebar: Sample calculation of spectral demands for a timebased
assessment using a seismic hazard curve. The seismic hazard curve below
illustrates the calculations associated with characterizing hazard for time
based assessment. The range of mean spectral acceleration was selected as
0.05 g to 1.23 g, where 1.23 g is the spectral acceleration (at a given period)
corresponding to a mean annual frequency of exceedance (MAFE) of 0.0002.
This range of mean spectral acceleration was split into eight equal intervals,
i
e A , of 0.1475 g, with the boundaries of 0.050, 0.198, 0.345, 0.493, 0.640,
0.788, 0.935, 1.083 and 1.23 g. The midpoint value in each interval
characterizes a target spectral demand for the scaling of ground motions. The
8 target spectral accelerations are identified in the figure by the symbol
April 15, 2009
ATC58 5: Ground Shaking Intensity 517
with values of (0.124, 0.271, 0.419, 0.566, 0.714, 0.861, 1.009, 1.156) g. The
mean annual frequency of spectral demand in each interval is computed as
the difference in the MAFEs at the boundaries of the interval (e.g., the mean
annual frequency of spectral acceleration (at 1.14 seconds) being between
0.788 and 0.935 g is 0.000499 [=0.001040.000543]). The eight
corresponding intervals of mean annual frequency,
1
ì A through
8
ì A , which
are required for timebased loss assessment, are 0.0452, 0.00775, 0.00318,
0.00163, 0.00050, 0.000497, 0.000199, and 0.000146, respectively.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Earthquake intensity, e (g)
0.0001
0.001
0.01
0.1
1
M
e
a
n
a
n
n
u
a
l
f
r
e
q
u
e
n
c
y
o
f
e
x
c
e
e
d
a
n
c
e
0.124 0.271 0.419 0.566 0.714 0.861 1.009 1.156
Ae
1
Ae
2
Ae
3
Ae
4
Ae
5
Ae
6
Ae
7
Ae
8
e
1
e
2
e
3
e
4
e
5
e
6
e
7
e
8
Aì
1
Aì
2
Aì
3
Aì
4
Aì
5
Aì
6
Aì
7
Aì
8
Two sets of procedures for selecting and scaling ground motion pairs are
presented below, the first for use with uniform hazard spectra and the second
for use with conditional mean spectra.
Earthquake hazard characterization using uniform hazard spectra involves
the following steps:
1. Develop a mean seismic hazard curve at T and adjust for soil type.
2. Compute the mean spectral accelerations
min
a
S and
max
a
S .
3. Split the range of mean spectral acceleration,
min
a
S to
max
a
S g, into m
intervals; identify the midpoint mean spectral acceleration in each
interval and the corresponding mean annual frequency of exceedance.
4. Calculate the mean annual frequency of spectral demand in each interval
from the hazard curve and record these frequencies for input to PACT.
5. Develop a uniform hazard spectrum for each mean annual frequency of
exceedance of step 3.
April 15, 2009
518 5: Ground Shaking Intensity ATC58
6. For each uniform hazard spectrum, select and scale suites of n ground
motions per steps 2 through 4 of Section 5.6.3 (intensitybased
assessment).
This procedure will results in n×m pairs of scaled earthquake histories.
Sidebar: Pairs of scaled ground motions. The spectral shapes at multiple
intensities will likely be similar and pairs of ground motions scaled to one
intensity of shaking could be scaled to an alternate intensity as well, thereby
substantially reducing the number of pairs of ground motions required.
Characterization of earthquake hazard using conditional mean spectra
involves the following steps:
1. Conduct steps 1 through 4 above for the uniform hazard spectra.
2. For each of the m midpoint mean spectral accelerations, ( )
ai
S T , 1, , i m =
deaggregate the spectral acceleration to determine the corresponding
modal [magnitude, distance, c] triple and generate the conditional means
spectrum.
3. For each of the m midpoint mean spectral accelerations, ( )
ai
S T ,
1, , i m = select suites of n pairs of ground motions, , 1,
ij
p j n = with
geomean spectra that are similar to the conditional means spectrum over
the userspecified range of period, which must include
1
X
T and
1
Y
T and
which should be sufficiently broad to account for period elongation due
to inelastic response.
4. For each of the m midpoint mean spectral accelerations, ( )
ai
S T , 1, , i m =
amplitude scale the geometric mean spectral acceleration of the two
components for each
ij
p , to ( )
ai
S T .
This procedure will results in n×m pairs of scaled earthquake histories.
5.6.5 ScenarioBased Assessments
For scenariobased assessment, seismic performance and loss are computed
for a userspecified combination of earthquake moment magnitude and site
tosource distance. Attenuation relationships are used to predict median
spectral demands and dispersions as a function of period. Section 5.5
provides introductory information on these relationships.
Scenariobased assessments utilize distributions in spectral demand for the
purpose of scaling earthquake ground motions for responsehistory analysis.
Ground motion intensity is assumed to be lognormally distributed with
median value u and dispersion . In this subsection, one procedure is
presented for generating pairs of ground motions in which dispersion is not
April 15, 2009
ATC58 5: Ground Shaking Intensity 519
considered in the development of pairs of ground motions but addressed
indirectly in the response computations of Chapter 6. An alternate procedure
in which dispersion is considered explicitly is presented in Appendix Section
E.8.
Identical to the procedures of Section 5.6.3, the user must specify a period
range (
max min
, T T ) for scaling pairs of ground motions. The characterization of
the earthquake hazard can involve the following steps:
1. Select the magnitude and sitetosource distance for the scenario event.
2. Select an appropriate attenuation relationship(s) for the region, site soil
type and source characteristics.
3. Establish the period range (
max min
, T T ) for scaling pairs of ground
motions.
4. Construct a median spectrum in the period range (
max min
, T T ) using the
attenuation relationship(s) of step 2.
5. Select n pairs of ground motions for amplitude scaling (a minimum of 11
is recommended).
6. Scale the set of n pairs of ground motions such that within the period
range of
min
T
to
max
T , the mean of the n geomean spectra approximates
the target spectrum of step 4.
This will result in n pairs of earthquake histories for use in response analysis.
A refinement to the above 6step procedure involves the use of conditional
mean spectra and the effective fundamental period T as defined in Section
5.6.1. Steps 4 through 6 above are altered per the discussion in Section 5.6.3.
5.7 Hazard Characterization for Use with Simplified
Analysis
5.7.1 Introduction
This section of the Guidelines presents acceptable procedures for
characterizing earthquake hazard for simplified linear analysis. Procedures
are presented in the following subsections for intensity, time and scenario
based assessments. The representations of seismic hazard are consistent with
those presented above for use with nonlinear responsehistory analysis and
the use of uniform hazard spectra. Refinements to the hazard representations
could be made but such improvements and the additional computational
effort are not warranted for simplified analysis.
April 15, 2009
520 5: Ground Shaking Intensity ATC58
5.7.2 IntensityBased Assessment
Earthquake intensity is defined by a 5%damped, elastic horizontal
acceleration response spectrum. Only values of the spectral response
acceleration are required for response computations. An intensitybased
assessment can be performed for any userspecified spectrum.
The characterization of the earthquake hazard involves the following steps:
1. select a response spectrum
2. determine the values of the spectral response acceleration at the periods
of the first translational mode along each axis of the building, namely,
1
X
T and
1
Y
T
5.7.3 TimeBased Assessment
Time based assessment makes use of seismic hazard curves adjusted for soil
type per Section 5.3. The seismic hazard curve can be established using the
USGS ground motion calculator, where linear interpolation can be used to
establish spectral demands at a period between those available in the
calculator. Site specific seismic hazard analysis can also be used.
Analysis is performed at m values (8 is recommended) of spectral
acceleration, ( )
a
S T , assumed here to be mean ( )
a
S T . The m values of
spectral acceleration should span the range of interest from no damage to
complete loss. One acceptable range is
 Minimum
min
( ) 0.05
a a
S T S = = g for 1 T s second and 0.05/ T g
otherwise.
 Maximum ( )
a
S T =
max
a
S =two times the spectral acceleration for an
annual frequency of exceedance =0.0004 (
4
4 10
÷
× ) but need not exceed
the median collapse capacity
ˆ
( )
a
S T per Chapter 7
The characterization of the earthquake hazard involves the following steps:
1. develop a mean seismic hazard curve at T and adjust for soil type
2. compute the mean spectral accelerations
min
a
S and
max
a
S
3. split the range of mean spectral acceleration,
min
a
S to
max
a
S g, into m
intervals; identify the midpoint mean spectral acceleration in each
interval and the corresponding mean annual frequency of exceedance
4. calculate the mean annual frequency of spectral demand in each interval
from the hazard curve and record these frequencies for input to PACT
April 15, 2009
ATC58 5: Ground Shaking Intensity 521
5. develop a uniform hazard spectrum for each mean annual frequency of
exceedance of step 3
6. for each of the m uniform hazard spectra, determine the values of the
spectral response acceleration at the periods of the first translational
mode along each axis of the building, namely,
1
X
T and
1
Y
T .
The product of step 6 will be 2m values of spectral acceleration for response
analyses.
5.7.4 ScenarioBased Assessment
The characterization of the earthquake hazard involves the following steps:
1. select the magnitude and site source distance for the scenario event
2. select an appropriate attenuation relationship for the region, site and
source characteristics
3. determine the median values of the spectral response acceleration at the
periods of the first translational mode along each axis of the building,
namely,
1
X
T and
1
Y
T .
April 15, 2009
ATC58 6: Response Analysis 61
Chapter 6
Response Analysis
6.1 Scope
This chapter presents acceptable structural analysis procedures to determine
seismic response of buildings, and seismic demands on structural and
nonstructural components and building contents. The procedures in this
chapter are not suitable for evaluation of building collapse and their
application is limited as outlined in Sections 6.2 and 6.3. Procedures for
predicting building collapse are presented in Chapter 7.
Section 6.2 presents nonlinear responsehistory analysis procedures. Section
6.3 presents a simplified analysis procedure. Both sections describe the
response data required for input to PACT.
6.2 Nonlinear ResponseHistory Analysis
6.2.1 Introduction
This section presents guidance on nonlinear responsehistory (dynamic)
analysis of building structures.
6.2.2 Mathematical Models of Components and Elements
Buildings should be modeled, analyzed and assessed as a threedimensional
assembly of components, including foundation and geocomponents as
appropriate. Nonstructural components that contribute significant stiffness
and/or strength should be modeled explicitly. The following subsections
present guidance on mathematical modeling of structural and nonstructural
components, and systems.
Structural Components
Nonlinear representation of forcedeformation behavior should typically be
used for all structural and geocomponents. A linear elastic component
representation can be substituted for a nonlinear component model only if
elastic response is anticipated and confirmed for the selected intensity of
shaking. Nonlinear models of components should be based on test data where
available. Strength and stiffness degradation in all components and elements
should be modeled explicitly unless response is insufficient to induce
degradation. Otherwise, hysteretic models must, as a minimum, account for
stiffness and strength degradation effects implicitly.
April 15, 2009
62 6: Response Analysis ATC58
Component forcedeformation relationships
Element and component models will generally be enveloped by the basic
form of the forcedisplacement relationships shown in Figure 61, where the
generalized force (axial, shear or moment), Q , is normalized by the yield
force,
y
Q .. In general, median values should be used to characterize the
forcedeformation parameters.
Sidebar: Mean versus median values. A goal of analysis is to determine
the central tendency of the response of the building and its components to
earthquake shaking. If the probability distributions of the parameters of the
forcedeformation relationship (e.g.,
y
Q ) are known, each parameter should
be defined using its median value. For parameters whose variability can be
represented by a normal distribution, the median value is equal to the mean
or expected value. Mean (expected) values are considered reasonable to
estimate the central tendency where insufficient information is available to
accurately characterize the distribution. However, where the parameters are
known to have lognormal or other (nonnormal) distributions, it is more
appropriate to use median rather than mean values.
The Type 1 curve is representative of ductile behavior characterized by an
elastic range (points 0 to 1 on the curve) and a plastic range (points 1 to 3).
The plastic range includes a strain hardening or softening range (points 1 to
2) and a strengthdegraded range with nonnegligible residual strength to
resist lateral and gravity loads (points 2 to 3). Loss of lateralforceresisting
capacity occurs at point 3, which is followed by loss of verticalforce
resisting capacity, which may occur at point 4 or beyond.
The Type 2 curve is representative of limited ductile behavior where there is
an elastic range (points 0 to 1 on the curve) and a plastic range (points 1 to
3), followed by loss of lateralforceresisting capacity at point 3 and loss of
verticalforceresisting capacity at point 4.
The Type 3 curve is representative of brittle or nonductile behavior where
there is an elastic range (points 0 to 1 on the curve) followed by loss of
lateralforceresisting capacity at point 3 and loss of verticalforceresisting
capacity at point 4.
April 15, 2009
ATC58 6: Response Analysis 63
Figure 61 Generalized forcedisplacement behavior of structural components
The forcedisplacement relationships described in Figure 61 address
deterioration of strength and stiffness in an approximate manner. More
rigorous hysteretic models are available and others are being developed for a
wide variety of structural components and construction materials. In general,
the responses of more rigorous hysteretic models are expected to provide
similar overall component responses to the generalized forcedisplacement
relationships that are calibrated to test data, but may provide substantially
different response for loading histories that are not well approximated by the
test loading protocol. Accurate implementations of such hysteretic models
should enable improved estimates to be made of peak transient story drifts,
residual story drifts and acceleration responses through to the point of
incipient building collapse. Information on models that address deterioration
of strength and stiffness can be found in ATC721 (2009), ATC63 (2009),
Haselton and Deierlein (2007), Ibarra et al. (2005), Ibarra and Krawinkler
(2005), Lignos (2008), Zareian and Krawinkler (2007).
Sidebar: Generalized forcedeformation relationship. The definition of
a generalized forcedeformation relationship for a structural component will
depend on the loading protocol used to derive the data and how the backbone
parameters are extracted from the test data. As shown in the figure below
from Gatto and Uang (2002), forcedeformation curves can generally be
distinguished between
those based on monotonic
loading versus cyclic
loading, where the latter
are usually defined as an
envelope of the cyclic
response. The force
deformation relationships
for seismic assessment by
nonlinear static analysis,
0.8Fc

F
c

L
o
a
d
Displacement
0.8Fc

F
c

L
o
a
d
Displacement
0.8Fc

F
c

L
o
a
d
Displacement
Backbone curve
Cyclic skeleton curve
April 15, 2009
64 6: Response Analysis ATC58
such as those presented in ASCE 41 (ASCE, 2006), are typically defined
based on a cyclic skeleton curve, where the components have been tested
under a standard loading protocol such as given by ATC24 (ATC, 1992).
Although cyclic skeleton curves are generally conservative, the cyclic
skeleton curves may be overly conservative for predicting structural collapse
under impulsive (near field) ground motions. Moreover, the internal forces
calculated based on the cyclic skeleton curves may underestimate the force
demands on forcecontrolled components. Therefore, it is generally desirable
to employ inelastic component models whose backbone curve evolves from
the initial monotonic curve to a degraded cyclic curve. Otherwise, where the
inelastic component model does not simulate explicitly such degradation, the
backbone curve must be defined to incorporate an appropriate amount of
cyclic degradation to represent the degradation that is expected to occur
under earthquake shaking.
Floor diaphragms
Floor diaphragms should be modeled using appropriate stiffness
assumptions. A diaphragm can be classified as rigid if the maximum lateral
deformation of the diaphragm is less than one half of the average story drift
above and below the subject diaphragm. Diaphragms that do not meet this
criterion and diaphragms in buildings containing outofplane offsets in the
vertical seismic framing system should not be considered rigid and should be
modeled explicitly using finite elements. Mass should be distributed across
the footprint of diaphragms to capture torsional response. If diaphragms are
incapable or remaining elastic at the response levels investigated, nonlinear
representation should be used.
Soilfoundationstructure interaction
Nonlinear models should be used to represent the foundation system
components where soilfoundationstructure interaction is judged to have a
significant effect on building response. Simplifying assumptions, such as
composite elements composed of vertical grouped foundation components
(e.g., piles associated with a single pile cap) can be used for nonlinear
response analysis. The hysteretic properties of each element under
generalized loadings should be developed based on test data and soil
mechanics principles. Foundations likely to uplift or slide should be modeled
to capture these effects.
Nonlinear models of the soil in the immediate vicinity of a building should
be included in the numerical model if (a) the soilstructurefoundation
April 15, 2009
ATC58 6: Response Analysis 65
interaction substantially modifies the freefield demand, or (b) the soil is
prone to liquefaction or ground failure.
Guidance on modeling soilfoundationstructure systems is presented in
Appendix E, together with a discussion of the likely impact of soil
foundationstructure interaction as a function of building and foundation type
and geometry, soil type and depth to bedrock, etc.
Foundation embedment
Seismic hazard assessments are routinely conducted for sites that are
assumed to be in the free field, namely, sites far removed from buildings or
other structures. Depending on the site and the building dynamic
characteristics, freefield motions may differ substantially from those
experienced by the embedded foundation of a building, with the greatest
differences for deep foundations and/or soft soil sites. Consideration should
be given to embedded foundations in terms of basement and foundation
modeling and characterization of seismic input, where the seismic input may
be generated by the propagation of ground motion from a rock outcrop at
depth through the soil column below the embedded foundation. Preliminary
guidance is provided in Appendix E.
Models of Nonstructural Components
Nonstructural components that contribute significant lateral strength and/or
stiffness to the building, such as infill masonry walls, should be modeled
using equivalent beamcolumn or strut elements, or appropriate finite
elements. Numerical models of nonstructural components should be based on
cyclic test data or other supporting evidence.
6.2.3 Analysis Procedures
Nonlinear responsehistory analysis should be performed using validated
software and best estimate structural and nonstructural component and
system models as described in Section 6.2.2.
Nonlinear analysis is performed to compute median values of demand
parameters (component deformations and forces, story drift, and floor
accelerations). As described below, the corresponding dispersions in the
demand parameters are usually estimated through judgments about the
uncertainty in calculating the response.
April 15, 2009
66 6: Response Analysis ATC58
Minimum Number of Pairs of Ground Motions for Estimating Median
Responses
For intensitybased and scenariobased assessments, nonlinear response
history analyses are conducted for n pairs of ground motions, which are
scaled to appropriate intensity values using the procedures of Section 5.6. It
is recommended that n be taken as 11 or more, to provide a suitably accurate
estimate of the mean response parameters.
Sidebar. Number of pairs of ground motions. Analysis performed using
eleven pairs of ground motions is expected to provide a good estimate
( 20 ± %) of median response (e.g., deformations, story drift, etc.) with
reasonable confidence (75% confidence, assuming the dispersion in the
response is of the order of 0.5, Huang et al. 2008). Larger suites of motions
will provide more accurate estimates of the median and smaller suites, and
less accurate estimates of the median response.
For timebased assessments, nonlinear analyses are conducted for m sets of a
n ground motion pairs each, scaled to appropriate intensity values using the
procedures of Section 5.6.4. It is recommended that m be taken as having a
value of 8 unless it is demonstrated that suitable stable results can be
obtained by a lesser number of intensities.
Dispersion in response due to modeling
Statistics (median and dispersion) to characterize the displacement and
acceleration response for a given intensity of earthquake shaking can be
established by either use of best estimate (deterministic) structural and
nonstructural component models to calculate the median response and a user
specified dispersion for modeling uncertainty, or development and
probabilistic analysis of a large family of building models whose component
mechanical properties are random variables with userspecified distributions.
Although probabilistic analysis is more rigorous than the bestestimate
deterministic analysis, the computational effort is increased by orders of
magnitude. Therefore, the use of the deterministic bestestimate analysis is
assumed here.
The dispersion in response for a given intensity of earthquake shaking due to
uncertainty in the component models should account for variations due to
building definition and quality assurance,
c
 , and the quality and
completeness of the nonlinear analysis model,
q
 . These sources of
uncertainty are grouped together and termed modeling uncertainty,
m
 . The
April 15, 2009
ATC58 6: Response Analysis 67
total modeling dispersion can be estimated using equation (61) and the
information presented in the tables below.
2 2
m c q
   = + (61)
Building definition and construction quality assurance. The value of the
dispersion,
c
 , is assigned on the basis of the level of definition of the
building and how well the design intent is (or will be) implemented through
close adherence to the construction documents. Table 61 provides
recommended values for
c
 under representative conditions. Performance
analysts should use individual judgment and guidance provided by the table
to assess
c
 for individual projects.
Table 61 Default descriptions and values for
c

Definition and Construction Quality Assurance
c

Building definition: The building is completely designed and well
represented by drawings and specifications (i.e., construction
documents are available).
Construction quality: The building was or will be constructed
using rigorous construction quality assurance measures, including
special inspection, materials testing and structural observation
.10
Building definition: The building is completely designed or nearly
completely designed but available drawings do not completely
define all details of construction (i.e., design development
documents or equivalent are available).
Construction quality: For existing buildings, the structure is sited
in a region and is of an age in which rigorous construction quality
assurance measures were likely implemented. For new
construction, the building will be constructed using reasonable
construction quality assurance measures.
0.25
Building definition: The design is incomplete (i.e., schematic
documents are available), or, for an existing building is known
only on the basis of limited field observation and testing.
Construction quality: For new or existing buildings, little
information on the construction quality assurance program is
available.
0.40
Model quality and completeness: The value of the dispersion,
q
 , should be
assigned on the basis of the completeness of the mathematical model and
how well the structural component deterioration and failure mechanisms are
understood and implemented. The dispersion should be selected on the basis
of the user’s understanding of how sensitive the response is to certain
structural parameters (e.g., strength/stiffness, deformation capacity, incycle
and between cycle degradation of structural components) and the likely
April 15, 2009
68 6: Response Analysis ATC58
degree of inelastic response and deterioration. Table 62 provides guidance
on the selection of
q
 values.
Table 62 Default descriptions and values for
q

Model Quality and Completeness
q

Model quality: The numerical model for each component is
robust over the anticipated range of displacement or deformation
response. Strength and stiffness deterioration and all likely failure
modes are modeled explicitly. Model accuracy is established with
data from largescale component tests through failure.
Completeness: The mathematical model includes all structural
components and nonstructural components in the building that
contribute strength and/or stiffness.
.10
Model quality: The numerical model for each component is
robust over the anticipated range of displacement or deformation
response. Strength and stiffness deterioration is fairly well
represented, though some failure modes are simulated indirectly.
Accuracy is established through a combination of judgment and
largescale component tests.
Completeness: The mathematical model includes most structural
components and nonstructural components in the building that
contribute significant strength and/or stiffness.
0.25
Model quality: The numerical model for each component is
based on cyclic skeleton curve models of ASCE 41 or comparable
guidelines, where strength and stiffness deterioration and failure
modes are incorporated indirectly.
Completeness: The mathematical model includes all structural
components in the lateralforceresisting system.
0.40
Sidebar: Estimating dispersion
q
 . The selection of a value for
q
 from
Table 62 should be guided by the expected degree of inelastic response and
likelihood of component failure and building collapse. For example, the
component models in a building that is shaken in the elastic range only (i.e.,
no structural damage) need not address component deterioration and failure
modes need not be simulated; a value of 0.1
q
 = might be appropriate in
this instance. Conversely, if collapse simulations are being performed, robust
numerical models are required for all components expected to suffer
significant damage. If such models have not been validated by largescale
testing, model confidence should not be high, and a value of
q
 in the range
of 0.3 to 0.5 is likely appropriate.
April 15, 2009
ATC58 6: Response Analysis 69
Table 63 Default Dispersions for RecordtoRecord Variability, Modeling Uncertainty
and Ground Motion Variability
1
, T T
(sec)
1
1
( )
=
a
y
S T W
S
V

A a

aa

m

gm
1
WNA CEUS PNW
0.20
s 1.00 0.05 0.10 0.25
0.60 0.53 0.80
2 0.35 0.10 0.25
4 0.40 0.10 0.35
6 0.45 0.10 0.50
> 8 0.45 0.05 0.50
0.35
s 1.00 0.10 0.15 0.25
0.60 0.55 0.80
2 0.35 0.15 0.25
4 0.40 0.15 0.35
6 0.45 0.15 0.50
> 8 0.45 0.15 0.50
0.5
s 1.00 0.10 0.20 0.25
0.61 0.55 0.80
2 0.35 0.20 0.25
4 0.40 0.20 0.35
6 0.45 0.20 0.50
> 8 0.45 0.20 0.50
0.75
s 1.00 0.10 0.25 0.25
0.64 0.58 0.80
2 0.35 0.25 0.25
4 0.40 0.25 0.35
6 0.45 0.25 0.50
> 8 0.45 0.25 0.50
1.0
s 1.00 0.15 0.30 0.25
0.65 0.60 0.80
2 0.35 0.30 0.25
4 0.40 0.30 0.35
6 0.45 0.30 0.50
> 8 0.45 0.25 0.50
1.50
s 1.00 0.15 0.35 0.25
0.67 0.60 0.85
2 0.35 0.35 0.25
4 0.40 0.30 0.35
6 0.45 0.30 0.50
> 8 0.45 0.25 0.50
2.0+
s 1.00 0.25 0.50 0.25
0.70 0.60 0.90
2 0.35 0.45 0.25
4 0.40 0.45 0.35
6 0.45 0.40 0.50
> 8 0.45 0.35 0.50
1. WNA = Western North America; CEUS = Central and Eastern United States; PNW =
Pacific North West. Values established using the attenuation relationships of Boore and
Atkinson (2008) and Campbell and Bozorgnia (2008) for WNA, Campbell (2003) for
CEUS, and Atkinson and Boore (2003) for the PNW; moment magnitudes between 6.5
and 7; and rock sites
April 15, 2009
610 6: Response Analysis ATC58
Table 63 presents default values of modeling dispersion,
m
 , based on the
data presented in Tables 61 and 62. Values are presented as a function of
effective fundamental period, T , and a strength ratio, S , which is given by
1
( )
a
y
S T W
S
V
= (62)
where
1 y
V is the (estimated) story yield strength at the base of the building,
and all other terms have been defined above or in Chapter 5. Linear
interpolation can be used to estimate  for intermediate values of
1
T and S.
Alternate values can be derived for performance assessment.
Sidebar: Estimation of values for
m
 in Table 63. The default values for
m
 in Table 63 were computed using equation (61) as follows. Values were
rounded to the closest multiple of 0.05.
a) The dispersion
c
 was set equal to 0.25, which assumes that design
development (DD) information is available to construct the mathematic
model. (If only scheme level information are available, use 0.4
c
 = .)
b) For [1,2] R = , 0.10
q
 = because near elastic response is anticipated.
c) For [4] R = 0.25
q
 = because inelastic response is anticipated but the
likelihood of strength and stiffness deterioration is low.
d) For [6,8] R = , 0.40
q
 = because significant inelastic response is
anticipated and the likelihood of strength and stiffness deterioration is not
low.
Dispersions in response due to ground motion recordtorecord
variability
In these Guidelines, pairs of earthquake ground motion records are scaled to
a value of spectral acceleration at an effective fundamental period. A suite of
earthquake ground motion pairs scaled in this manner will produce
distributions of responses in elastic and inelastic (i.e., nonlinear) multi
degreeoffreedom systems. (The dispersion will only be zero for elastic
single degreeoffreedom systems.) Unless a very large number of ground
motion pairs, on the order of 20 or more, are used to predict median response
parameter values, dispersion values obtained from the suites of analyses will
generally not accurately represent the true dispersion due to recordtorecord
variability. For this reason, these Guidelines recommend the use of default
values of recordtorecord variability, denoted
a
 . The default values
presented in columns 2 and 3 of Table 63 were established using data
presented in Huang et al. (2009) and RuizGarcia and Miranda (2003).
Values are presented for story drift (
a

A
) and floor acceleration (
aa
 )
April 15, 2009
ATC58 6: Response Analysis 611
responses and vary as a function of effective fundamental period and strength
ratio. Alternate values may be used.
Dispersions in ground motion conditioned on magnitude and distance
Chapter 5 introduced attenuation relationships that describe the distribution
in spectral demand as a function of moment magnitude, sitetosource
distance and other seismological parameters. Table 63 lists default values
for dispersion in ground motion spectral demand,
gm
 , computed using
widely accepted attenuation relationships for Western North America
(WNA); the Central and Eastern United States (CEUS); and the Pacific North
West (PNW), assuming subductionzone earthquakes. The values are
independent of effective fundamental period and strength ratio. Alternate
values of
gm
 can be used. Dispersion
gm
 is used only for scenariobased
assessments.
6.2.4 Residual Drift
Experience from past earthquakes suggests that residual (permanent) story
drift can substantially influence repair cost and, in some cases, the viability
of repair. Prior research, which is summarized in Appendix H, indicates that
residual drifts predicted by nonlinear analysis are highly variable and
sensitive to the assumed component models including the postyield
hardening/softening slope and unloading response and often are not a good
representation of the actual residual drifts that could occur. Residual drifts
may be larger for nearfault records in the forward directivity region.
Accurate statistical simulation of residual drifts requires the use of advanced
component models with careful attention paid to cyclic hysteretic response
and a large number of ground motion pairs, greater than the minimum
specified in these Guidelines.
If advanced component models are not used and/or the minimum number (11
or fewer) of ground motions pairs are used for analysis, equation (63) can be
used to estimate the median residual story drift, which is denoted
r
A :
0
0.3( ) 4
( 3.1 ) 4
r y
r y y y
r y y
A = A s A
A = A ÷ A A < A < A
A = A ÷ A A > A
(63)
where A is the median peak transient story drift calculated by analysis and
y
A is the median story drift at yield calculated by analysis. The dispersion
in the residual drift is generally larger than that of the peak story drift.
Unless substantiating data are available, the total dispersion in the residual
April 15, 2009
612 6: Response Analysis ATC58
drift due to both recordtorecord variability,
a
 , and modeling,
m
 , should
be set equal to 0.8 (RuizGarcia and Miranda, 2005).
Sidebar: Calculation of yield drift ratio. For the purpose of calculating
residual drifts, the yield drift ratio is defined based on the onset of significant
yielding in the structure. For concentrically steel braced frame systems (or
for squat wall systems), the yield drift can be calculated as the story drift
associated with imposed story shear forces equal to the expected yield
strength of the braces (or the expected shear strength of the walls). For
moment frame systems, the yield drift can be calculated as the story drift
associated with story shear forces that cause (a) the beams and/or columns to
reach their expected plastic moment capacity, taking into account the effect
of axial forces in the members, or (b) the beamcolumn joint panel to reach
its expected yield strength. The distribution of forces in the story under
consideration can be calculated either based on the forces induced using
equivalent lateral forces, or alternatively, by applying a shear force at the
floor above the story being evaluated with a lateral support imposed at the
floor below the story being evaluated. In shear wall systems whose behavior
is dominated by flexural yielding, the yield drift is equal to the story drift
associated with the wall reaching its expected flexural strength under a lateral
load distribution that reflects the dynamic response of the building.
6.2.5 Response Data Input to PACT
IntensityBased Assessment
For intensitybased assessment, the maximum absolute value of each demand
parameter of interest, story drift in each direction, and floor acceleration at
each level are assembled into a k×n matrix, where k is the total number of
demand parameters and n is the number of response history analyses that ran
to completion (i.e., did not trigger collapse). This matrix is input to PACT.
The dispersions in the drift response,
IBA

A
, and acceleration response,
IBAa
 , are computed as follows:
2 2
IBA a m
  
A A
= + ;
2 2
IBAa aa m
   = + (64)
where
a

A
,
aa
 and
m
 are presented in Table 63. These dispersions are
input to PACT.
ScenarioBased Assessments
For scenariobased assessment, the maximum absolute value of each demand
parameter of interest, story drift in each direction, and floor acceleration at
each level are assembled into a k×n matrix, where k is the total number of
April 15, 2009
ATC58 6: Response Analysis 613
demand parameters and n is the number of response history analyses that ran
to completion. This matrix is input to PACT.
The dispersions in the drift response,
SBA

A
, and acceleration response,
SBAa
 , are computed as follows:
2 2 2
SBA a m gm
   
A A
= + + ;
2 2 2
SBAa aa m gm
    = + + (65)
where
gm
 is the dispersion in ground motion per Table 63 and other terms
are defined above. These dispersions are input to PACT.
TimeBased Assessments
For timebased assessment, m sets (one per spectral intensity interval per
Section 5.6.4) of k×n matrices of the maximum absolute value of each
demand parameter of interest are assembled for input to PACT, together with
the corresponding mean annual frequency of shaking in each spectral
interval.
The dispersions in the drift response,
TBA

A
, and acceleration
response,
TBAa
 , at the given intensity of shaking (and the corresponding
value of S) are computed as:
2 2
TBA a m
  
A A
= + ;
2 2
TBAa aa m
   = + (66)
where no allowance is made for uncertainty in ground motion because this
uncertainty is already captured in the hazard calculations used to derive the
mean hazard curve. These dispersions are input into PACT for each intensity
of shaking.
6.3 Simplified Analysis
6.3.1 Introduction
The simplified procedures in this section use linear mathematical structural
models, an estimate of the lateral yield strength of the building frame, and a
simplified analysis procedure to estimate responses at global, story, floor,
and component levels. Structural framing is assumed to be uncoupled along
each horizontal axis (denoted X and Y in Chapter 5). Separate analysis is
performed along each axis.
The simplified procedure is based on a series of assumptions including: (a)
the building framing is independent along each horizontal axis, permitting
the framing systems along each axis to be uncoupled and torsional response
to be ignored; (b) the building is regular in plan and elevation (i.e., there are
no substantial discontinuities in lateral strength and stiffness), which was
April 15, 2009
614 6: Response Analysis ATC58
assumed for the mathematical models used to develop the simplified
procedure; (c) story drifts do not exceed 4 times the corresponding yield
drift, so that the assumptions of elasticplastic (bilinear) response at the
component level is not compromised by excessive degradation in strength
and stiffness; (d) the story drifts are less than 4%, beyond which P Aeffects
may become important; and (e) the building is less than 15 stories in height.
If the building being analyzed does not fall within this design space, the
simplified procedure should not be used.
The use of linear mathematical models and the simplified method of analysis
provide values of maximum story drift and peak floor acceleration that are
representative of the median estimates of response values. Residual story
drifts are estimated using equation (63). However, the use of these
simplified procedures entails much greater uncertainty both as to the true
value of these medians and also as to the dispersion associated with the
demands. As such dispersions in the drift and acceleration response
computed using simplified procedures should be greater than those
associated with the nonlinear procedures of Section 6.2 and will result in
projections of performance that entail greater uncertainty than when the
nonlinear methods are used. Guidelines for calculating the dispersion for the
simplified procedure are presented in later sections.
6.3.2 Mathematical Models of Components and Elements
For meaningful assessment, the mathematical model of the building must
represent the distribution of mass, lateral strength and lateral stiffness in the
building. The following subsections provide guidance on modeling structural
and nonstructural components for assessment using the simplified analysis
procedure.
Structural Components and Elements
All elements that contribute significantly to either the building’s lateral
strength or stiffness should be considered in developing the model, whether
or not these elements are considered to be structural or nonstructural, or part
of the seismic forceresisting system. ASCE Standard 41 provides guidance
for modeling the strength and stiffness of typical building elements.
The numerical model and framingsystem data are used to establish the first
mode period and shape of the building (in the direction under consideration),
which is required to compute the pseudo lateral forces acting on the building
frame and the resulting story drifts. An estimate of the structure’s lateral
yield strength is also required, which can be estimated by either plastic
analysis or, for a new codecompliant building, using the response
April 15, 2009
ATC58 6: Response Analysis 615
modification factor, R and the strength factor,
0
O (see sidebar below).
Nonlinear static or plastic analysis should provide a better estimate of the
yield strength at the base of the building than that calculated using R and
0
O .
The affects of vertical earthquake shaking and soilfoundationstructure
interaction is neglected. The response of seismically isolated structures
should be assessed using the equivalent linear procedures of ASCE 41.
Sidebar: Estimation of yield strength. For a new building structure, the
building codes require design for a minimum strength given by a base shear
coefficient, calculated as the product of the spectral acceleration at the first
mode period of the building and the reactive weight of the building, divided
by the quantity (R/I), where is R is the response modification coefficient and
I is the occupancy importance factor. Most framing systems will have
strength in excess of the minimum value for a number of reasons, including:
construction using materials that are stronger than the minimum specified
strength, members that are oversized to meet drift limitations, resistance
factors used to calculate design strengths, and redundancy in the frame. The
factor
0
O , which varies by framingsystem type, is a crude measure of the
likely strength in the frame in excess of the design strength and is expressed
as the ratio of the maximum strength to design strength in the building frame.
The effective yield strength of a building frame is typically between 1.25 and
0
O times the codebased design strength. For performance assessments to
be meaningful, it is important that the strength of the structure be determined
accurately. Use of R and
0
O factors to assess building strength is not
encouraged. Instead, yield strength should be determined by nonlinear static
analysis.
Models of Nonstructural Components
Nonstructural components that contribute significant mass, lateral strength
and/or lateral stiffness to the building frame should be modeled with a level
of detail that is consistent with the mathematical models adopted for the
structural frame. Components that contribute only mass need not be modeled
explicitly aside from assigning reactive mass to the appropriate location.
6.3.3 Analysis Procedure
The simplified elastic analysis procedure described in this section can be
used to estimate story drifts and floor accelerations. Two analyses are
performed, one along each horizontal axis of the building.
Figure 62 presents the definitions of floor and story numbers and story
height adopted for use with the analysis procedures of this section.
April 15, 2009
616 6: Response Analysis ATC58
Base
Roof, Floor N+1
Floor N
Story N
Floor N1
Story N1
Floor 3
Floor 2
Story 2
Floor 1
Story 1
h
h
h
h , H
1
2
N1
N
Figure 62 Definition of floor and story numbers and story height
Pseudo Lateral Force
A pseudo lateral load, V, is used to compute median story drifts. The load V
is computed using equation (67) and distributed over the height of the
mathematical model as described below.
1 2 1 1
( )
a
V C C S T W = (67)
where
1
C is an adjustment factor for inelastic displacements;
2
C is an
adjustment factor for cyclic degradation;
1
( )
a
S T is the 5% damped spectral
acceleration at the fundamental period of the building in the direction under
consideration, for the selected level of ground shaking; and
1
W is the first
modal weight of the building but cannot be taken as less than 80% of the
total reactive weight, W.
The first modal weight,
1
W , can be calculated as
1
1
2
1
2
1 1
2
2
( )
j
N
j j
j
N
j
j
w
W
w


+
=
+
=
=
¿
¿
(68)
where
j
w is the lumped weight at floor level j (see Figure 62),
1 j
 is the jth
floor ordinate of the first mode deflected shape and N is the number of floors
in the building above the base. In the absence of calculation per (68), and
subject to the 80% limit,
1
W can be set equal to
sm
C W , where
sm
C is defined
in ASCE 41.
The adjustment factor
1
C is computed as:
April 15, 2009
ATC58 6: Response Analysis 617
1 1
1 2
1
1
1
1 for 0.2sec
0.04
1
1 for 0.2 1.0sec
=1 for 1.0sec
S
C T
a
S
T
aT
T
÷
= + s
÷
= + < s
>
(69)
where
1
T is the fundamental period of the building in the direction under
consideration, a is a function of the soil site class per Chapter 5 (=130, 130,
90, 60 and 60 for site classes A, B, C, D and E, respectively) and S is a
strength ratio given by equation (62) with the effective fundamental period
replaced by
1
T .
The adjustment factor
2
C is computed as:
2
2 1
2
1 2
1
1
( 1)
1 for 0.2sec
32
1 ( 1)
1 for 0.2 0.7sec
800
=1 for 0.7sec
S
C T
S
T
T
T
÷
= + s
÷
= + < s
>
(610)
If the building incorporates viscous or viscoelastic damping devices, the 5%
damped spectral acceleration ordinate should be replaced by the value that
corresponds to the first mode damping in the building. The procedures of
ASCE 41 can be used to estimate the damping in a building equipped with
viscous or viscoelastic energy dissipation devices and to adjust the response
spectrum for the effects of damping other than 5%.
Median Peak Transient Story Drift Calculation
For a given spectral acceleration at the first mode period of the building, the
story drifts,
*
i
A , are computed in three steps:
1. Distribute the pseudo lateral force over the height of the building.
2. Apply the lateral forces to the elastic model of the building frame,
compute the floor displacements and then compute the uncorrected story
drifts,
i
A , as the difference in the displacements of adjacent floors.
3. Correct the story drifts to account for inelastic action and higher mode
effects
1
.
Each step in the drift calculation procedure is described below.
1
The corrections of step 3 are based on analysis of 9 buildings of varying
construction with first mode periods ranging between 0.2 and 2.5 seconds.
See the project technical note by Huang et al. (2009).
April 15, 2009
618 6: Response Analysis ATC58
The pseudo lateral force, V, is distributed over the height of the building to
compute the lateral loads at floor level x,
x
F :
x vx
F C V = (611)
using the vertical distribution factor,
vx
C :
1
1
1
2
k
x x
vx N
k
j j
j
w h
C
w h
÷
+
÷
=
=
¿
(612)
where
j
w is the lumped weight at floor level j;
1 j
h
÷
(
1 x
h
÷
) is the height above
the effective base of the building to floor level j; and k is equal to 2.0 for a
first mode period greater than 2.5 seconds and equal to 1.0 for a first mode
period less than or equal to 0.5 second (linear interpolation can be used for
intermediate periods).
The lateral forces,
x
F , are applied to the linearly elastic model of the building
frame to compute the uncorrected floor displacements and story drifts,
i
A .
Correct the story drifts,
i
A , to compute demands for performance
assessment,
*
i
A , as follows:
*
1
( , , , )
i i i i
H S T h H
A
A = A (613)
where
1
( , , , )
i i
H S T h H
A
is the drift correction factor for story i that is
computed using equation (614) as a function of S; the first mode period,
1
T ;
the height of the floor immediately above the story for which the drift is
being calculated, as defined in Figure 62,
i
h ; and the total height of the
building, H:
2
0 1 1 2 3 4
ln ( )
i i
i
h h
H a a T a S a a
H H
A
= + + + + 1, S i > = 1 to N (614)
where the values of the coefficients
0
a through
4
a are presented in Table 64
for braced frame, momentframe and shearwall buildings.
Table 64 StoryDrift and FloorAcceleration Correction Factors
Frame type
0
a
1
a
2
a
3
a
4
a
S
t
o
r
y
D
r
i
f
t
Braced 0.72 0.048 0.012 2.64 2.09
Moment 0.65 0.027 0.010 2.58 2.30
Wall 1.12 0.22 0.059 2.70 1.29
F
l
o
o
r
a
c
c
e
l
e
r
.
Braced 0.57 0.16 0.089 0 0
Moment 0.67 0.28 0.080 0 0
Wall 0.33 0.22 0.081 0.53 0
April 15, 2009
ATC58 6: Response Analysis 619
Median Residual Story Drift Calculation
Residual story drifts are calculated using the peak transient story drifts
computed above and equation (63). Note that the simplified elastic
procedure should not be used for computing story drifts in excess of four
times the yield story drift. The yield story drift should be computed as
discussed previously.
Median Floor Acceleration Calculation
For a given peak ground acceleration (filtered at a frequency of 25 Hz), the
peak acceleration at floor level i,
*
i
a , is computed as:
*
i ai i
a H a = i = 2 to N+1 (615)
where
1
( , , , )
ai i
H S T h H is the acceleration correction factor for floor i (see
Figure 62) calculated using equation (616); and
i
a is the peak ground
acceleration. The acceleration at the base of the building, floor level 1 per
Figure 62, shall be set equal to the peak ground acceleration. The
acceleration correction factor
2
is given by
1
0 1 1 2 3
ln
i
ai
h
H a a T a S a
H
÷
= + + + 1, S i > = 2 to N+1 (616)
where the values of the coefficients
0
a through
3
a are presented in Table 64
for braced frame, momentframe and shearwall buildings.
Dispersions in response calculations
To conduct performance assessments, it is necessary to develop distributions
of story drift and floor acceleration, where the distributions capture the
uncertainty in ground motion,
gm
 , analysis,
a

A
and
aa
 , and
modeling,
m
 . Table 63 lists default values of the dispersion to be used with
the simplified analysis procedure. Linear interpolation can be used to
estimate  for intermediate values of
1
T and S.
6.3.4 Response Data Input to PACT
For intensitybased and scenariobased assessment, one vector of absolute
values of median maximum demand parameters is input to PACT, together
with the dispersions for drift and acceleration computed per Section 6.3.3
using equations (64) and (65), respectively.
2
The acceleration correction factors are based on analysis of 9 buildings of
varying construction with first mode periods ranging between 0.2 and 2.5
seconds; see Huang et al. (2009) for details.
April 15, 2009
620 6: Response Analysis ATC58
For a timebased assessment, one family of eight vectors of absolute values
of median maximum demand parameters, one per spectral interval, are input
to PACT, together with the corresponding dispersions for drift and
acceleration and the mean annual frequencies of shaking in each spectral
interval.
April 15, 2009
ATC58 7: Procedures for Collapse Assessment 71
Chapter 7
Procedures for Collapse
Assessment
7.1 Scope
This chapter presents two alternative procedures to establish building
specific collapse fragility functions that relate the probability of building
collapse to a ground motion intensity parameter. The fragility functions are
defined in the form of a lognormal distribution defined by a median value
ˆ
( )
a
S T and a dispersion  .
7.2 Collapse Analysis
7.2.1 Introduction
Collapse fragilities can be developed for any building using Incremental
Dynamic Analysis (IDA) procedure described herein. Collapse fragilities for
lowrise buildings can also be approximated using a simplified procedure,
also described below.
Collapse fragility functions expresses the probability of collapse conditioned
on a ground motion intensity parameter: ( ( )
a
P c S T . The intensity parameter
is spectral acceleration at the first mode period or averaged first mode period
(T ). The collapse fragility curve is assumed to follow a lognormal
distribution and so is fully defined by a median value (
ˆ
( )
a
S T , the median
collapse capacity) and dispersion, .
Building collapse is generally associated with failure of the gravityload
framing system, either locally or globally. There are few examples of
buildings overturning, aside from those that suffered soil or foundation (pile)
failures. Excessive story displacements, which can lead to a nearcomplete
loss of lateral stiffness, can trigger a failure of both vertical and horizontal
components, whether such components are included in the threedimensional
building model or not. In these procedures, collapse is defined as the more
critical of the following two limit states:
1. sidesway failure as characterized by loss of lateral stiffness and
associated excessive building drifts, and
2. loss of vertical load carrying capacity of gravity framing
members due to earthquakeinduced building drifts.
April 15, 2009
72 7: Procedures for Collapse Assessment ATC58
Vertical collapse is not simulated directly because currently available
numerical tools are unable to do so. However, loss of vertical capacity in
gravity framing components is included in the collapse assessment and
development of collapse fragility functions.
7.3 Mathematical Models for Collapse Analysis
7.3.1 Introduction
Two types of mathematical models can be used for collapse analysis: (1) a
best estimate model, and (2) a simplified best estimate model.
7.3.2 Best Estimate Models
Bestestimate models should be used for collapse assessment. A best
estimate model involves the use of median properties of all force
deformation parameters (e.g., yield strength, yield deformation, postyield
stiffness, maximum or capping strength, deformation corresponding to
maximum strength) for each component in the building.
Sidebar: Best estimate models should be based on median values of the
model parameters, since this is consistent with calculation of the median
collapse capacity. However, it is not unreasonable to use mean (or expected
values) in place of the median values since (a) many of the modeling
parameters have a normal distribution, in which case the mean and median
are equal, (b) given the scarcity of available data for many of the parameters,
it is not possible to discern the difference between median and mean values,
and (c) given the many other approximations and assumptions, the difference
between mean and median modeling parameters is not expected to have a
significant impact on the final results.
Where possible, collapse fragilities should be developed using complete
simulation models. A complete simulation model includes all structural
components in the building frame (regardless of whether they are considered
to be part of the lateralforceresisting or gravityloadresisting systems or
both), those nonstructural components that contribute strength and stiffness
to the building, and can simulate all possible failure modes in all modeled
components.
An incomplete simulation model is any multidegreeoffreedom model that
is not complete as described above. Use of an incomplete simulation model
for collapse fragility development requires the user to address outside of the
analytical model all components and/or modes of failure that are not directly
simulated (nonsimulated) in the development of the collapse fragility
parameters.
April 15, 2009
ATC58 7: Procedures for Collapse Assessment 73
7.3.3 Simplified Model
A simplified model can be prepared using the results of nonlinear static
analysis, as described below. Such models are approximate only and should
only be considered for analysis of buildings that are regular in plan and
whose response is dominated by the first translational mode along each
horizontal axis, thereby restricting its use to buildings of approximately 10
stories and less in height.
An equivalent singledegreeoffreedom (SDOF) model is prepared for each
horizontal axis of the building using nonlinear static (pushover) analysis.
Such analysis should be based on the use of a best estimate model of the
multidegreeoffreedom structure and account explicitly for stiffness and
strength degradation and P A effects. All possible failure modes should be
simulated.
The results of the nonlinear analysis are represented by two global force
displacement relationships, one per building axis. These relationships are
plotted as lateral displacement of a reference node (typically at the roof of the
building) versus the applied lateral force (or base shear). ASCE 41 presents
procedures that can be used to perform such analysis and to develop the
forcedisplacement relationship, commonly termed a pushover curve.
After the lateral forceroof displacement relationship has been developed,
several control points are extracted from the relationship to represent the
behavior of the building, namely,
1. Yield point for the structure: defined by the total applied force,
y
V , and
the displacement of the reference node,
y
A , at the point where
substantial yielding has occurred in the framing.
2. Capping (peak force) point: defined by the maximum applied force in the
nonlinear analysis,
m
V and the corresponding displacement,
m
A .
3. Residualstrength point: defined by the postcapping strength
r
V and the
corresponding displacement
r
A associated with the onset of the residual
strength plateau.
4. Failure point: defined by force
r
V and displacement
u
A at which the
structure loses an ability to carry gravity forces, resulting either from
local failure modes or sidesway collapse.
Figure 71 illustrates the development of the global forceroof displacement
relationship. This relationship ( , ) V A can be converted into an equivalent
SDOF system
1 1
( , ) V A by setting
1
/
m
V V C = and
1 1
A = I A , where
m
C is the
effective first mode mass factor ( 1 s ) per ASCE 41, and
1
I is the first mode
participation factor ( 1 > ). (This transformation is similar to that performed
April 15, 2009
74 7: Procedures for Collapse Assessment ATC58
for analysis using the Capacity Spectrum Method.) Figure 71 shows the
equivalent SDOF representation of the global forcedisplacement
relationship, computed assuming that 0.75
m
C = and
1
1.25 I = .
Sidebar:The factor
m
C is a firstmode mass participation factor and
represents the fraction of the structure’s mass that is effective in the first
mode. The factor
1
I is the first mode participation factor and relates the
displacement of the reference node to the first mode spectral displacement.
These factors can be computed as follows:
2
1
1 1
n
i i
i
m n n
i i i
i i
w
C
w w


=
= =
(
(
¸ ¸
=
¿
¿ ¿
1
1
n
i i
i
n
r i
i
w
w


=
=
¿
I =
¿
where w
i
is the weight at level i, 
i
is the displacement of level i when the
structure is displaced in the first mode shape and 
r
is the first mode
displacement of the reference node. For regular structures, 1/
m
C I = .
Two forcedisplacement relationships are typically required for building
collapse analysis, one per axis, producing a twodegreeoffreedom (TDOF)
model. The (first mode) periods along each axis of the TDOF model,
denoted
1
X
T and
1
Y
T in Chapter 5, are used for the calculation of T and
selection and scaling of ground motions for collapse assessment.
Figure 71 Forcedisplacement relationships
April 15, 2009
ATC58 7: Procedures for Collapse Assessment 75
7.4 Analysis and Modeling Considerations
7.4.1 Three and TwoDimensional Models
Threedimensional mathematical models should be used for collapse
assessment of buildings.
Twodimensional (planar) models may be appropriate for developing
fragility curves for buildings with regular geometries where the response in
each orthogonal direction is uncoupled. The SPO2IDA tool of Section 7.6.5
assumes response in one plane only. If twodimensional models are used, the
building collapse fragility should be determined from the weaker of the two
directions.
Elements of the gravity system should be incorporated into the mathematical
model for collapse assessment if such elements contribute appreciable lateral
strength and stiffness to the building. Irrespective of whether modeled or not,
the collapse assessment must account for potential collapse of the gravity
framing. Assessment of the collapse potential of the gravity framing system
in an incomplete simulation model may be accomplished using limit state
criteria for nonsimulated collapse modes as described in Section 7.6.2.
7.4.2 Mathematical Models of Structural Components
Component models should incorporate nonlinear properties for all elements
that cannot be shown to remain fully elastic in the nearcollapse condition.
Guidance for modeling structural components across the entire range of
response, from elastic behavior through complete loss of resistance, is
presented in Section 6.2.2.
Ideally, nonlinear models of building components should simulate all or
nearly all possible modes of deterioration and failure (e.g., axial, flexure,
flexureaxial interaction, shear, and flexureshear interaction). All such
modes of deterioration and failure are denoted as simulated. If a mode of
deterioration or failure is not included in a component model, its effect on the
results of a collapse analysis must be accounted for indirectly as a non
simulated deterioration mode. Some building components (e.g., gravityload
resisting columns) may not be included in the building model but must be
checked to either ensure that they do not fail at story drifts smaller than those
predicted by analysis at incipient building collapse, or the results of the
analysis must be adjusted to account for the nonsimulated failures. Non
simulated deterioration in components included in the mathematical model
and components not included in the mathematical model should be assessed
using appropriate collapse (loss of loadcarrying capacity) limit state checks
against demand parameters.
April 15, 2009
76 7: Procedures for Collapse Assessment ATC58
7.4.3 Damping
Energy dissipation will occur in a building prior to the onset of inelastic
structural response and damage. Such dissipation in structural and
nonstructural components may be modeled through equivalent viscous
damping and implemented in a variety of ways, including Rayleigh damping
and modal damping. For typical buildings, damping under earthquake
motions is generally assumed to range between 2% and 5% of critical,
depending on the building height, construction materials, and nonstructural
components, with smaller values expected for highrise, steelframed
buildings and larger values for lowrise, concrete framed buildings. Care
must be exercised so as not to overestimate the effects of equivalent viscous
damping following yielding of the structural frame because hysteretic
damping is included explicitly in the nonlinear analysis.
7.5 Ground Motion Characterization for Collapse
Assessment
Incremental Dynamic Analysis (IDA) enables the user to plot maximum
story drift versus spectral acceleration at the effective first mode period of
the building for each of a number of ground motions. Figure 72 presents the
results of a sample IDA for 22 different ground motions, where the
horizontal axis is maximum drift in any story of the sample building and the
vertical axis is spectral acceleration at a first mode period (2.2 seconds in this
example). Collapse intensity can be characterized by a lognormal
distribution, as shown by the probability density plot on the righthand
vertical axis of the figure.
Figure 72 Results of Incremental Dynamic Analysis
Two methods are presented below for selecting and scaling pairs of ground
motions for use in the IDA method of collapse assessment. One method uses
April 15, 2009
ATC58 7: Procedures for Collapse Assessment 77
uniform hazard spectra and the other method uses conditional mean spectra,
both of which are described in Chapter 5 and Appendix E.
In IDA, each ground motion pair is scaled to a geomean spectral response
acceleration for the building, at which the building will remain linear and a
response history analysis is performed to determine the peak drift in each
story. Then, the geomean spectral acceleration for each ground motion pair
at period T should be incremented a number of times and response history
analysis repeated, until collapse is triggered. For each ground motion pair,
the maximum drift in any story along either axis should be plotted as a
function of geometric mean spectral acceleration. Collapse can be inferred to
occur where small increment in spectral acceleration produces large
increment in spectral displacement or when demands predicted by the
analysis indicate that nonsimulated collapse modes will occur.
Method A – Uniform Hazard Spectra: A minimum of eleven pairs of
ground motions with appropriate spectral shape are recommended for IDA at
each increment of ( )
a
S T . The following procedure should be used to select
the appropriate pairs of motions:
1. estimate the median collapse capacity,
ˆ
( )
a
S T and the mean annual
frequency of exceedance of
ˆ
( )
a
S T
2. develop a Uniform Hazard Spectrum consistent with the mean annual
frequency of exceedance of step 1
3. select a minimum of 11 pairs of ground motions with geometric mean
spectra that have a similar shape to the Uniform Hazard Spectrum of step
2
If the estimate of median collapse capacity in step 1 is far removed from the
value computed by IDA, use the updated value of median collapse capacity
to estimate a new mean annual frequency of exceedance and repeat the
ground motion selection and IDA.
Each pair of motions is used twice for IDA: once in the original coordinate
system and once in a 90degree rotated coordinate system, producing a
minimum of 22 sets of ground motion pairs.
Method B – Conditional Mean Spectrum: A minimum of eleven pairs of
ground motions with an appropriate spectral shape are used for IDA at each
increment of ( )
a
S T . The following procedure should be used to select the
appropriate pairs of motions:
April 15, 2009
78 7: Procedures for Collapse Assessment ATC58
1. estimate the magnitude, distance and c [M, r, c] associated with the
median collapse capacity,
ˆ
( )
a
S T ; the value can be selected by
deaggregation of the ( )
a
S T hazard as described in Appendix E.
2. generate a Conditional Mean Spectrum (CMS) per Appendix E for the
[ , , ] M r c of step 1 using an appropriate attenuation function; normalize
the CMS so that the peak ordinate is equal to 1.0
3. select a minimum of 11 pairs of ground motions with geometric mean
spectra that have a similar shape and period at peak spectral response to
the CMS.
If the estimate of median collapse capacity in step 1 is far removed from the
value computed by IDA, use the updated value of median collapse capacity
to estimate a new [ , , ] M r c triple and repeat the ground motion selection and
IDA.
Each pair of motions is used twice for IDA: once in the original coordinate
system and once in a 90degree rotated coordinate system producing a
minimum of 22 sets of motions.
7.6 Development of Collapse Fragility Curves
7.6.1 Introduction
Collapse fragility curves can be developed from the results of Incremental
Dynamic Analysis (IDA) or using the SPO2IDA tool described in Section
7.6.5.
7.6.2 Truncation of IDA Curves for NonSimulated Failure Modes
and Components
If an incomplete simulation model is used for nonlinear responsehistory
analysis, the results of each IDA must be adjusted to accommodate the
collapse of nonsimulated components (e.g., components of the gravityload
resisting system) and modes of deterioration or failure not explicitly
addressed in modeled components (e.g., shear failure in a column modeled as
a beamcolumn element with lumped plastic flexural hinges at each end).
The limiting deformation associated with the failure of each nonsimulated
component and nonsimulated failure mode must be established and used to
correct each IDA as required.
April 15, 2009
ATC58 7: Procedures for Collapse Assessment 79
Sidebar: The inclusion of a nonsimulated mode of failure in the results of
IDA is illustrated in the example below for which column shear failure was
not simulated.
The results of the IDA ignoring
column shear failure are shown by
the solid blue line that indicates a
collapse capacity of 0.48 g. Analysis
of the shear capacity of the columns
in the building indicates failure at a
drift of 2.8%, which is associated
with a spectral acceleration of 0.42 g
and smaller than 0.48 g. In this
instance, the collapse capacity is
adjusted downwards to 0.42 g to capture the nonsimulated deterioration (or
failure). A similar procedure could be used to assess components not
included in the mathematical model.
7.6.3 Construction of a Collapse Fragility Curve
The results of IDA for 2n pairs of ground motions (n pairs × 2 orientations),
after adjustment for nonsimulated modes are mined to generate a vector of a
minimum of 2n threshold spectral accelerations that correspond to the
collapse limit state. Generally, these are identified from analyses that result
in either sidesway instability, where the maximum drift in any story and
either horizontal direction increases in an unbounded manner for no
significant increase in spectral acceleration, or loss in vertical load carrying
capacity, where the collapse limit state deformation demand is exceeded in
one or more components of the gravity load resisting system. For collapse
analysis using simplified models per Section 7.3.3, two values of drift are
reported per groundmotion simulation, one per axis.) The threshold
accelerations are shown as open circles in Figure 72. The median value of
spectral acceleration at collapse, namely, the median collapse capacity, can
be computed for this example by organizing the 22 values of spectral
acceleration in ascending order and selecting the acceleration midway
between the 11th and 12th values. For analysis using 2n pairs of ground
motions, the median collapse capacity shall be the average of the n and (n+1)
values of spectral acceleration.
Alternately, if responsehistory analysis is performed in spectralacceleration
order (i.e., analysis is performed for all 2n ground motion pairs at one
intensity level before incrementing the spectral acceleration), the response
April 15, 2009
710 7: Procedures for Collapse Assessment ATC58
history analysis can be terminated once 50% collapses are observed at a
given value of ( )
a
S T because sufficient information will be available to
establish the median collapse capacity
ˆ
( )
a
S T .
Figure 73 presents the cumulative distribution function (CDF) associated
with lognormal distribution of the values of Figure 72. Note that the
dispersion in the CDF of Figure 73, which is based on analysis of a
mathematical model constructed using best estimate mechanical properties,
does not include all sources of randomness and uncertainty. These alternate
sources are identified below and implemented in the collapse fragility
function.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5
Collapse Spectral Acceleration, S
a
(T)
ˆ
( )
a
S T =0.26g
=0.33
Figure 73 Transforming results of IDA to a collapse fragility curve
7.6.4 Collapse Fragility Parameters
7.6.4.1 Introduction
The two collapse fragility parameters are the median (
ˆ
a
S , 50th percentile)
and logarithmic standard deviation or dispersion (  ). The value of 
can
either be taken as the default value given below or computed using the
procedure of Section 7.6.4.3.
7.6.4.2 Median
The computation of the median collapse capacity is described in Section
7.6.3.
7.6.4.3 Dispersion
The collapse fragility dispersion should account for three sources of
uncertainty: variation in collapse capacity among different ground motion
records,
a

A
, which is also termed recordtorecord uncertainty; variations
April 15, 2009
ATC58 7: Procedures for Collapse Assessment 711
due to building definition and construction quality assurance,
c
 ; and quality
and completeness of the nonlinear analysis model,
q
 . The second and third
sources of uncertainty are grouped together in Chapter 6 and termed
modeling uncertainty,
m
 . The dispersion can be estimated using equation
(71) and the information presented below. Alternately, a default dispersion
value of 0.6 can be used.
2 2 2 2 2
( )
a c q a m
     
A A
= + + = + (71)
a. Recordtorecord uncertainty. The dispersion in collapse capacity
among ground motion records,
a

A
, can either be selected from the
values presented as a function of structural period and geographic region
in column 3 of Table 63, or taken as 0.45 in the absence of computation.
b. Building definition and construction quality assurance. Table 6.1
provides default values for
c
 .
c. Model quality and completeness: Table 6.2 provides default values for
q
 . Given that few robust models that account for strength and stiffness
deterioration through component failure have been validated by large
size experimentation,
q
 will likely range between 0.3 and 0.5.
7.6.5 Tools for Collapse Analysis of Simplified, Planar Models
Vamvatsikos and Cornell (2005) developed an analysis tool, known as
SPO2IDA, to facilitate the development of incremental dynamic analysis
(IDA) curves. The spreadsheet tool, which is available for download at
http://www.ucy.ac.cy/~divamva/software.html, uses a userdefined global
forcedisplacement relationship, such as that presented in Figure 74, and a
database of results from largescale analyses of nonlinear SDOF models of
varying profiles to generate IDA curves, which include some recordto
record variability. This tool was developed by analysis of SDOF systems, so
its use should be limited to lowrise planar frames that are regular in both
plan and elevation, whose seismic response is dominated by the first
translational mode. The implementation of SPO2IDA is described below.
Sidebar: The SPO2IDA tool has not been formally validated. The user
should review the technical basis for the methodology as proposed by
Vamvatsikos and Cornell (2005) and sample results (e.g., Hanson (2007)) to
understand the limitations and utility of the tool.
To use the SPO2IDA tool, the global baseshear (V ) versus roof
displacement (A ) relationship is transformed to a relationship between a
normalized spectral acceleration and global displacement ductility. Figure 7
4 presents sample information. The spectral acceleration SA is computed as
April 15, 2009
712 7: Procedures for Collapse Assessment ATC58
/ ( / )
m
SA V C W g = , where all terms have been defined previously. The
spectral acceleration is normalized by a variable
y
S , which is equal in value
to SA at the first control point as seen in the figure. The global displacement
ductility, µ , is the ratio of the roof displacement A to the roof yield
displacement. The coordinate of the yield point (first control point) in the
SPO2IDA space is ( , / 1,1)
y
SA S µ = .
Figure 74 Global forcedisplacement relationship in the SPO2IDA space
Figure 75 presents sample output from the SPO2IDA tool, which includes
the results of the nonlinear static analysis per Figure 74 and the IDA
predictions, including the 16th, 50th (median) and 84th percentile
predictions. The median collapse capacity,
ˆ
a
S , for this example is
approximately 5.1
y
S . Since a best estimate model is used to derive the
relationship of Figure 74, dispersions due to recordtorecord variability and
modeling (see Section 7.6.4) shall be computed and used to derive the
collapse fragility curve. The default dispersion of 0.6 is recommended for use
with the SPO2IDA tool.
April 15, 2009
ATC58 7: Procedures for Collapse Assessment 713
Figure 75 Sample SPO2IDA results for the example of Figure 74
7.7 Collapse Descriptions for Casualty Assessments
To facilitate casualty assessments, a brief qualitative narrative should be
provided to describe the likely mode(s) and extent of the collapse. Each
narrative should include information on the dominant deterioration modes
and triggers of collapse, whether the collapse is dominated by sidesway or by
loss in vertical load carrying capacity, and whether the collapse is likely to
involve the entire building area or to affect just portions of the structure (e.g.,
select stories or regions of the plan area). Information on the likely stories
involved in the collapse should be provided, if available.
Collapse area ratios could be used to supplement the collapse fragility curves
with information on the portion of the building area that is likely to be
involved in the collapse. These ratios take the form of a probability equation
to predict the total amount of collapsed floor area as the sum of the fractional
collapse on each floor of the building. These fractions are presented
independently of the ground motion intensity using the format of Table 71
below. The sum of the probabilities in the table total 1.00. The fractional area
of collapse area is the sum of (probability)× (% of stories)× (% plan area).
Using the sample probabilities in Table 71 below, the expected fractional
area of the building involved in collapse would be 0.31.
The procedure described above can be used to develop buildingspecific
collapse descriptions for casualty assessments. In this case, the user should
revise the table based on the configuration of their building and must identify
the entries in each cell.
April 15, 2009
714 7: Procedures for Collapse Assessment ATC58
Table 71 Sample probabilities that the specified portion of a
building will be involved in the collapse
Amount of Plan Area
25% of
plan area
50% of
plan area
100% of
plan area
Number of
stories involved
in the collapse
25% of stories 0.05 0.15 0.05
50% of stories 0.1 0.3 0.1
100% of stories 0.05 0.15 0.05
April 15, 2009
ATC58 8: Example Applications 81
Chapter 8
Example Applications
8.1 Introduction
This chapter presents intensity, scenario and timebased performance
assessments of example buildings using the procedures presented in Chapters
4, 5 and 6 of these Guidelines. The purpose of these example applications is
to demonstrate the practical application of these procedures as they could be
implemented in design practice.
Reader Note: This example application makes use of a steel momentframe
building and nonlinear responsehistory analysis that is unchanged from prior
drafts of these Guidelines, pending completion of an updated version of the
Performance Assessment Calculation Tool (PACT). Subsequent drafts will
include updated examples and assessments of other types of buildings using
the new version of the software tool, and will illustrate the use of the
simplified analysis procedure.
8.2 Example Building #1
8.2.1 Introduction
This example application uses the building shown in Figures 81, 82, and 8
3 as a prototype. The building shown is located in Berkeley, California, near
the campus of the University of California. It is representative of modern
Class A office space in the Greater Bay Area of California. Construction
was completed in 2004. Some of the features of the actual building were
modified and simplified for application in this example. The key features of
the building model are:
Height: 3 stories; 14 ft. floor to floor; 42 ft total above grade; no
basement
Area: 22,736 sq. ft. per floor; 68,208 sq. ft. total (actual building
slightly larger)
Occupancy: General office space (B2)
April 15, 2009
82 8: Example Applications ATC58
Figure 81 Photograph of the example building
Figure 82 Schematic structural framing plan for example building (no scale)
April 15, 2009
ATC58 8: Example Applications 83
Figure 83 Elevation of typical steel moment frame for example building (no scale)
8.2.2 Seismic Hazard Characterization
Introduction
This section presents the characterizations of seismic hazard for the enhanced
intensity, scenario and timebased performance assessment of example
building #1. The building is located 1 km from the Hayward Fault on a soft
rock site that is classified as Site Class B per ASCE 705. The
latitude/longitude coordinates of the site, which are required for intensity
and timebased assessments, are 37.875º/122.267º. The fundamental
translational period of the building, which is required information for scaling
earthquake ground motions, is 1.14 seconds, along each principal axis.
Ground Motions for IntensityBased Assessment
An intensitybased assessment was performed for a 5% damped design
earthquake response spectrum for the site. The spectrum was developed per
the 2006 International Building Code (ICC, 2006) for Site Class B. The
spectral acceleration at a period of 1.14 seconds for this spectrum is 0.51 g.
The USGS ground motion calculator and the lat/long coordinates noted
above were used to compute the design earthquake spectral demand:
1
0.874 ,
M
S g =
1
0.582
D
S g = ,
1.14
0.582/1.14 0.51
D
S g = =
Table 81 lists the eleven pairs of nearfault ground motions (Bin 1) that were
randomly selected from the nearfault groundmotion bin for response
history analysis. For these pairs of motions, the moment magnitude ranges
April 15, 2009
84 8: Example Applications ATC58
between 6.2 and 7.6 and the sitetosource distance varies between 1 and 10.7
km. The eleven pairs of seed motions in Table 81 were amplitude scaled to
0.51 g at 1.14 seconds per Section 5.6.4 of these Guidelines. This bin of
scaled motions was denoted as Bin I1. Figure 84 presents the acceleration
response spectra for the 22 scaled motions.
Table 81 Seed Ground Motions for ResponseHistory Analysis
Designation Event Station
W
M r
NF3, NF4 Loma Prieta 1989
SAC 2/50 for Los
Angeles
7.0 3.5
NF7, NF8 Northridge 1994 6.7 6.4
NF13, NF14 Elysian Park 2 (simulated) 7.1 10.7
NF17, NF18
Palos Verdes 1
(simulated)
7.1 1.5
NF21, NF22
Cape Mendocino
04/25/92 18:06
89156 Petrolia 7.1 9.5
NF27, NF28 ChiChi 09/20/99 TCU068 7.6 1.1
NF35, NF36 Erzinkan 03/13/92 17:19 95 Erzinkan 6.9 2.0
NF41, NF42
Imperial Valley 10/15/79
23:16
942 El Centro Array
#6
6.5 1
NF43, NF44 Kobe 01/16/95 20:46 Takarazu 6.9 1.2
NF45, NF46
Morgan Hill 04/24/84
04:24
57191 Halls Valley 6.2 3.4
NF47, NF48 Northridge 1/17/94 12:31 24279 Newhall 6.7 7.1
0 1 2 3
Period (sec)
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
Figure 84 Spectral accelerations for Bin I ground motions
April 15, 2009
ATC58 8: Example Applications 85
Ground Motions for ScenarioBased Assessment
The moment magnitude associated with the rupture of both segments of the
Hayward Fault is approximately 7.0. The scenario selected for assessment of
the building was a moment magnitude 7 earthquake at a sitetosource
distance of 1 km.
The ChiouYoungs (CY) Next Generation Attenuation (NGA) relationship
was used to predict the spectral demand for the [
W
M , r ] pair because this
relationship predicts spectral demands at 1.0, 1.1 and 1.2 seconds (Chiou and
Youngs 2006). Table 82 presents the median spectral acceleration (u ) and
the logarithmic standard deviation (  ) at 1, 1.1 and 1.2 seconds predicted by
the CY NGA relationship for
W
M =7, r =1 km, strikeslip faulting and a
shear wave velocity
30
760 v = m/s. Figure 85 presents the median CY
spectral acceleration in the period range of 1 to 1.5 seconds for this
combination of variables and a curve for spectral acceleration inversely
proportional to period and anchored to the median CY spectral demand at 1
second (=0.47 g). The CY spectral demands are virtually identical to
0.47/T in this period range. Accordingly, for the scenariobased
assessment, ( 1.14) 0.47/1.14 0.41 T g u = = = and 0.64.  = Fault rupture
directivity effects were ignored for this example (but should not be ignored
for project applications.)
Per Section 5.6.5 of these Guidelines, 11 values of target spectral
acceleration are used to characterize the distribution of seismic demand for
scenariobased assessments using Equation (57).
Table 82 Spectral Demand Per the ChiouYoungs Relationship
Period
(sec)
u (g)

1 0.47 0.63
1.1 0.43 0.64
1.2 0.4 0.64
April 15, 2009
86 8: Example Applications ATC58
1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4 1.45 1.5
Period (sec)
0
0.1
0.2
0.3
0.4
0.5
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
0.47
0.43
0.4
0.36
0.34
0.31
CY NGA relationship
for M=7 and r=1 km
1/T curve
Figure 85 Spectral accelerations per the ChiouYoungs NGA relationship for
W
M =7, r = 1
km, strikeslip faulting and
30
v = 760 m/s, and varying as 0.47 T
The eleven target spectral accelerations are (0.14, 0.20, 0.25, 0.30, 0.35, 0.41,
0.48, 0.55, 0.66, 0.83, 1.22) g per Equation (57). Per step 5 of the scaling
procedure of Section 5.6.5, a pair of motions was randomly selected from
Table 71 and each component was amplitude scaled to one of the eleven
target values. The process was repeated 10 times for the remaining target
spectral ordinates and groundmotion pairs in Table 81. The resultant scaled
motions were denoted as Bin S1 motions. Figure 86 presents the spectra for
the Bin S1 ground motions, the 16
th
, 50
th
and 84
th
percentiles of spectral
acceleration and the eleven target spectral ordinates (denoted ×).
Figure 87 presents the 16
th
, 50
th
and 84
th
percentiles of spectral accelerations
for the Bin S1 ground motions and those predicted by the CY NGA
relationship for
w
M =7, r =1 km, strikeslip faulting and
30
760 v = m/s.
The distribution of spectral demand for Bin S1 ground motions is comparable
to that predicted by the CY relationship for periods greater than 0.6 second.
For periods less than 0.6 second, the median spectral accelerations are lower
than those predicted by the CY relationship.
April 15, 2009
ATC58 8: Example Applications 87
0 1 2 3
Period (sec)
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
Bin S1 ground motions
16
th
, 50
th
and 84
th
percentiles
11 target spectral ordinates
Figure 86 Spectral accelerations for Bin S1 ground motions, 16
th
, 50
th
and 84
th
percentiles of
spectral acceleration and the 11 target spectral ordinates for Bin S1 motions
0 1 2 3
Period (sec)
0
0.5
1
1.5
2
2.5
3
3.5
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
16
th
, 50
th
and 84
th
percentiles of
spectral accelerations for
Bin S1 ground motions
CY NGA relationship
for M=7, r=1 km
Figure 87 Sixteenth, 50
th
and 84
th
percentiles of spectral acceleration for Bin S1 motions and
demands predicted by the ChiouYoungs NGA relationship
Ground Motions for TimeBased Assessment
The USGS ground motion parameter calculator was used to generate a
seismic hazard curve for the example building at a period of 1 second.
Figures 88 and 89 present a screen capture from the ground motion
calculator and the resultant hazard data, respectively.
April 15, 2009
88 8: Example Applications ATC58
Figure 88 Screen capture from USGS ground motion calculator for generating a
onesecond hazard curve for the site of the building
Figure 89 Onesecond seismic hazard curve
April 15, 2009
ATC58 8: Example Applications 89
Figure 85 shows that the median spectral demand predicted by the CY
NGA relationship
w
M =7 and r =1 km is inversely proportional to period
in the range of 1 to 1.5 seconds. Table 82 shows that the dispersion in
spectral demand for that scenario case is essentially constant between 1 and
1.2 seconds. On the basis of these results (and noting that it is unlikely that
the
W
M 7 earthquake at r =1 km will dictate the spectral demand across all
mean annual frequencies of exceedance), the 1.14second seismic hazard
curve for the site was generated by dividing the 1second values shown in
Figure 89 by 1.14. Figure 810 presents four curves. Curves A and D are the
1 and 2second seismic hazard curves, respectively, generated by the USGS
ground motion calculator for the site. Curves B and C were developed by
dividing the spectral demands for curve A by 1.14 and 2, respectively.
The spectral ordinates for curves C and D are virtually identical for spectral
acceleration less than 0.3 g. The spectral demands for curve C are slightly
greater than those for curve D for spectral acceleration greater than 0.3 g—
indicating that the 1/
a
S T · is slightly conservative in the period range of 1
to 2 seconds. Curve B was used for the timebased assessment of the
building. The values of spectral acceleration and the corresponding mean
annual frequencies of exceedance (MAFE) for curve B are listed in Table
83.
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Acceleration(g)
1x10
4
1x10
3
1x10
2
1x10
1
1x10
0
A
n
n
u
a
l
F
r
e
q
u
e
n
c
y
o
f
E
x
c
e
e
d
a
n
c
e
Curve A: T =1 s
Curve B: (T=1s)/1.14
Curve C: (T=1s)/2
Curve D: T =2 s
Figure 810 Seismic hazard curves for the site of the building
The seed ground motions in Table 81 were scaled per the procedure
described in Section 5.6.6 of the Guidelines for timebased assessments.
Figure 811 illustrates the procedure. The range of mean spectral acceleration
April 15, 2009
810 8: Example Applications ATC58
was selected as 0.05 g to 1.23 g, where 1.23 g is the spectral acceleration at a
period of 1.14 seconds corresponding to a MAFE of 0.0002. This range of
mean spectral acceleration was split into eight equal intervals,
i
e A , of
0.1475 g. The midpoint value in each interval characterizes a target spectral
demand for the scaling of ground motions. The 8 target spectral accelerations
are identified in Figure 811 by the symbol with values of (0.124, 0.271,
0.419, 0.566, 0.714, 0.861, 1.009, 1.156) g. All of the seed ground motions of
Table 81 were scaled to each of the 8 target spectral accelerations at a
period of 1.14 seconds. The 22 scaled ground motions at the 8 target spectral
accelerations,
i
e , were denoted as Bin Ti, i =1, 8. Figure 812 shows the 8
target spectral ordinates and the spectra for the Bin T1 and Bin T8 ground
motions.
Table 83 Mean Hazard Curve
MAFE Spectral acceleration (g)
4.64E01 1.75E03
3.99E01 3.51E03
3.27E01 5.26E03
2.57E01 7.02E03
1.94E01 1.14E02
1.44E01 1.67E02
1.04E01 2.54E02
7.40E02 3.77E02
5.19E02 5.61E02
3.58E02 8.42E02
2.38E02 1.26E01
1.47E02 1.89E01
8.17E03 2.84E01
3.90E03 4.27E01
1.54E03 6.40E01
4.72E04 9.56E01
1.01E04 1.44E+00
1.23E05 2.16E+00
1.63E07 3.24E+00
April 15, 2009
ATC58 8: Example Applications 811
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Earthquake intensity, e (g)
0.0001
0.001
0.01
0.1
1
M
e
a
n
a
n
n
u
a
l
f
r
e
q
u
e
n
c
y
o
f
e
x
c
e
e
d
a
n
c
e
0.124 0.271 0.419 0.566 0.714 0.861 1.009 1.156
Ae
1
Ae
2
Ae
3
Ae
4
Ae
5
Ae
6
Ae
7
Ae
8
e
1
e
2
e
3
e
4
e
5
e
6
e
7
e
8
Aì
1
Aì
2
Aì
3
Aì
4
Aì
5
Aì
6
Aì
7
Aì
8
Figure 811 Characterizing seismic hazard for timebased assessment
0 1 2 3
Period (sec)
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
Bins T1 and T8
Target spectral ordinates
for Bins T1 through T8
Figure 812 Target spectral ordinates for Bins T1 through T8 and spectra for scaled ground
motions in Bins T1 and T8
The 8 values of mean annual frequency of spectral demand in interval i,
i
ì A ,
shown in Figure 811, are required for timebased assessment. Losses are
computed at each of the 8 midpoint spectral demands and the annualized loss
is calculated as the sum of the products of the losses conditioned on the
intensity measure (midpoint spectral demand in each interval) and the
corresponding mean annual frequency of shaking in the interval. The 9
April 15, 2009
812 8: Example Applications ATC58
values of MAFE that define the 8 intervals of spectral acceleration are listed
in Table 84. The MAFE corresponding to a boundary between two ranges
can be computed using linear interpolation and the data listed in Table 83.
For example, the spectral boundary between
2
e and
3
e in Figure 811 is
0.345 g. The MAFE corresponding to 0.345 g can be computed using the
shaded entries in Table 83 as follows:
3
0.0039 0.00817
0.00817 (0.345 0.284) 0.00635
0.427 0.284
ì
÷
= + × ÷ =
÷
The 9 values of MAFE listed in Table 84 were calculated in this manner.
The mean annual frequency (MAF) of spectral demand in each interval is
computed as the difference in the MAFEs at the boundaries of the interval.
For example, the mean annual frequency of spectral acceleration at 1.14
seconds being between 0.788 and 0.935 g is 0.000497 (=0.001040.000543).
Table 85 lists the mean annual frequencies (MAF),
i
ì A , and bounding
spectral values (min, max) for each interval of shaking intensity.
Table 84 Spectral Accelerations and MAFE (
i
ì ) for
Boundaries on the Seismic Hazard Curve
i Spectral acceleration (g) ì
i
1 0.050 0.0593
2 0.198 0.0141
3 0.345 0.00635
4 0.493 0.00317
5 0.640 0.00154
6 0.788 0.00104
7 0.935 0.000543
8 1.083 0.000344
9 1.230 0.000198
April 15, 2009
ATC58 8: Example Applications 813
Table 85 Mean Annual Frequency (MAF) for Spectral
Intervals
Interval MAF Spectral range (g)
1
1
ì A 0.0452 0.050, 0.198
2
ì A 0.00775 0.198, 0.345
3
ì A 0.00318 0.345, 0.493
4
ì A 0.00163 0.493, 0.640
5
ì A 0.000500 0.640, 0.788
6
ì A 0.000497 0.788, 0.935
7
ì A 0.000199 0.935, 1.083
8
ì A 0.000146 1.083, 1.230
1. Range is [min, max] spectral acceleration.
8.2.3 ResponseHistory Analysis and Demand Parameter
Matrices
This section presents the matrices of demand parameters for the enhanced
intensity, scenario and timebased performance assessment of the example
building.
The scaled ground motion pairs in Bins I1, S1 and T1 through T8 were used
as input to the OpenSees model of the building for intensity, scenario and
timebased assessments, respectively. Since the building is symmetric, a two
dimensional model was used for analysis: the first component of the pair was
assigned the X direction of the building and the second component was
assigned to the Y direction.
Seven demand parameters, three story drifts and four peak floor accelerations
were used to characterize performance. Results are attached in Tables 86,
87, and 88. The notation used for drift is dr_x_y, where dr denotes drift, x
denotes the direction (1 or 2) and y denotes the story, with story 1
immediately above the base of the building. For accelerations, the notation is
a_x_z, where a denotes acceleration, x denotes direction (1 or 2) and z
denotes the floor (g is base and 2 is the first supported level above the base).
Figure 63 illustrates these definitions of floor and story numbers. Story
drifts and floor accelerations are organized by directions.
April 15, 2009
814 8: Example Applications ATC58
Table 86 Demand Parameters for IntensityBased Assessment
Number
stories
=
3
Number
of GM
pairs =
11 Demand parameters in units of inches and g
GM
pair
dr_1_1 dr_1_2 dr_1_3 a_1_g a_1_2 a_1_3 a_1_4 dr_2_1 dr_2_2 dr_2_3 a_2_g a_2_2 a_2_3 a_2_4
1 2.61 3.59 3.91 0.40 0.64 0.56 0.65 1.44 2.44 2.76 0.17 0.34 0.31 0.47
2 4.57 5.66 5.52 0.62 0.67 0.82 0.61 2.64 2.93 4.02 0.50 0.91 0.76 0.69
3 1.85 2.34 2.41 0.36 0.77 0.80 0.66 1.89 2.72 2.92 0.33 0.60 0.54 0.51
4 2.91 3.63 3.39 0.29 0.46 0.46 0.68 2.68 3.67 3.60 0.27 0.27 0.33 0.45
5 1.43 2.66 3.40 0.72 1.14 0.85 0.86 1.72 2.64 2.73 0.44 0.60 0.59 0.64
6 2.54 3.28 3.09 0.32 0.33 0.42 0.40 1.94 2.71 2.50 0.31 0.36 0.33 0.38
7 2.36 3.42 3.41 0.47 0.66 0.75 0.57 1.95 2.75 2.59 0.32 0.33 0.34 0.44
8 2.43 3.55 3.73 0.55 0.58 0.55 0.53 3.05 3.52 3.36 0.54 0.90 0.78 0.61
9 2.09 3.13 3.09 0.48 0.56 0.55 0.59 2.02 2.34 2.51 0.40 0.78 0.73 0.59
10 1.37 2.32 2.53 0.51 1.12 0.87 0.82 1.71 2.94 3.52 0.43 0.55 0.52 0.53
11 1.97 2.21 2.44 0.42 0.77 1.01 0.79 1.53 2.39 2.84 0.25 0.43 0.50 0.53
April 15, 2009
ATC58 8: Example Applications 815
Table 87 Demand Parameters for ScenarioBased Assessment
Number
stories
=
3 Demand parameters in units of inches and g
Stripe dr_1_1 dr_1_2 dr_1_3 a_1_g a_1_2 a_1_3 a_1_4 dr_2_1 dr_2_2 dr_2_3 a_2_g a_2_2 a_2_3 a_2_4
1 0.59 0.89 0.82 0.11 0.18 0.17 0.23 0.55 0.88 0.87 0.05 0.09 0.11 0.19
2 1.01 1.37 1.54 0.25 0.35 0.40 0.40 0.99 1.30 1.76 0.20 0.40 0.40 0.43
3 1.21 1.55 1.54 0.18 0.45 0.51 0.45 1.00 1.58 1.78 0.17 0.36 0.37 0.43
4 1.21 1.91 1.81 0.17 0.27 0.29 0.34 1.19 1.92 1.83 0.16 0.22 0.30 0.37
5 1.03 2.00 2.81 0.48 0.70 0.57 0.73 1.06 1.99 2.32 0.29 0.43 0.40 0.48
6 1.86 2.62 2.58 0.26 0.30 0.39 0.42 1.57 2.49 2.50 0.25 0.29 0.33 0.39
7 2.29 3.31 3.30 0.46 0.65 0.74 0.56 1.88 2.67 2.49 0.31 0.32 0.34 0.43
8 3.14 4.27 4.37 0.61 0.60 0.55 0.55 3.79 4.34 4.22 0.60 0.99 0.81 0.59
9 2.31 3.66 3.88 0.62 0.75 0.60 0.70 2.27 2.71 3.14 0.52 1.07 0.78 0.74
10 1.87 2.93 3.70 0.83 1.80 1.30 0.95 3.13 5.02 6.20 0.71 0.90 0.63 0.72
11 3.79 3.91 4.56 0.99 1.30 1.51 0.91 2.85 4.77 5.81 0.59 0.84 0.77 0.79
April 15, 2009
816 8: Example Applications ATC58
Table 88 Demand Parameters for TimeBased Assessment
Number
stories =
3
Number of
intensities
(8 min) =
8
Number of
GM pairs =
11 Demand parameters in units of inches and g
GM
pair
MAF dr_1_1 dr_1_2 dr_1_3 a_1_g a_1_2 a_1_3 a_1_4 dr_2_1 dr_2_2 dr_2_3 a_2_g a_2_2 a_2_3 a_2_4
Intensity_1
1
0.0452
0.53 0.79 0.73 0.10 0.16 0.15 0.21 0.49 0.79 0.77 0.04 0.08 0.10 0.17
2 0.61 0.81 0.92 0.15 0.22 0.25 0.25 0.60 0.79 1.06 0.12 0.25 0.25 0.26
3 0.62 0.77 0.78 0.09 0.23 0.25 0.23 0.52 0.79 0.88 0.08 0.18 0.18 0.21
4 0.50 0.78 0.73 0.07 0.11 0.12 0.17 0.50 0.79 0.76 0.07 0.07 0.12 0.17
5 0.39 0.75 0.98 0.17 0.22 0.17 0.32 0.43 0.74 0.83 0.11 0.16 0.15 0.21
6 0.55 0.81 0.81 0.08 0.11 0.13 0.19 0.48 0.77 0.78 0.07 0.09 0.11 0.17
7 0.54 0.79 0.90 0.11 0.18 0.23 0.25 0.53 0.78 0.76 0.08 0.09 0.14 0.16
8 0.55 0.74 0.88 0.13 0.19 0.19 0.22 0.54 0.76 0.74 0.13 0.21 0.17 0.16
9 0.58 0.78 0.72 0.12 0.17 0.15 0.16 0.48 0.77 0.97 0.10 0.20 0.23 0.24
10 0.48 0.77 0.76 0.12 0.28 0.23 0.24 0.50 0.76 0.82 0.10 0.13 0.15 0.21
11 0.69 0.79 0.98 0.10 0.26 0.28 0.28 0.48 0.78 0.87 0.06 0.14 0.15 0.21
Intensity_2
1
0.00775
1.17 1.68 1.61 0.21 0.34 0.34 0.44 1.03 1.71 1.72 0.09 0.17 0.22 0.35
2 1.37 1.94 2.10 0.33 0.44 0.50 0.48 1.34 1.75 2.33 0.26 0.49 0.52 0.49
3 1.28 1.62 1.64 0.19 0.48 0.54 0.47 1.04 1.67 1.87 0.18 0.39 0.39 0.44
4 1.10 1.72 1.64 0.15 0.23 0.25 0.36 1.08 1.73 1.65 0.14 0.14 0.24 0.35
5 0.83 1.61 2.17 0.38 0.52 0.41 0.64 0.91 1.61 1.83 0.23 0.35 0.32 0.43
6 1.19 1.79 1.80 0.17 0.25 0.28 0.39 1.03 1.69 1.74 0.16 0.19 0.25 0.37
7 1.18 1.75 1.88 0.25 0.37 0.48 0.48 1.13 1.71 1.67 0.17 0.19 0.29 0.34
8 1.17 1.64 2.00 0.29 0.40 0.40 0.45 1.19 1.63 1.56 0.29 0.47 0.36 0.35
9 1.28 1.69 1.55 0.25 0.35 0.32 0.32 1.03 1.68 2.15 0.21 0.43 0.51 0.47
10 1.00 1.63 1.65 0.27 0.62 0.51 0.51 1.07 1.63 1.79 0.23 0.29 0.31 0.42
11 1.36 1.54 2.02 0.22 0.54 0.61 0.58 1.03 1.69 1.92 0.13 0.27 0.33 0.44
April 15, 2009
ATC58 8: Example Applications 817
Table 88 Demand Parameters for TimeBased Assessment (continued)
Number of
GM pairs =
11 Demand parameters in units of inches and g
GM
pair
MAF dr_1_1 dr_1_2 dr_1_3 a_1_g a_1_2 a_1_3 a_1_4 dr_2_1 dr_2_2 dr_2_3 a_2_g a_2_2 a_2_3 a_2_4
Intensity_3
1
0.00318
1.81 2.41 2.83 0.33 0.52 0.42 0.58 1.37 2.39 2.57 0.14 0.26 0.28 0.43
2 3.03 3.90 3.65 0.50 0.57 0.71 0.53 2.32 2.66 3.44 0.41 0.74 0.69 0.62
3 1.55 1.94 2.24 0.29 0.66 0.70 0.61 1.59 2.55 2.67 0.27 0.49 0.51 0.48
4 1.99 2.79 2.69 0.24 0.39 0.36 0.55 1.90 2.82 2.69 0.22 0.22 0.29 0.41
5 1.22 2.16 3.21 0.59 0.90 0.71 0.80 1.30 2.13 2.60 0.36 0.53 0.50 0.53
6 1.89 2.65 2.61 0.26 0.30 0.40 0.41 1.60 2.51 2.51 0.25 0.30 0.33 0.39
7 1.88 2.73 2.71 0.38 0.57 0.66 0.54 1.53 2.22 2.05 0.26 0.27 0.35 0.40
8 1.67 2.47 2.86 0.45 0.49 0.48 0.51 2.08 2.65 2.46 0.44 0.73 0.60 0.54
9 1.78 2.63 2.55 0.39 0.52 0.51 0.51 1.69 2.21 2.62 0.33 0.65 0.68 0.53
10 1.28 2.19 2.31 0.42 0.95 0.76 0.73 1.36 2.41 2.76 0.36 0.45 0.46 0.50
11 1.63 1.93 2.26 0.34 0.65 0.89 0.73 1.32 2.16 2.45 0.20 0.37 0.46 0.53
Intensity_4
1
0.00163
3.19 4.39 4.61 0.45 0.70 0.58 0.67 1.48 2.49 2.86 0.19 0.39 0.35 0.49
2 5.39 6.64 6.54 0.68 0.74 0.87 0.65 2.81 3.17 4.33 0.55 0.98 0.79 0.72
3 2.12 2.53 2.46 0.40 0.82 0.84 0.69 2.07 2.68 2.96 0.37 0.65 0.62 0.52
4 3.41 4.14 3.78 0.32 0.47 0.51 0.74 3.17 4.23 4.28 0.30 0.30 0.34 0.47
5 1.56 2.96 3.46 0.79 1.27 0.92 0.89 2.01 2.96 3.07 0.48 0.63 0.63 0.69
6 3.02 3.78 3.46 0.36 0.36 0.42 0.41 2.10 2.73 2.47 0.34 0.40 0.34 0.39
7 2.66 3.86 3.82 0.52 0.70 0.79 0.58 2.24 3.12 2.98 0.35 0.36 0.33 0.45
8 3.13 4.26 4.36 0.61 0.60 0.55 0.55 3.78 4.33 4.21 0.60 0.99 0.81 0.59
9 2.19 3.35 3.35 0.53 0.62 0.56 0.63 2.13 2.49 2.65 0.44 0.88 0.74 0.64
10 1.41 2.37 2.76 0.57 1.22 0.91 0.85 1.94 3.26 3.96 0.48 0.61 0.54 0.56
11 2.22 2.37 2.52 0.46 0.84 1.06 0.81 1.64 2.53 3.04 0.27 0.45 0.50 0.56
April 15, 2009
818 8: Example Applications ATC58
Table 88 Demand Parameters for TimeBased Assessment (continued)
Number of
GM pairs =
11 Demand parameters in units of inches and g
GM
pair
MAF dr_1_1 dr_1_2 dr_1_3 a_1_g a_1_2 a_1_3 a_1_4 dr_2_1 dr_2_2 dr_2_3 a_2_g a_2_2 a_2_3 a_2_4
Intensity_5
1
0.000500
4.89 6.52 6.99 0.56 0.82 0.61 0.69 2.02 2.76 3.24 0.24 0.49 0.39 0.54
2 7.18 8.70 8.60 0.86 0.82 0.87 0.78 3.44 3.81 4.98 0.69 1.20 1.04 0.82
3 2.81 3.13 3.14 0.50 0.97 0.89 0.77 2.99 3.16 3.65 0.47 0.80 0.85 0.62
4 4.59 5.44 5.13 0.40 0.58 0.65 0.89 4.59 6.09 6.44 0.38 0.38 0.38 0.60
5 1.91 3.81 4.41 1.00 1.60 1.05 0.97 2.88 3.93 4.25 0.61 0.84 0.74 0.82
6 4.62 5.62 5.33 0.45 0.45 0.52 0.45 2.67 3.50 3.28 0.43 0.50 0.43 0.48
7 3.42 5.00 4.95 0.66 0.76 0.82 0.60 3.30 4.22 4.00 0.44 0.46 0.40 0.49
8 4.63 5.76 5.65 0.76 0.64 0.54 0.60 6.34 7.20 7.59 0.75 1.12 0.78 0.60
9 2.35 3.77 4.12 0.67 0.81 0.67 0.74 2.32 2.88 3.39 0.56 1.14 0.85 0.78
10 1.65 2.73 3.30 0.72 1.56 1.13 0.91 2.64 4.23 5.19 0.61 0.77 0.61 0.65
11 2.71 2.77 2.87 0.58 0.89 1.13 0.86 1.83 2.82 3.51 0.35 0.53 0.61 0.62
Intensity_6
1
0.000497
6.07 7.97 8.83 0.68 0.91 0.75 0.74 2.80 3.60 3.80 0.29 0.51 0.42 0.57
2 8.79 10.40 10.02 1.04 1.01 0.83 0.86 4.10 4.34 5.50 0.83 1.48 1.31 0.94
3 2.92 3.09 3.49 0.60 1.09 0.90 0.82 3.86 4.35 5.16 0.56 0.96 1.03 0.68
4 5.63 6.51 6.61 0.48 0.66 0.70 1.01 6.23 8.34 8.82 0.45 0.46 0.46 0.72
5 2.20 4.69 5.67 1.21 1.82 1.20 1.13 3.85 5.01 5.69 0.74 1.06 0.83 0.93
6 7.06 8.25 7.85 0.55 0.52 0.62 0.48 3.44 4.43 4.31 0.52 0.61 0.51 0.53
7 3.97 5.88 5.88 0.79 0.86 0.80 0.67 4.48 5.52 5.25 0.54 0.55 0.48 0.53
8 5.79 6.94 6.78 0.92 0.83 0.63 0.68 10.32 12.06 11.65 0.91 1.15 0.90 0.71
9 2.41 4.05 4.77 0.81 1.00 0.89 0.84 2.84 3.23 4.10 0.67 1.32 1.05 0.86
10 1.95 2.93 3.78 0.86 1.84 1.33 0.96 3.23 5.21 6.45 0.73 0.93 0.64 0.73
11 3.04 2.83 3.22 0.70 1.00 1.15 0.89 2.09 3.33 4.18 0.42 0.61 0.68 0.66
April 15, 2009
ATC58 8: Example Applications 819
Table 88 Demand Parameters for TimeBased Assessment (continued)
Number of
GM pairs =
11 Demand parameters in units of inches and g
GM
pair
MAF dr_1_1 dr_1_2 dr_1_3 a_1_g a_1_2 a_1_3 a_1_4 dr_2_1 dr_2_2 dr_2_3 a_2_g a_2_2 a_2_3 a_2_4
Intensity_7
1
0.000199
6.60 8.76 10.08 0.79 1.01 0.86 0.82 3.66 4.54 4.81 0.34 0.56 0.50 0.59
2 10.43 12.32 11.84 1.22 1.17 0.94 0.87 4.63 5.06 6.15 0.98 1.69 1.59 1.07
3 3.74 4.42 4.62 0.71 1.14 0.98 0.82 4.22 5.04 6.04 0.66 0.95 1.01 0.79
4 6.58 7.49 8.06 0.57 0.59 0.76 1.12 7.87 10.69 11.23 0.53 0.53 0.54 0.81
5 2.52 5.54 6.85 1.41 1.89 1.36 1.26 4.85 6.21 7.25 0.86 1.20 0.95 1.02
6 9.81 11.54 11.26 0.64 0.58 0.70 0.53 4.49 5.29 5.22 0.61 0.70 0.54 0.55
7 4.38 6.64 6.80 0.93 1.04 0.91 0.71 5.73 7.02 6.94 0.63 0.65 0.56 0.54
8 6.44 7.43 7.75 1.08 1.07 0.80 0.80 16.55 18.70 18.66 1.07 1.16 0.97 0.68
9 2.47 4.27 5.28 0.95 1.11 1.01 0.94 3.41 3.64 4.69 0.79 1.43 1.14 0.90
10 2.27 2.94 4.20 1.01 2.02 1.46 0.98 3.68 6.08 7.61 0.86 1.09 0.69 0.79
11 3.35 3.25 3.53 0.82 1.13 1.32 0.91 2.45 3.96 4.92 0.49 0.70 0.73 0.68
Intensity_8
1
0.000146
7.32 9.61 11.34 0.91 1.15 1.02 0.87 4.29 5.35 5.89 0.39 0.64 0.53 0.63
2 11.72 13.97 13.79 1.39 1.45 1.04 0.98 5.01 5.95 7.42 1.12 1.84 1.87 1.18
3 5.72 6.63 7.01 0.81 1.28 1.03 0.84 4.82 6.45 6.99 0.75 1.02 1.05 0.87
4 7.43 8.48 9.42 0.65 0.75 0.83 1.24 9.53 12.89 13.60 0.61 0.61 0.62 0.90
5 2.97 6.36 7.93 1.62 2.11 1.56 1.36 5.88 7.50 8.89 0.99 1.26 1.07 1.11
6 13.30 15.57 15.38 0.73 0.69 0.79 0.57 6.41 7.40 7.35 0.70 0.79 0.60 0.57
7 5.04 7.30 7.72 1.06 1.21 1.11 0.76 7.03 8.64 8.76 0.72 0.74 0.63 0.56
8 7.91 9.70 10.29 1.24 1.26 0.94 0.91 22.44 25.84 26.59 1.22 1.25 1.02 0.79
9 2.97 4.98 6.20 1.09 1.11 1.07 1.03 3.88 4.10 5.31 0.91 1.48 1.19 0.92
10 2.51 3.47 4.67 1.16 2.12 1.58 0.99 4.06 6.84 8.71 0.98 1.24 0.79 0.85
11 3.65 3.73 4.23 0.94 1.25 1.46 0.91 2.81 4.60 5.62 0.56 0.83 0.76 0.75
April 15, 2009
820 8: Example Applications ATC58
8.2.4 Input of Data into the Performance Assessment
Calculation Tool (PACT)
The input screens for PACT are accessed through an input hub as shown in
Figure 813. To begin click on Edit General Info.
Figure 813 PACT Input Hub
This will bring up the general information entry screen, shown in Figure 8
14. Note that this screen will show userspecific information, based on data
entered when PACT is first installed on the computer. Enter data as follows:
 Enter a project identification in the top data entry box. This can be a
building name, project number or similar identification.
 In the second data entry box, enter a brief description of the building.
This is not actually used by the program, but is provided to allow later
identification of the data file as to the building assessed.
 In the third data entry box enter the client’s name.
April 15, 2009
ATC58 8: Example Applications 821
 In the fourth data entry box, enter the name of the engineer performing
the assessment.
 Leftclick on the “return” button to store the data and return to the input
hub.
Figure 814 PACT General Info Screen
To begin entering quantitative data on the building configuration, leftclick
on the “enter building info” button on the input hub screen. This will bring
up the building information screen as shown in Figure 815.
In the Building Information Screen (Figure 815) enter data as follows:
 In the top data left hand data entry window, enter the number of stories
above grade.
 In the dropdown window below the story data entry box, select the
building occupancy template. The user may specify the use of an
occupancy template that is included in PACT. When this alternative is
selected, PACT “seeds” the performance groups and quantities of
performance groups for the building using typical quantities based on
architectural standards. The user can start from scratch and simply enter
the performance group information directly as explained below.
April 15, 2009
822 8: Example Applications ATC58
 Beneath the occupancy window, enter the typical area of a single floor in
square feet. Note that for buildings with setbacks, it is possible to
modify this area for different floors. PACT simplifies the data input
process using the Most Typical quantities designation. For example, in
Figure 41 the user has entered 11.5 as the Most Typical Story Height.
The datasheet in the lower panel is interactive and initially includes data
for each story based on the Most Typical quantities. It is possible to
override the Most Typical quantities. In this example, the height for the
first story has been changed from the Most Typical value of 11.5 feet to
14 feet.
Figure 815 PACT Building Information Screen
 In the data entry box located to the right of the floor area data box, enter
the typical story height. Note that this can also be modified later for
atypical stories.
 In the two boxes below the entry for Typical Floor Area, enter the total
length of perimeter wall in the eastwest (Direction 1) and northsouth
(Direction 2) in feet, for the most typical floor. This can be modified
later for individual floors.
April 15, 2009
ATC58 8: Example Applications 823
Figure 816 PACT Performance Group Quantity Screen(s)
 In the data entry boxes to the left, enter the structural system, in
accordance with Section 2.3 in each of the eastwest (direction 1) and
northsouth (direction 2) directions of response.
 Enter the total estimated replacement cost for the structure in current
U.S. dollars.
 Leftclick the “Return” button to return to the input hub.
April 15, 2009
824 8: Example Applications ATC58
The next step in the process is to formulate performance groups for the
building. Click “Edit Performance Groups” on the input hub. This brings up
the “Performance Group Quantity Screen(s)” as shown in Figure 816
In this example, PACT automatically forms default performance groups and
assigns default fragility specifications based on the occupancy, structural
type, and basic building information provided. The user specifies this option
by clicking on “fill data based on chosen template for steel moment frame
office building. Appendices D and E contain a more complete listing of the
default performance groups and fragilities contained within the PACT
database.
Performance groups are summarized by those subject to damage cause by
displacement in each of the two orthogonal directions (Direction 1 and
Direction 2) and those sensitive to acceleration in any direction (Direction
independent). The radio buttons switch among these three options. It is
possible to edit any of the quantities in the performance groups and change or
add fragility specifications to correct for conditions in the actual building.
Leftclick the Return button to return to the input hub.
The next step in the process is to enter the results of the response analyses.
Click on “enter analysis results for all scenario or intensities. This will bring
up the “View Analysis Cases” screen shown in Figure 817.
For each intensity or scenario, the results of the response analysis, in terms of
drifts and accelerations from the tables in the previous section, are entered
into the program. There are results for each orthogonal direction. The
nondirectional data can be “autofilled.” This results in the use of the
maximum response in either direction, which is useful for acceleration
sensitive performance groups. When all the data is entered, click on
“Return” to go back to the input hub.
April 15, 2009
ATC58 8: Example Applications 825
Figure 817 PACT View Analysis Cases Screen
8.2.5 Loss Computations Using PACT
Once all of the data is input into PACT, the user clicks on “perform
assessment for all scenarios or intensities” in the input hub. For scenario or
intensity based assessments the losses are viewed using the screen shown in
Figure 818.
The analysis to be viewed is selected from the upper dropdown menu on the
left. The loss curve shows the total losses for the building subjected to the
specified scenario or intensity. The losses are plotted on the horizontal axis.
The probability of the losses being equal to or less than any specific value is
read from the vertical axis – P (total repair cost<=$C). In the lower left of
the display screen are controls that facilitate the reading of losses from the
curve. If a loss value is typed into the dialog box, the corresponding
probability of nonexceedance will appear in the adjacent box and the cursor
will move to the designated point on the plot.
April 15, 2009
826 8: Example Applications ATC58
Figure 818 PACT Intensity and Scenarios Based Loss Screen
Alternatively, the probability of nonexceedance can be entered and the loss
will be automatically provided in the adjacent box. Also, it is possible to
simply move the cursor to the desired point along the graph to obtain
readouts in the output boxes. In Figure 818 the total losses for the intensity
or scenario have a 50% probability of being less than or equal to $5,300,000
(median loss). The bar chart above the loss curve shows the contribution of
each performance group to the loss. Below “analysis” menu there are three
boxes that allow the user to select which losses are to be displayed in the bar
chart. The first box lists the various types of performance groups, for which
losses can be individually viewed; the second box allows specification of the
direction of shaking; and the lower box designates the story within the levels
of the building for which the losses are to be displayed. In the figure, as an
example, direct economic losses are plotted for all performance groups from
all directions, located on all floors of the building.
April 15, 2009
ATC58 8: Example Applications 827
To view the results of a timebased assessment, click on “Perform Time
Based Assessment” on the input hub. This will bring up the screen shown in
Figure 819.
The vertical axis for a time based assessment is the annual probability that
the loss on the horizontal axis will be exceeded. The integration of the time
based loss curve results in the total annualized loss (i.e., the total loss that
would be expected spread out over time). This information is shown in the
upper right corner of the screen.
Figure 819 PACT TimeBased Assessment Loss Screen
April 15, 2009
ATC58 A: Probability, Statistics & Distributions A1
Appendix A
Probability, Statistics &
Distributions
A.1 Introduction
This appendix provides a brief tutorial on probability and statistics including
methods of expressing probabilities in the form of distributions. It is
intended to provide readers who are unfamiliar with these topics basic
information essential to an understanding of the process used to account for
the uncertainties inherent in performance assessment. Interested readers may
wish to obtain additional information on this subject by referring to textbooks
on probability and statistics, and on structural reliability theory. In
particular, Benjamin and Cornell contains a wealth of information on this
topic.
A.2 Statistical Distributions
A statistical distribution is a mathematical representation of the probability of
encountering a specific outcome, or an outcome that is either greater than or
less than the specific outcome, given a set of possible outcomes. The set
representing all possible outcomes is termed a population. There are
generally two broad types of populations considered in statistical studies.
The first of these is a finite population of outcomes, where each possible
outcome has a discrete value representing one of the finite number of
possible outcomes. The second of these is an infinite population of possible
outcomes. Each is discussed separately, below.
A.2.1 Finite Populations and Discrete Outcomes
Consider the classic case of a coin thrown in the air to determine an outcome.
One of two possible outcomes will occur each time the coin is tossed. One
potential outcome is that the coin will land “headsup” and the other that it
will land “headsdown.” Which way the coin will land on a given toss is a
function of a number of factors including which way we hold the coin before
we toss it; the technique we use to toss the coin; how hard or high we toss it;
and how, it lands. We could never hope to precisely simulate each of these
factors, and therefore, the occurrence of a “headsup” or “headsdown”
outcome in a given coin toss appears to be a random phenomena, that is not
predictable. There is an equal chance that the coin will land “heads up” or
April 15, 2009
A2 A: Probability, Statistics & Distributions ATC58
“heads down,” and we can not know before tossing the coin, which way it
will land.
If we toss a coin in the air one time there are two possible outcomes. These
two outcomes – coin lands “headsup” and con lands “headsdown”
completely define all possibilities for one coin toss and therefore, the
probability that the coin will land either heads up or heads down is 100%. In
essence, this illustrates the total probability theorem, that is, that the sum of
the probabilities of all possible outcomes will be 1.0.
The probability that the coin will land “headsup” is 0.5, or 50%. The
probability that it will land “headsdown” is the inverse of this, calculated as
one minus the probability of “headsup,” or (10.5) =0.5, also 50%.
If we toss the coin in the air many times, we would expect that we would get
a “headsup” outcome in half of these tosses and “headsdown” outcome in
the other half. This does not mean that every second toss of the coin will
have a “headsup” outcome. We might obtain several successive “headsup”
outcomes or several successive “headsdown” outcomes, however, over a
great many tosses, we should have approximately the same number of each
possible outcome.
A.2.2 Combined Probabilities
A combined probability is the probability of experiencing a specific
combination of two or more independent outcomes. We can calculate the
combined probability of two independent events as the product of the
probability of outcome 1 and the probability of outcome 2. That is:
) ( ) ( ) ( B P A P B A P  = + (A1)
To illustrate this, if we toss the coin into the air two times, there is a 50%
chance, each time, that the coin will land “headsup.” The chance that the
coin will land “headsup” both times is calculated, using equation A1, as the
probability of landing “headsup” the first time (0.5) multiplied by the
probability that it will land “headsup” the seconds time (0.5), or (0.5 x 0.5)
=0.25, or 25%.
This probability means that if we toss a coin into the air twice in succession,
a large number of times, and record the number of “headsup” outcomes, in
each pair of tosses, approximately 25% of the total number of pairs of tosses
will have two successive headsup outcomes. The 25% probability does not
mean that every fourth time we make a pair of coin tosses we will get two
successive headsup outcomes. There is some possibility that we will get
two successive headsup outcomes several times in a row and there is also a
April 15, 2009
ATC58 A: Probability, Statistics & Distributions A3
possibility that we will have to make more than four pairs of coin tosses to
obtain an outcome of two headsup in any of the pairs of tosses. However,
over a very large number of pairs of tosses, one fourth of the pairs should be
successive headsup outcomes.
If we toss the coin into the air three times, there is again a 50% chance on
each toss that the coin will land “headsup.” The probability that the coin
will land “headsup” all three times is given by the probability that it will
land “headsup” the first two times (25%) multiplied by the probability that it
will land “headsup” the third time (50%), or (0.25 x 0.5) =0.125, or 12.5%.
If we repeat this exercise with 4 coin tosses, the probability that all four will
land “headsup” will be the probability that the first three tosses will land
“headsup” (12.5%) times the probability that the fourth toss will land
“headsup” or (0.125 x 0.5 =0.0625) or 6.25%. That is, there is
approximately a 6% chance that we will have two successive pairs of coin
tosses both having two headsup outcomes.
A.2.3 Mass Distributions
A probability mass distribution is a plot of the probability of occurrence of
each of the possible outcomes in a finite population of discrete outcomes,
such as a coin landing either “headsup” or “headsdown” a specified number
of times in Nthrows.
Consider the case of four coin tosses. As shown in the previous section,
there is a 6.25% chance that all four coin tosses will land “headsup.” There
is also a 6.25% chance that all four coin tosses will land “headdown” or that
none of the coin tosses will land “headsup.” The chance that three of the
four coin tosses will land “headsup” is equal to the combined chance of any
of the following outcomes: “T, H, H, H”; “H, T, H, H”; “H, H, T, H” or “H,
H, H, T,” where “H” indicates “headsup” and “T” “headsdown .” The
chance of each of these outcomes is the same as having all four tosses
landing “headsup” or 6.25%. Therefore, the chance of exactly three out of
four tosses landing “headsup” is equal to 6.25% for the “T, H, H, H”
combination, plus 6.25% for the “H, T, H, H” combination, plus 6.25% for
the “H, H, T, H” combination, plus 6.25% for the “H, H, H, T” combination,
or a total of 25% (6.25% +6.25% +6.25% +6.25%). Similarly, the
probability that exactly three of the four tosses will be “headsdown,” or that
only one of the tosses will land “headsup” is 25%. We can use similar
approaches, to show that the probability that exactly half the tosses will land
“headsup” is 37.5%.
April 15, 2009
A4 A: Probability, Statistics & Distributions ATC58
Figure A1 Probability mass function indicating the probability of “n”
numbers of “headsup” outcomes in four successive coin tosses
Figure A1 is a plot that shows the probability of obtaining zero, one, two,
three, or four “headsup” outcomes from four coin tosses. Plots of this type
are termed probability mass distributions. By entering the plot along the
horizontal axis, at a particular outcome, for example – “1 headsup” out of
four throws, we can read vertically to see the probability of this outcome,
which is 25%. The probability of having not more than one “headsup” toss
in four tosses, is calculated as the sum of the probability of no “headsups”,
which is 6.25% plus the probability of 1 “headsup” which is 25%, for a
cumulative probability of 31.25%. Another way to say this is that the
probability of nonexceedance for “1 headsup in 4 coin tosses” is 31.25%.
The inverse of this is that the cumulative probability of more than “1 heads
up” is 1–0.3125 or 68.75%. This could also be called the probability of
exceedance of “1 headsup toss in 4 tosses.”
A.2.4 Continuous Distributions
The distribution of possible “headsup” throws in a finite number of coin
tosses is an example of a discrete distribution. That is, there are only a finite
number of possible outcomes, and these have discrete values, in the example
above, 0, 1, 2, 3 or 4 “headsup” outcomes in 4 coin tosses.
For many situations, the possible outcomes are not a finite number of discrete
possibilities but rather a continuous range of possibilities. An example of
such a continuous distribution is that of possible compressive strengths
obtained from concrete cylinder compression tests, where all cylinders are
from concrete mixed using the same mix design. Such a distribution will
have the form shown in Figure A2, where the dispersion in strength is due to
minor variability in the amounts of cement, amount of water, strength of
April 15, 2009
ATC58 A: Probability, Statistics & Distributions A5
aggregates, cylindercasting technique, curing technique etc. There are an
infinite number of possible outcomes for the strength of any particular
cylinder test. The form of the distribution shown in Figure A2 is termed a
probability density function.
The area under the curve of a probability density function, between any two
points along the horizontal axis gives the probability that the value of any
member of the population will be within the range defined by the two values.
Figure A3 illustrates this. In the figure, the probability that a single cylinder
test, conforming to the population represented by Figure A2 will have
strength that is greater than 4,000 psi but less than or equal to 5,000 psi is
given by the area under the curve between the two strength values, which in
this case has a calculated probability of 54%. That is, there is a 54% chance
that any cylinder test in this population will have strength between 4,000 psi
and 5,000 psi.
0
0.0001
0.0002
0.0003
0.0004
0.0005
0.0006
0.0007
2000 3000 4000 5000 6000 7000
Concrete 28day strength
P
r
o
b
a
b
i
l
i
t
y
D
e
n
s
i
t
y
Figure A2 Distribution of possible concrete cylinder strengths for a
hypothetical mix design
April 15, 2009
A6 A: Probability, Statistics & Distributions ATC58
Figure A3 Calculation of probability that a member of the population will
have a value within a defined range.
A.3 Common Forms of Distributions
A.3.1 Normal Distributions
The probability density function illustrated in Figure A2 has a special set of
properties known as a normal distribution. First, the “median” outcome, that
is, the outcome that is exceeded 50% of the time is also equal to the average
or “mean” outcome, which is the total value of all possible outcomes, divided
by the number of possible outcomes. In the example above, the median
outcome is that the concrete has a compressive strength of 4,500 psi. The
average strength of all cylinders tested is also 4,500 psi. Normal
distributions are also symmetric. That is, there is an equal probability of
having a value at a defined measure above the average, say 1.5 times the
average, as there is of having a value the same defined distance below the
average, say .5 times the average.
Two parameters are used to uniquely specify the characteristics of a normal
distribution, namely, the mean value x and standard deviation, o .
Equations A2 and A3, define these values for a random variable x (e.g.
compressive strength of concrete as measured by cylinder testing) and a
sample of size N:
N
x
x
N
i
i ¿
=
=
1
(A2)
( )
2
1
1
1
N
i
i
x x
N
o
=
= ÷
÷
¿
(A3)
April 15, 2009
ATC58 A: Probability, Statistics & Distributions A7
where N is the size of the population. If a parameter is normally distributed,
that is, its population of possible outcomes is represented by a normal
distribution, having mean value x and standard deviation o , it is possible to
determine the value x of an outcome that has a specific probability of
exceedance, based on the number of standard deviations that the value x lies
away from the mean, x . For example, there is a 97.7% chance that the value
of any single outcome x will be greater than x 2o , an 84.1% chance that
the value of any single outcome x will be greater than x o , a 50% chance it
will be greater than x , a 15.8% chance it will exceed x +o and a 2.3%
chance that it will be greater than x +2o . Standard tables that are available
in most texts on probability theory indicate the probability that the value of
any outcome in a normally distributed population of potential outcomes will
exceed a value that is a defined number of standard deviations from the
mean. The number of standard deviations above or below the mean that will
have a specified probability of exceedance is sometimes termed the Gaussian
variate and the probability tables that give these values, Gaussian tables.
Structural engineers use normal distribution to represent the distribution of
values for many random quantities (e.g., concrete compressive strength). In
addition to the mean, median and standard deviation, another oftenused
parameter to characterize the properties of a normal distribution is the
coefficient of variation (COV). The coefficient of variation is calculated as
the standard deviation, o , divided by the mean value x . It is useful
because it represents a normalized measure of the scatter inherent in a
normally distributed population. Figure A4 plots probability density
functions for several normal distributions, all having mean values of 1.0 and
coefficients of variation of 0.1, 0.25, and 0.5, respectively.
0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 0.5 1 1.5 2
Val ue x
P
r
o
b
a
b
i
l
i
t
y
D
e
n
s
i
t
y
Coef. Of Variation =0.1
Coef. Of Variation =0.25
Coef. Of Variation =0.5
Figure A4 Probability density function plots of normal distributions with
mean values of 1.0 and coefficients of variation of 0.1, 0.25 and
0.5.
April 15, 2009
A8 A: Probability, Statistics & Distributions ATC58
A.3.2 Cumulative Probability Functions
An alternative means of plotting probability distributions is in the form
known as a cumulative probability function. A cumulative probability
function plot shows the probability that an outcome in the population of
possible outcomes will have a value that is less than or equal to the specified
value x. This probability is sometimes termed the probability of non
exceedance. Cumulative density plots are obtained by integrating over the
probability density function to determine the area under the curve between an
x value of zero and any other value of x. Figure A5 presents cumulative
probability plots for the same normally distributed populations previously
shown in Figure A4. From either series of plots, it is possible to observe
that the larger the coefficient of variation becomes, the greater the amount of
scatter in possible outcome values.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.5 1 1.5 2 2.5
Val ue x
P
r
o
b
a
b
i
l
i
t
y
o
f
N
o
n

E
x
c
e
e
d
a
n
c
e
Coef. Of Variation =0.1
Coef. Of Variation =0.25
Coef. Of Variation =0.5
Figure A5 Cumulative probability plots of normal distributions with
coefficients of variation of 0.1, 0.25, and 0.5.
A.3.3 Lognormal Distributions
Although normal distributions represent some random variables well, they do
not represent all variables well. Some variables have skewed distributions.
In skewed distributions, the mean value, x , will be either greater than or less
than the median value. In structural reliability applications, such as
performance assessment, it is common to use a specific type of skewed
distribution known as a lognormal distribution. This is because the skew
inherent in lognormal distributions can reasonably represent the distributions
observed in many structural engineering phenomena, such as the distribution
of strength in laboratory specimens.
The lognormal distribution has the property that the natural logarithm of the
values of the population In( ) x are normally distributed. The mean and the
April 15, 2009
ATC58 A: Probability, Statistics & Distributions A9
median of the natural logarithms of the population have the same value,
which is equal to In( ) u , where u is the median value. In these guidelines,
the standard deviation of the natural logarithms of the values
In( ) x
o is called
the dispersion and is denoted by the symbol . For relatively small values of
dispersion, the value of  is approximately equal to the coefficient of
variation for x. Together, the values of u and  completely define the
lognormal distribution for a population.
Figure A6 plots probability density functions for three lognormally
distributed populations having median values of 1.0 and dispersions of 0.1,
0.25 and 0.5 respectively. Figure A7 plots this same data in the form of
cumulative probability functions.
0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 1 2 3
Val ue x
P
r
o
b
a
b
i
l
i
t
y
D
e
n
s
i
t
y
Dispersion =0.1
Dispersion =0.25
Dispersion =0.5
Figure A6 Probability density function plots of lognormal distributions with
median values of 1.0 and dispersions of 0.1, 0.25 and 0.5.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 1 2 3 4 5
Val ue x
P
r
o
b
a
b
i
l
i
t
y
o
f
N
o
n
E
x
c
e
e
d
a
n
c
e
Dispersion =0.1
Dispersion =0.25
Dispersion =0.5
Figure A7 Cumulative probability plots of lognormal distributions with
median values of 1.0 and dispersions of 0.1, 0.25, and 0.5.
April 15, 2009
A10 A: Probability, Statistics & Distributions ATC58
As can be seen from the plots, for small values of the dispersion, the
distribution approaches the shape of the normal distribution, but as the
dispersion increases in value, the distribution becomes more and more
skewed. This skew is such that there is nil probability of incurring a negative
value in the distribution (because the logarithm of a negative number is
positive) and extreme values above the median are substantially more
probable than extreme values below the median. Consider the distribution of
possible actual yield strengths of various steel parts, all conforming to a
specific ASTM specification and grade. Since this variable (yield strength)
cannot take on values of less than zero, the lognormal distribution could be
used to describe the variation in possible steel strength using a median value
and dispersion. In this example, the median value will be substantially
higher than the minimum specified value, which, is intended to have a very
low probability of nonexceedance by any steel conforming to the
specification.
The procedures presented in these guidelines use lognormal distributions to
represent the distributions of intensity for a scenario earthquake; the values
of response parameters given an intensity; the probability of incurring a
damage state as a function of response parameter, herein termed a fragility;
and the probability of incurring a specific level of direct economic loss,
downtime and casualties, given a damage state. These lognormal
distributions are derived in a variety of ways. Specifically,
 Intensity distributions are obtained based on statistical analysis of actual
ground motion data recorded from past earthquakes, and represented in
the form of equations known as attenuation relationships
 Response parameter distributions are obtained by performing suites of
structural analyses using multiple ground motion recordings and varying
the strength, stiffness, damping, and ductility capacity of the structural
elements
 Fragility distributions are obtained primarily through laboratory testing
or observation of actual damage in past earthquakes
 Loss distributions are obtained through performing multiple calculations
of loss, varying the factors that affect the loss, such as contractor
efficiency, and number of persons likely to be occupying a building at
the time of the earthquake
Regardless of how these distributions are derived, they are represented by
two variables, the median value, u and dispersion .
April 15, 2009
ATC58 A: Probability, Statistics & Distributions A11
The mean
Y
m and standard deviation
Y
o of a lognormal distribution Y, can
be calculated from
Y
u and
Y
 as follows:
2
exp
2
Y
Y Y
m

u
 
=

\ .
(A4)
( )
2 2 2
exp 1
Y Y Y
m o 
(
= ÷
¸ ¸
(A5)
The coefficient of variation,
Y
v , is given by
2
exp( ) 1
Y
Y Y
Y
m
o
v  = = ÷ (A6)
Many common spreadsheet applications also have embedded within them,
functions that will automatically solve lognormal distribution problems. For
example, in Microsoft Excel, the LOGNORMDIST function will determine
the cumulative probability of nonexceedance of any value in a population Y,
based on input of the value, y, the natural logarithm of the median ln
Y
u and
the dispersion,
Y
 . The Excel input is of the form =LOGNORMDIST (y,
ln
Y
u ,
Y
 ). Similarly, the LOGINV function can be used to determine the
value of y, at a given cumulative probability of nonexceedance, based on the
desired probability, p, which must be less than 1.0, the natural logarithm of
the median ln
Y
u and the dispersion,
Y
 . The Excel input is of the form
=LOGINV(p, ln
Y
u ,
Y
 ).
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B1
and Population Models
Appendix B
Fragility Group, Normative
Quantity and Population Models
B.1 Fragility Group Classifications
Table B1 provides a listing of all fragility groups, structural and non
structural contained within the PACT databases, together with the Uniformat
Coding applied to these fragility groups. Default fragility specifications are
provide for each of these fragility groups within the PACT Databases.
B.2 Occupancy Default Assignment Tables
Table B2 and B3 itemize default descriptions and fragility IDs
corresponding to the default inventory quantities assigned to lowrise
commercial office buildings and lowrise, multifamily residential buildings.
The logic used to determine default quantities is provided in Section B.2, and
default fragility values for nonstructural components are provided in
Appendix D.
Reader note: At this stage only 2 tables are provided. In subsequent versions
of these guidelines, it is planned that the tables we will be provided for all 8
occupancies identified in Sec 2.2.2 of Part B and for each of the factors that
would affect the seismic fragilities. It is also expected that more items will be
considered in the list of default inventory items. When completed, it is
expected there will be several hundred of these tables that will be included in
updated versions of PACT.
B.3 Occupancy Default Normative Quantity Logic
B.3.1 Commercial Office
Table B4 below indicates the default normative quantities for nonstructural
components related to commercial office occupancies. These quantities
apply only to buildings that are 5 stories or less in height. The fragility data
corresponding to the fragility specification is provided in Appendix D.
B.3.2 Residential
Table B5 indicates the default normative quantities for nonstructural
components related to residential occupancies. These quantities apply only
to lowrise buildings.
April 15, 2009
B2 B Fragility Group, Normative Quantity ATC58
and Population Models
B.4 Population Models
Tables B6 presents the default population models for commercial office and
multifamily occupancies. Table B7 presents the default monthly variation
of occupancy, relative to peak, for these same occupancies.
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B3
and Population Models
Table B1 Default Fragility Groups
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
A Substructure
B Shell B10 Super
Structure
B1010 Floor Construction B1016 Floor Raceway
Systems
B1016.000 Floor Raceway Systems 
For Fragility see D5093.0
B1017 Horizontal
framing
(diaphragms)
B1017.CSD
2a
Concrete Slab on Ribbed
Steel Deck; Welded wire
fabric reinforcement;
decking attached to
framing by puddle welds
B1017.CSD
3a
Concrete Slab on Ribbed
Steel Deck; Connected to
supporting framing through
welded shear studs,
reinforcing bars at critical
locations for shear and in
plane flexure in slab
B1017.SD1a Ribbed Steel Deck;
Connected by intermittent
plates and screws to
seismic framing system.
April 15, 2009
B4 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1017.SD2a Ribbed Steel Deck;
Connected by studs or
puddle welds to provide
diaphragm continuity and
connection to the seismic
framing system.
B1017.CS2a CastInPlace Concrete
Slab; Connected to
supporting framing by
nominal reinforcement.
B1017.CS3a CastInPlace Concrete
Slab; Connected to
supporting framing by
reinforcing bars or other
anchorage, slab reinforcing
bars at critical locations for
shear and inplane flexure
in slab
B1017.PC1a Precast Concrete Slab or
Slab/Joists; Connections
between slab/joists and at
boundary connections
designed for strength only.
B1017.PC2a Precast Concrete Slab or
Slab/Joists; Detailed for
ductile connections
between slab/joists and at
boundary connections
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B5
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1017.PC2b Precast Concrete Slab or
Slab/Joists; Same as PC3a,
but with cast in place
topping slab with
provisions made for
bond/shear transfer
between topping and
precast panels with welded
wire fabric.
B1017.PC3a Precast Concrete Slab or
Slab/Joists; Same as PC3a,
but with cast in place
topping slab with
provisions made for
bond/shear transfer
between topping and
precast units using rebar.
B1017.SPA1a Wood Sheathing with
Structural Panels; Structural
sheathing panels with
minimum nailing to
supporting framing.
B1017.SPA2a Wood Sheathing with
Structural Panels; Structural
sheathing panels with close
nailing to supporting
framing to provide
continuity and shear
transfer.
April 15, 2009
B6 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1017.SPA2b Wood Sheathing with
Structural Panels; Structural
sheathing panels with close
nailing to supporting
framing to provide
continuity and shear
transfer with concrete
topping slab
B1017.SPL1a Wood Sheathing with
Structural Planks; Diagonal
or straight plank sheathing
with nailing designed for
shear transfer.
B1017.SPL2a Wood Sheathing with
Structural Planks; Diagonal
or straight plank sheathing
with nailing designed for
shear transfer and with a
concrete topping slab
B1017.SR1a Steel Bracing; Diagonal rod
or other steel bracing
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B7
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1017.SR2a Steel Bracing; Diagonal rod
or other steel member
bracing designed using
capacity design principles
to prevent connection
failure prior to significant
yielding either in bars or
the seismic lateral system
B1018 Gravity
framing
B1018.MD2a Concrete over metal deck
or metal deck supported
by rolled steel beams or
openweb joists and HSS
section columns
B1018.MD3a Concrete over metal deck
or metal deck supported
by rolled steel beams or
open web joists and rolled
steel columns
B1018.CIP1a Lightly reinforced and post
tensioned flat twoway
slabs not designed to
transfer shear/flexural loads
into columns
B1018.CIP2
2a
Lightly reinforced and post
tensioned flat twoway
slabs not specifically
designed to transfer
shear/flexural loads into
columns, but include
structural integrity
reinforcement
April 15, 2009
B8 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1018.CIP2
3a
Lightly reinforced and post
tensioned flat twoway
slabs specifically designed
to transfer shear/flexural
loads into columns,
including structural
integrity reinforcement
B1018.CIP1
1a
Oneway slab with
reinforcement in one
direction supported by
concrete joists, beams, and
columns
B1018.CIP1
2a
Oneway slab with
reinforcement in both
directions supported by
concrete joists, beams, and
columns
B1018.PC1a Precast concrete planks
without topping slab
supported by concrete
beams and columns
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B9
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1018.PC2a Precast concrete planks
with topping slab
integrating the planks and
the supporting beams,
supported by concrete
columns
B1018.W2a Wood structural flooring
supported by wood joists
beams, columns, and
bearing walls
B1018.W3a Wood structural flooring
supported by wood joists,
beams, and columns
B1030 Structural Steel B1031 Steel moment
resisting
frames (S1)
B1031.S1MF
1a
Basic strength design w/o
special detailing;2005 AISC
Specification (w/o seismic
detailing); pre1985 UBC
(strength design)
B1031.S1MF
2a
S1MF1a with masonry or
concrete infill walls
B1031.S1MF
2b
Strength design with
limited rotation capacity in
beamcolumn
connections;2005 AISC 
OMF
B1031.S1MF
2c
Seismically designed for
high ductility but with pre
Northridge connection
details;19851994 UBC
(preNorthridge)
April 15, 2009
B10 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1031.S1MF
3
Strength design with pre
qualified capacity design
provisions for beam
column connections and
member slenderness;2005
AISC  IMF
B1031.S1MF
4a
Seismically designed and
detailed with prequalified
capacity design
provisions;Wshape
columns shallower than
W18 or box columns;2005
AISC  SMF
B1031.S1MF
4b
S1MF4a with Wshape
columns deeper than W18
B1032 Steel braced
frames (S2)
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B11
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1032.S2BF
1
Concentric or eccentrically
braced system; basic
strength design without
special detailing;2005 AISC
Specification (w/o seismic
detailing)
B1032.S2
CBF2a
Concentrically braced
system; strength design
with capacity design
procedures for
connections, compact
bracing member
compactness;2005 AISC 
OCBF
B1032.S2
CBF2b
S2CBF2a with braces
other than HSS members
B1032.S2
CBF2c
S2CBF2a with Kbrace
configuration
B1032.S2
CBF3a
Concentrically braced
system; seismically
designed and detailed with
slenderness/compactness
limits on braces and
capacity design
requirements for
connections, beams, and
columns;2005 AISC  S
CBF
April 15, 2009
B12 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1032.S2
CBF3b
S2CBF3a with braces
other than HSS members
B1032.S2
CBF3c
S2CBF3a with chevron
configuration braces
B1032.S2
CBF3d
S2CBF3a with X
configuration braces
B1032.S2
CBF3e
S2CBF3a with beam
bracecolumn connections
designed as moment
resisting
B1032.S2
CBF4
Concentrically braced
system with buckling
restrained braces and
capacity design of
connections, beams, and
columns;2005 AISC  BRBF
B1033 Steel
eccentrically
braced frames
(S3)
B1033.S3
EBF3a
Eccentrically braced system
designed and detailed with
capacity design provisions;
flexural links remote from
column;2005 AISC  EBF
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B13
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1033.S3
EBF3b
Eccentrically braced system
designed and detailed with
capacity design provisions;
flexural links adjacent to
column;2005 AISC  EBF
B1033.S3
EBF4a
Eccentrically braced system
designed and detailed with
capacity design provisions.
Shear links adjacent to
column;2005 AISC  EBF
B1033.S3
EBF4b
S3EBF1a but with shear
links away from
column;2005 AISC  EBF
B1033.S3
EBF4c
S3EBF1a but with shear
links away from column
with beambracecolumn
connections designed as
moment resisting;2005
AISC  EBF
B1034 Steel light
frame (S4)
B1034.S4
LMF1
Steel moment frame
structure with light gravity
load (i.e., a metal building).
Basic strength design w/o
special detailing;2005 AISC
Specification (w/o seismic
detailing)
April 15, 2009
B14 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1034.S4
LMF2
Steel framed structure with
light gravity load (i.e., a
metal building). Basic
strength design with
capacity design of
connections;2005 AISC 
OMF for systems with low
gravity load
B1035 Steel infilled
frames (S5)
B1035.S5
RCIF2
Concrete infill; Basic
strength design w/o special
detailing;2005 AISC
Specification and ACI318
Building Code (w/o seismic
detailing)
B1035.S5
URMIF1
Unreinforced masonry
infill; Basic strength design
w/o special detailing; 2005
AISC Specification and
Masonry Code? (w/o
seismic detailing)
B1039 Miscellaneous
steel
B1039.001 Gravity shear connections
B1039.002 Column base plates
B1039.003 Welded column splices
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B15
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1040 Reinforced Concrete B1041 Concrete
Moment
Frame (C1)
B1041.C1
MF1
Basic strength design w/o
special detailing;2005 ACI
318 (w/o seismic detailing)
B1041.C1
MF2
Basic strength design with
continuous rebar at beam
column connections;2005
ACI318, OMF
B1041.C1
MF3
Strength design with
continuous rebar at beam
column connections and
column confinement
rebar;2005 ACI318, IMF
B1041.C1
MF4
Seismically designed and
detailed with capacity
design provisions, including
precast systems that utilize
prequalified
connections;2005 ACI
318, SMF
B1042 Concrete
Shear Wall
(C2)
B1041.C2W
1a
Basic strength design with
deficient (nonconforming)
reinforcement details;Pre
1976 UBC (strength
design)
April 15, 2009
B16 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1041.C2W
2a
Low gravity load, basic
strength design w/o seismic
detailing;2005 ACI318
(nonseismic)
B1041.C2W
2b
High gravity load, basic
strength design w/o seismic
detailing;
B1042.C2W
2c
Precast shear wall
(including tiltup), basic
strength design w/o seismic
detailing;
B1042.C2W
3a
Low gravity load, squat
shear wall, strength
design;2005 ACI318,
including Chapter 21
provisions
B1042.C2W
3b
Low gravity load, slender
shear wall, strength design
with seismic detailing for
shear and boundary zone
confinement;
B1042.C2W
3c
Same as C2W3b but with
coupling beams;
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B17
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1042.C2W
3d
High gravity load, squat
shear wall, strength design;
B1042.C2W
3e
High gravity load, slender
shear wall, strength design
with seismic detailing for
shear and boundary zone
confinement;
B1042.C2W
3f
Same as C2W3e but with
coupling beams;
B1042.C2W
3g
Precast shear wall, strength
design with seismic
detailing for wall
anchorage, shear and
boundary zone
confinement;
B1043 Concrete
Frame with
Unreinforced
Masonry Infill
Walls (C3)
B1043.C31a Basic strength design w/o
special detailing;2005 ACI
318 and Masonry Code
(w/o seismic detailing)
April 15, 2009
B18 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1049 Miscellaneous
reinforced
concrete
B1049.001 Reinforced concrete slab
column connections
without shear
reinforcement
B1049.002 Reinforced concrete slab
column connections with
shear reinforcement
B1049.003 Posttensioned slab
column connections
without shear
reinforcement
B1049.004 Posttensioned slab
column connections with
shear reinforcement
B1050 Reinforced masonry B1051 Reinforced
Masonry
Bearing Walls
(RM)
B1051.RM1a Basic strength design w/o
special detailing;
B1051.RM2a Strength design with
seismic detailing;
B1060 Unreinforced masonry B1061 Unreinforced
Masonry
Bearing Walls
(URM)
B1061.URM
1a
Basic strength design w/o
special detailing; existing
URM construction
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B19
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1061.URM
1b
Same as URM1a, but with
tiebacks to diaphragm to
help avoid out of plane
failures; retrofitted URM
construction in high
seismic regions, or new
URM in low seismic
regions
B1070 Timber B1071 Light Wood
Frame Shear
Wall (W1)
B1071.WSW
1a
Structural panel exterior
sheathing (plywood or
OSB) shear walls with
interior gypsum wallboard,
without holddowns, basic
design strength
B1071.WSW
1b
Structural panel exterior
sheathing (plywood or
OSB) with stucco exterior
and gypsum Wallboard
interior, basic strength
design
B1071.WSW
1c
Stucco on gypsum
wallboard;
April 15, 2009
B20 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B1071.WSW
2a
Structural panel exterior
sheathing (plywood or
OSB) shear walls with
interior gypsum Wallboard,
with holddowns, basic
design strength
B1071.WSW
2b
structural panel exterior
sheathing (plywood or
OSB) shear walls with
stucco exterior and gypsum
Wallboard interior,
strength design with
seismic tie downs and
nail/screw details
B1072 Light Wood
Frame,
Diagonal Strut
Bracing
B1072.WDS1 light wood frame, diagonal
strut bracing
B20 Exterior
Enclosure
B2010 Exterior Walls B2011 Exterior Wall
Construction
See Section
B10 for all
structural walls
B2011.001a Precast concreteWest
Coast Practice
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B21
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B2011.001b Precast concreteEast Coast
Practice
B2011.002 GFRC cladding
B2011.003 EIFS Cladding
B2012 Parapets B2012.001a Unreinforced masonry 
unbraced
B2012.001b Unreinforced masonry 
retrofit per ASCE 4105
B2020 Exterior Windows B2021 Windows B2021.001 Window, aluminum frame,
sliding, standard glass, 125
square foot pane
B2021.002 Window, aluminum frame,
fixed, standard glass,
80"x80" pane
B2021.003 Windows, wood, double
hung, standard glass, 3'
1.5"x4'
B2021.004 Window, aluminum frame,
sliding, heavy sheet glass,
4'0x2'6"x3/16"
B2022 Curtain Walls
April 15, 2009
B22 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B2022.001a Annealed Monolithic, 6:5
aspect ratio (H:W), 1/4 in.
(6mm) thick glass, 0.43 in.
(11 mm) glass‐to‐frame
clearance, Kawneer 1600
curtain wall framing
B2022.001b Annealed Monolithic, 6:5
aspect ratio (H:W), 1/4 in.
(6 mm) thick glass, 0 in. (0
mm) glass‐to‐frame
clearance, Kawneer 1600
curtain wall framing
B2022.001c Annealed Monolithic, 6:5
aspect ratio (H:W),1/4 in.
(6mm) thick glass, 1/8 in.
(3 mm) glass‐to‐frame
clearance, Kawneer 1600
curtain wall framing
B2022.001d Annealed Monolithic, 6:5
aspect ratio (H:W),1/4 in.
(6 mm) thick glass, 1/4 in.
(6 mm) glass‐to‐frame
clearance, Kawneer 1600
curtain wall framing
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B23
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B2022.001e Annealed Monolithic, 2:1
aspect ratio (H:W), 1/4 in.
(6mm) thick glass, 0.43 in.
(11 mm) glass‐to‐frame
clearance, Kawneer 1600
curtain wall framing
B2022.001f Annealed Monolithic, 1:2
aspect ratio (H:W), 1/4 in.
(6 mm) thick glass, 0.43 in.
(11 mm) glass‐to‐frame
clearance, Kawneer 1600
curtain wall framing
B2022.002a Annealed Insulating Glass
Unit, 1/4 in. (6mm) thick
glass, 0.43 in. (11 mm)
glass‐to‐frame clearance,
Kawneer 1600 curtain wall
framing
B2022.002b Annealed IGU, 1/4 in.
(6mm) thick glass, 1/4 in.
(6 mm) glass‐to‐frame
clearance, Kawneer 1600
curtain wall framing
April 15, 2009
B24 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B2022.003 Asymmetric Annealed
Insulating Glass Unit, 1/4
in. (6 mm) inner AN mono
glass, 1/8 in. (3 mm) outer
AN lam. (0.030 PVB) outer
glass, avg. 0.43 in. (11 mm)
glass‐to‐frame clearance,
Kawneer 1600 curtain wall
framing
B2022.004 Asymmetric Annealed
Insulating Glass Unit, 1/4
in. (6 mm) inner AN mono
glass, 1/8 in. (3 mm)outer
AN lam. (0.060 PVB) outer
glass, avg. 0.43 in. (11 mm)
glass‐to‐frame clearance,
Kawneer 1600 curtain wall
framing
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B25
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B2022.005 Asymmetric Annealed
Insulating Glass Unit, 1/4
in. (6 mm) inner AN mono
glass, 1/4 in. (6 mm) outer
AN lam. (0.030 PVB) outer
glass, avg. 0.43 in. (11
mm) glass‐to‐frame
clearance, Kawneer 1600
curtain wall framing
B2022.006 Annealed Laminated, 1/4
in. (6 mm) thick glass
(0.030 PVB) square
corners, cut corner finish,
cut edge finish, 0.43 in.
(11 mm) glass‐to‐frame
clearance Kawneer 1600
curtain wall framing
B2023 Storefronts B2023.001 Annealed Monolithic,
square corners, cut corner
finish, cut edge finish, 5 ft
W x 6 ft H (1.5 m x 1.8 m)
size, 6:5 aspect ratio
(H:W), 1/4 in. (6 mm) thick
glass, 0.41 in. (10 mm) avg.
glass‐to‐frame clearance,
Kawneer TriFab II 451
Storefront curtain wall
framing
April 15, 2009
B26 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B2023.002 Annealed Insulating Glass
Unit, square corners, cut
corner finish, cut edge
finish, 5 ft W x 6 ft H (1.5
m x 1.8 m) size, 6:5 aspect
ratio (H:W), 1 in. (25 mm)
thick glass, 0.594 in. (15
mm) avg. glass‐to‐frame
clearance, Kawneer TriFab
II 451 Storefront curtain
wall framing
B2023.003 Annealed Laminated, 5 ft
W x 6 ft H (1.5 m x 1.8 m)
size, 6:5 aspect ratio
(H:W), 1/4 in. (6 mm) thick
glass (0.030 PVB) square
corners, cut corner finish,
cut edge finish, 0.41 in.
(10 mm) avg. glass‐to
frame clearance, Kawneer
TriFab II 451 Storefront
curtain wall framing
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B27
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
B30 Roofing B3010 Roof Coverings B3011 Roof Finishes B3011.001a Concrete roofing tiles that
are not individually
fastened to the roof
sheathing
B3011.001b Concrete roofing tiles that
are individually fastened to
the roof sheathing per
mfg's instructions (UBC
1994)
B3011.002a Clay and slate roofing tiles
that are not individually
fastened to the roof
sheathing
B3011.002b Clay and slate roofing tiles
that are individually
fastened to the roof
sheathing per mfg's
instructions (UBC 1994)
C Interiors C10 Interior
Construction
C1010 Partitions See Section
B10 for all
structural walls
C1011 Fixed
Partitions
C1011.001 Gypsum wallboard
partitions on metal studs
C20 Stairs C2010 Stair Construction C2011 Regular Stairs C2011.000
C2011.000a Stairs without slip joints
April 15, 2009
B28 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
C2011.000b Stairs with slip joints to
accommodate interstory
drift
C30 Interior
Finishes
C3010 Wall Finishes
C3020 Floor Finishes
C3027 Access
Pedastal
Flooring
C3027.000
C3027.001 Pedestals glued to floor
C3027.002 Pedestals anchored w/bolts
C3027.003 Pedestals anchored w/bolts
& braced and stringers tied
to pedestals.
C3030 Ceiling Finishes
C3032 Suspended
Ceilings
C3032.000
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B29
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
C3032.001 Lightweight soft DropIn
Tile (tiles weigh less than 2
psf) Ceilings; Aluminum T
bar Grid  note additional
variables may be depth of
plenum and area of ceiling
C3032.001a CISCA Zone 02 Compliant
Design and Installation
C3032.001b CISCA Zone 34 Compliant
Design and Installation
C3032.001c ASCE 705 SDC D
Compliant Design and
Installation w/o flexible
sprinkler drops
C3032.001d ASCE 705 SDC D
Compliant Design and
Installation with flexible
sprinkler drops
C3032.002 Heavyweight hard DropIn
Tile (tiles weigh more than
2 psf) ; Aluminum Tbar
Grid  note additional
variables may be depth of
plenum and area of ceiling
April 15, 2009
B30 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
C3032.002a CISCA Zone 02 Compliant
Design and Installation
C3032.002b CISCA Zone 34 Compliant
Design and Installation
C3032.002c ASCE 705 SDC D
Compliant Design and
Installation w/o flexible
sprinkler drops
C3032.002d ASCE 705 SDC D
Compliant Design and
Installation with flexible
sprinkler drops
D Services D10 Conveying D1010 Elevators & Lifts D1011 Passenger
Elevators
D1011.000
D1011.001a Traction elevators; Pre
1973 ASME 17.1 Elevator
Code Design & Installation
D1011.001b Traction elevators; Post
1973 ASME 17.1 Elevator
Code Design
D1011.002a Hydraulic elevators; Pre
1973 ASME 17.1 Elevator
Code Design & Installation
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B31
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D1011.002b Hydraulic elevators; Post
1973 ASME 17.1 Elevator
Code Design & Installation
D10 Conveying D1020 Escalators & Moving
Walks
D1021 Escalators D1021.000
D1021.001 Escalator not designed &
installed to accommodate
interstory drift of building
D1021.002 Escalator designed &
installed to accommodate
interstory drift of building
D20 Plumbing
D2020 Domestic Water
Distribution
D2021 Cold Water
Service
D2021.000
D2021.001a Piping; w/o seismic design
installation
D2021.001b Piping; ASME Design &
Installation
D2021.001c Piping; UBC/IBC Design &
Installation with Braces
April 15, 2009
B32 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D2021.001d Piping; Rod hung with
Thrust Restraints
D2022 Hot Water
Service
D2022.000
D2022.001a Piping; w/o seismic design
installation
D2022.001b Piping; ASME Design &
Installation
D2022.001c Piping; UBC/IBC Design &
Installation with Braces
D2022.001d Piping; Rod hung with
Thrust Restraints
D2023 Domestic
Water Supply
Equipment
D2023.000
D2023.001a Cold Water Supply Motor
Driven Pumps; Single Pad
Installations Unanchored
D2023.001b Cold Water Supply Motor
Driven Pumps; Single Pad
Installations Anchored
D2023.002a Hot Water Supply Motor
Driven Pumps; Single Pad
Installations Unanchored
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B33
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D2023.002b Hot Water Supply Motor
Driven Pumps; Single Pad
Installations Anchored
D2023.003a Water heaters; Self
standing
D2023.003b Water heaters; Lighttape
lateral restraint
D2023.003c Water heaters; Heavy
lateral restraint
D2030 Sanitary Waste D2031 Waste Piping
D2031.002a w/o seismic design
installation
D2031.002b ASME Design & Installation
D2031.002c UBC/IBC Design &
Installation with Braces
D2031.002d Rod hung with Thrust
Restraints
D2090 Other Plumbing Systems D2091 Gas
Distribution
D2091.000
D2091.000a w/o seismic design
installation
D2091.000b ASME Design & Installation
April 15, 2009
B34 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D2091.000c UBC/IBC Design &
Installation with Braces
D30 HVAC D3010 Energy Supply
D3012 Gas Supply
System
D3012.000 see fragilities for
D2091.000
D3020 Heat Generating Systems D3021 Boilers D3021.000
D3021.000a Unanchored
D3021.000b Anchored
D3022 Boiler Room
Piping &
Specialties
D3022.000 see fragilities for
D2021.000 and
D2022.000
D3030 Cooling Generating
Systems
D3031 Chilled Water
Systems
D3031.000
D3031.001 Packaged Circulating
Water Chillers
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B35
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D3031.001a Floor Mounted 
unanchored
D3031.001b Floor Mounted  anchored
D3031.002 Chilled water piping see
Fragilities for D2021.000
D3040 Distribution Systems D3041 Air
Distribution
Systems
D3041.000
D3041.001a Fans; Restrained
D3041.001b Fans; Isolated
D3041.002a HVAC ductwork rod hung
D3041.002b HVAC ductwork with sway
braces
D3041.003a Air Diffuser  w/o safety
wires
D3041.003b Air Diffuser  w/safety wires
D3042 Exhaust
Ventilation
Systems
D3042.000
D3042.001 Unreinforced brick
chimneys
D3042.002 Reinforced masonry and
precast reinforced concrete
chimneys
April 15, 2009
B36 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D3042.003 Lightweight chimneys
(insulated metallined) flues
in lightframe construction
D3043 Steam
Distribution
Systems
D3043.000 Steam Distribution
Systems, all see Fragilities
for D2022.000
D3044 Hot Water
Distribution
D3044.000 Hot Water Distribution, all;
see Fragilities for
D2022.000
D3045 Chilled Water
Distribution
D3045.000 Chilled Water Distribution,
all see Fragilities for
D3031.002
D3050 Terminal & Package
Units
D3051 Terminal Self
Contained
Units
D3051.000
D3052 Package Units D3052.000
D40 Fire
Protection
D4010 Sprinklers D4011 Sprinkler
Water Supply
D4011.000
D4011.001a Unbraced Sprinklers and
not installed per NFPA 13
D4011.001b Installed per NFPA 13 prior
to 1991
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B37
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D4011.001c Installed per NFPA 13
between 1991 and 2002
D4011.001d Installed per NFPA 13
between 2002 and 2007
D4011.001e Installed per NFPA 13 after
2007
D4012 Sprinkler
Pumping
Equipment
D4012.000
D4012.001 Motor Driven Fire Water
Pumps; See Fragilities for
D2022.001
D4020 Standpipes D4021 Standpipe
Water Supply
D4021.000
D4021.001a Installed w/o flexible
couplings and not per
NFPA13
D4021.001b Installed per NFPA13
D4022 Pumping
Equipment
D4022.000
D4022.001 Motor Driven Fire Water
Pumps; See Fragilities for
D2022.001
April 15, 2009
B38 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D50 Electrical D5010 Electrical Service &
Distribution
D5011 High Tension
Service & Dist.
D5011.000
D5011.001a Rod hung  not braced
D5011.001b Braced per UBC/IBC
D5012 Low Tension
Service & Dist.
D5012.000
D5012.002a Switchgear; Wall mounted
D5012.002b Switchgear; Floor mounted
 unanchored
D5012.002c Switchgear; Floor mounted
 anchored
D5012.003a Electrical Cabinets
including Motor Control
Centers; Unanchored
D5012.003b Electrical Cabinets
including Motor Control
Centers; Anchored w/o
plate washers
D5012.003c Electrical Cabinets
including Motor Control
Centers; Anchored per
IBC/ASCE705
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B39
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D5020 Lighting & Branch Wiring
D5022 Lighting
fixtures
D5022.000
D5022.001a Layin fluorescent lighting
fixtures without two or
more slack safety wires
D5022.001b Layin fluorescent lighting
fixtures with two or more
slack safety wires
D5022.002a Stemhung pendant
fluorescent fixtures without
safety wires in the stems
D5022.002b Stemhung pendant
fluorescent fixtures with
safety wires in the stems
D5022.002c Stemhung pendant
fluorescent fixtures with
safety wires in the stems +
lateral restraint wires
D5090 Other Electrical System
April 15, 2009
B40 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D5092 Emergency
Light & Power
Systems
D5092.000
D5092.001a Standby generator;
Unanchored
D5092.001b Standby generator;
Anchored w/o isolation
D5092.001c Standby generator;
Vibration isolated w/o
seismic stops
D5092.001d Standby generator;
Vibration isolated
w/seismic stops
D5092.002a Battery racks; Not
seismically designed
D5092.002b Battery racks; Braced for
seismic and batteries
secured
D5093 Floor Raceway
Systems, all
D5093.000
D5093.000a Not laterally braced
D5093.000b Laterally braced per
SMACNA
D5093.000c Laterally braced per
IBC/ASCE 7
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B41
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
D5094 Other Special
Systems &
Devices, all
D5094.000
D5094.001 Motor control center; See
Fragilities for D5012.003
E Equipment &
Furnishings
E10 Equipment E1010 Commercial Equipment
E1018 Office
Equipment
E1018.000
E1018.001a Computer equipment in
racks; unanchored
E1018.001b Computer equipment in
racks; anchored
E1018.002a Bookshelves; Unanchored
E1018.002b Bookshelves; Weakly
Anchored
E1018.002c Bookshelves; Well
Anchored/Restrained
E1018.003a File cabinets; Unanchored
E1018.003b File cabinets;
Anchored/Restrained
April 15, 2009
B42 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
E1020 Institutional Equipment
E1022 Library
Equipment
E1022.000
E1022.001a Shelving; unanchored
E1022.001b Shelving; anchored
E1028 Medical
Equipment
E1028.000
E1028.001a Cart supported items;
Unrestrained
E1028.001b Cart supported items;
Restrained by chains
E1028.001c Cart supported items;
Wheels locked
E1040 Industrial equipment E1041 Chemical
tanks
E1041.001a Unanchored
E1041.001b Anchored
E1042 Compressed
gas cylinders
E1042.001a Unanchored
E1042.001b Anchored
April 15, 2009
ATC58 B: Fragility Group, Normative Quantity B43
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 (Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3 (Category
description)
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 (Fragility
description)
E1043 Pallet storage
racks
E2022.030a Unanchored
E2022.030b Anchored but not installed
per RMI 16.1
E2022.030c Anchored  Designed and
Installed to RMI 16.1prior
to 2008
E2022.030d Anchored  Designed and
installed to RMI 16.1 after
2008
E20 Furnishings
E2020 Movable Furnishings E2021 Movable
Artwork
E2021.000
E2021.001 China and art on shelves
E2021.001a Unsecured and cabinet
doors unlatched
E2021.001b Secured or cabinet doors
latched
E2021.002a Wall mounted artwork;
standard installation
E2021.002b Wall mounted artwork;
seismic designed
installation
E2021.003 Miscellaneous Countertop
Items; See Fragility for
E2021.001a
April 15, 2009
B44 B Fragility Group, Normative Quantity ATC58
and Population Models
Table B1 Default Fragility Groups (continued)
Level
1 ID
Level 1 Basic
construction
group)
Level
2 ID
Level 2
(Group
category)
Level 3
ID
Level 3
(Categorydescription
Level 4
ID
Level 4
(Systems)
Level 5 ID
(Fragility)
Level 5 description
Fragility
E2022 Furniture &
Accessories
E2022.000
E2022.001 Large freestanding storage
furniture subject to
overturning
E2022.001a Unanchored or Secured
E2022.001b Weakly anchored/secured
E2022.001c Well anchored secured
E2022.002 Large freestanding
household electrical
appliances(unsecured)
E2022.003 Home entertainment
systems
E2022.004 Wall Mounted Televisions
and Video Monitors
E2022.005 Desktop computer system
unit and CRT/LCD monitor
E2022.005a Unrestrained
E2022.005b Restrained
April 15, 2009
ATC58 B: Fragility Group Normative Quantity B45
and Population Models
Table B2 Default Assignment Commercial Office
General Age of Construction = New
Number of Stories = Less than or equal to 5
Local Seismic Design Practice = High
Local Seismic Installation Practice = Average
Facility Seismic Safety Management Practice = Low
Default Inventory Item Assigned Description Fragility ID
Exterior Enclosure – Walls Highrise curtain wall system with annealed glass –
square corners
B2022.001
Interior Walls Drywall finish, 5/8in., 2 sides on 35/8in metal
studs, screw – no slipjoint –
electrical embedded
C1011.001a
Interior Ceilings lightweight acoustical ceiling 4’x2’ , aluminum tee
bar grid – installed, in accordance with ASCE 705
for high seismic – sprinkler drops protruding
C3032.001
Conveying – Elevator Hydraulic Passenger Elevator D1011.002
HVAC – Package Units Heating/Cooling Air Handling Units D3063.000
Desk top Computers Unsecured computer and CRT monitor E2022.011
Servers and Network Equip. 6 foot high unsecured rack of equipment E2022.011a
Tall Filing Cabinets 4 high front loading filling cabinets – not anchored. E2022.026a
Tall Book Cases 7 foot high unanchored book case E2022.029a
Table B3 Default Assignment Multifamily Residential
General Age of Construction = New
Construction Type = Wood frame
Number of Stories = Less than or equal to 2
Local Seismic Design Practice = High
Local Seismic Installation Practice = Average
Personal Seismic Safety Management Practice = Low
Default Inventory Item Assigned Description Fragility ID
Roof Covering 10 psf concrete, clay or slate roofing that are
individually fastened to the roof sheathing per
manufacturers instructions
B3011.002
Ceilings Gypsum Wall Board on wood joists C3033.001
Furniture & Accessories Unsecured china cabinet and refrigerator E2022.000
Household Entertainment Unsecured television and stereo units E2022.004
Desk top Computers Unsecured computer and CRT monitor E2022.011
April 15, 2009
B46 B: Fragility Group, Normative Quantity ATC58
and Population Models
Table B4 Normative Quantities for Commercial Office Occupancies
Fragility Specification Direction Normative Quantities
Number Name
B2022.001 Highrise curtainwall systems
with annealed glass
Direction 1 Direction 1 perimeter times story height at
each level
Direction 2 Direction 2 perimeter times story height at
each level
C1011.009a Drywall finish, 5/8in., 2 sides,
on 35/8in metal stud, screws
Direction 1 0.025 times the floor area times story
height; includes one door and one window
per 600sf floor area
Direction 2 0.025 times the floor area times story
height; assume one door and one window
per 600sf floor area
D1011.002 Hydraulic passenger elevators Unidirectional one elevator per 30,000 floor area (except
for single floor
C3032.001 Lightweight acoustical ceiling
4'x2' aluminum teebar grid
SEE COMMENT
Unidirectional Floor area below
D3063.000 Heating/Cooling Air Handling
Units, all
Unidirectional For 70000sf bldg: Hundred ton cooling
tower + one 600 cfm fan 20000 cfm air
handler, one 15 T multi zone AC. Prorate by
sf for larger or smaller by decimal unit (e.g.
100,000sf =100,000/70,000=1.4 units
E2022.011 Desktop computer system unit
and CRT monitor
Unidirectional Floor area divided by 400
E2022.026a Tall file cabinets Unidirectional Floor area divided by 300
E2022.029 Unanchored bookcases Unidirectional Floor area divided by 300
E2022.011a Computer system servers and
network equipment
Unidirectional One office system per 20,000sf
($50k).Prorate by square foot for larger or
smaller by decimal unit (e.g. 100,000sf
=100,000/20,000=5.0 units
Table B5 Normative Quantities for Residential Occupancy
Fragility Specification Direction Normative Quantities
Number Name
B3011.002 Concrete, clay, and slate roofing tiles
that are individually fastened to the
roof sheathing
Unidirectional 1.15 times the upper floor
area
E2022.011 Desktop computer system unit and
CRT monitor
Unidirectional 1.0 set per 600sf
E2022.004 Household entertainment
equipment
Unidirectional 1.0 set per 600sf
E2022.000 Furniture & Accessories, all 1.0 units per 1200sf with
$20k value
April 15, 2009
ATC58 B: Fragility Group Normative Quantity B47
and Population Models
Table B6 Default Time of Day and Day of Week Population Variations
(relative to Expected Peak Population) for Office and
Residential Occupancies
Office occupancy Residential occupancy
Time of Day Weekdays Weekends Weekdays Weekends
12:00 AM 0% 0% 100% 100%
1:00 AM 0% 0% 100% 100%
2:00 AM 0% 0% 100% 100%
3:00 AM 0% 0% 100% 100%
4:00 AM 0% 0% 100% 100%
5:00 AM 0% 0% 100% 100%
6:00 AM 0% 0% 80% 100%
7:00 AM 25% 0% 60% 100%
8:00 AM 50% 5% 40% 100%
9:00 AM 75% 5% 20% 75%
10:00 AM 100% 5% 20% 50%
11:00 AM 100% 5% 20% 50%
12:00 PM 50% 5% 20% 50%
1:00 PM 50% 5% 20% 50%
2:00 PM 100% 5% 25% 50%
3:00 PM 100% 5% 30% 50%
4:00 PM 75% 5% 35% 50%
5:00 PM 50% 5% 50% 50%
6:00 PM 25% 0% 67% 50%
7:00 PM 0% 0% 84% 50%
8:00 PM 0% 0% 100% 50%
9:00 PM 0% 0% 100% 75%
10:00 PM 0% 0% 100% 100%
11:00 PM 0% 0% 100% 100%
April 15, 2009
B48 B: Fragility Group, Normative Quantity ATC58
and Population Models
Table B7 Monthly Population Variations (Relative to Expected Peak
Population) for Office and Residential Occupancies
Office MultiUnit Residence
Month Weekday Weekday Weekday Weekend
Jan. 91% 100% 100% 100%
Feb. 95% 100% 100% 100%
March 100% 100% 100% 100%
April 100% 100% 100% 100%
May 95% 100% 100% 100%
June 100% 100% 100% 100%
July 95% 100% 100% 100%
Aug. 100% 100% 100% 100%
Sept. 95% 100% 100% 100%
Oct. 100% 100% 100% 100%
Nov. 95% 100% 100% 100%
Dec. 95% 100% 100% 100%
Reader’s note: The occupancy templates in PACT provide population modeling data as described in this
section for the following occupancies:
1 Commercial office
2a Education (k12): Elementary Schools
2b Education (k12): Middle Schools
2c Education (k12): High Schools
3 Healthcare
4 Hospitality
5 Multiunit residential
6 Research
7 Retail
8 Warehouse
April 15, 2009
ATC58 C: Default Structural Fragility Data C1
Appendix C
Default Structural Fragility
Data
C.1 Introduction
ASCE705, Minimum Design Loads on Buildings and Other Structures
(ASCE, 2005) identifies seventyeight lateralforceresisting systems for new
building construction. Many other framing systems have also been
constructed over the course of the past 50 years. In addition, these various
types of lateralforceresisting systems can be combined with many gravity
load carrying systems to form a very large number of possible building types.
Although it is possible to apply the methodology contained in these
Guidelines to a building of any construction type, to do so would require a
combination of laboratory test data and expert opinion on the behavior of
components of each structural system. Many users of these Guidelines are
unlikely to have either sufficient laboratory test data or expert opinions at
their disposal to develop such fragility specifications, so default fragility data
for common structural components and systems are provided.
A classification is used to group all possible framing systems into a modest
number of systems for which default fragility and loss data are provided. The
classification system includes categorization of the gravityloadresisting
system as well as two components of seismic structural framing, vertical and
horizontal (diaphragms).
Reader Note: This draft of the Guidelines contains default data for only two
structural systems: momentresisting steel frames supporting concrete filled
metal deck and light wood frame bearing wall construction with plywood
sheathed diaphragms. Subsequent versions of these Guidelines will contain
additional fragility data.
C.2 Vertical Seismic Framing Systems
The classification system for vertical seismic framing systems is presented in
Table C1. The classification system is composed of the following four
levels:
Level I: Structural Material/System Designation: defines the system based on
the primary construction materials and system type/configuration.
April 15, 2009
C2 C: Default Structural Fragility Data ATC58
Level II: Seismic Ductility: up to four seismic ductility classes are identified
for each structural system/material type, which will generally reflect the
extent to which inelastic system behavior will be limited to predictable and
ductile behavioral modes. In modern design procedures, capacity design and
detailing rules are invoked to control these properties for some structural
systems. However many modern buildings, particularly in zones of low and
moderate seismicity and most archaic buildings have not been designed in
this manner. Given the wide variation in deformation capacities among
various structural system/material types, the seismic ductility classes are
defined primarily in a relative sense, although, some consideration is given to
parity between systems and materials. The four seismic ductility classes are
as follows:
Limited (that a structure has definable lateral strength but no specific
detailing or provisions to ensure ductility)
Low (limited detailing to provide inelastic response capability,
comparable to requirements for ordinary systems requirements per
ASCE 705)
Moderate (comparable to requirements for intermediate systems per
ASCE 705)
High (extensive detailing and other provisions to assure ductile behavior,
comparable to requirements for special systems per ASCE 705)
Level III: Seismic Behavior Characteristics: distinguishes between
behavioral aspects that are not sufficiently reflected in the seismic
performance class. For example, the level III designation is used to
distinguish between alternative bracing configurations for concentrically
braced steel frames (e.g. “X”braced and “V”braced patterns) that might
significantly impact behavior and are reflected in current building codes.
Level IV: Numerical Designation: provides a structured alphanumeric
description for each framing system of level III.
The fifth column in Table C1 provides supplemental guidance to the user for
the purpose of selecting appropriate default fragility curves to represent a
particular building structure. The guidance is not absolute because adherence
to a specific code or guideline does not guarantee a certain level of
performance because the performance of codecompliant buildings has never
been quantified. The sixth column identifies each Level IV vertical seismic
framing system by number.
Reader Note: Default fragility specifications will eventually be provided for
all listed Level II systems and some Level III systems.
April 15, 2009
ATC58 C: Default Structural Fragility Data C3
C.3 Horizontal Seismic Framing Systems (Diaphragms)
Table C2 presents the classification system for horizontal framing systems.
The intent of the classifications presented in Table C2 are to articulate
attributes that distinguish between diaphragm systems insofar as they affect
the overall performance of the seismic framing system, including the
likelihood of partial or total building collapse, and affect the assessment of
structural damage and the associated repair measures and losses. The floor
diaphragm systems can be matched with any vertical seismic framing system
(Table C1) and gravity framing system (Table C3).
The diaphragm classification system focuses on defining the materials and
characteristics of the system, particularly with regard to toughness and
tendency for damage either at joints within the diaphragm or at attachments
to the supporting gravity and lateral force resisting systems. The
classification system is not intended to explicitly address diaphragm
flexibility and its effect on building response, since the calculation of
diaphragm deformations should be addressed by analysis (see Chapter 6).
Whereas stiffness and the damageability are related, the two aspects of
response are uncoupled herein. The diaphragm classification system is
composed of the following three levels:
Level I: Structural Material/System Designation: defines the system based on
the primary construction materials and diaphragm configuration.
Level II: Seismic Toughness: toughness of a diaphragm is classified as low,
moderate or high. Toughness is a somewhat subjective classification, which
relates to whether the diaphragms and their connections are expected to be a
weak link in the seismic framing system. For systems with high toughness,
diaphragm damage should be minimal. For systems designated as having low
toughness, the diaphragm performance and damage to its components are
expected to have a major effect on the damage and loss. Systems with
moderate toughness lie between these two extremes.
Level III: Seismic Behavioral Characteristics: provides a mechanism to
distinguish between behavioral aspects that are not sufficiently reflected in
the seismic toughness class.
C.4 Gravity Framing Systems
The intent of the gravity framing system classifications are to articulate
attributes that distinguish between gravity systems insofar as they affect the
overall building seismic performance. The gravity framing system
April 15, 2009
C4 C: Default Structural Fragility Data ATC58
classifications can be matched with various lateral and diaphragm systems,
although in practice the number of variants is limited.
The key aspects of the gravity framing system classification schemes relate
to their ability to undergo deformations associated with seismic framing
system and their tendency for damage at the slab/beamtocolumn
connections or for damage in the vertical support elements (columns or
walls). The proposed classification system is composed of the following
three levels:
Level I: Structural Material/System Designation: defines the system based on
the primary construction materials and configuration.
Level II: Deformation Class: three classes of deformation capacity are used,
low, moderate and high, which relate to the ability of the system to sustain
deformations associated with story drift before major damage is reached.
Level III: Seismic Behavior Characteristics: identifies the components of the
system associated with the designated deformation class for common
construction.
C.5 Default Fragility Data
Figures C1 and C2 present default fragility data for steel momentresisting
frames and light wood frame bearing wall construction, respectively.
April 15, 2009
ATC58 C: Default Structural Fragility Data C5
Table C1 Classification System for Vertical Seismic Framing Systems
Material/System
Designation
Seismic
Ductility
Seismic Behavior Characteristics
Numerical
Designation
Code Equivalency No.
Steel Moment Frame (S1)
Limited Basic strength design w/o special detailing S1MF1a
2005 AISC Specification
(w/o seismic detailing);
pre1985 UBC (strength
design)
1
Low
S1MF1a with masonry or concrete infill walls S1MF2a 2
Strength design with limited rotation capacity in beam
column connections.
S1MF2b 2005 AISC – OMF 3
Seismically designed for high ductility but with pre
Northridge connection details
S1MF2c
19851994 UBC (pre
Northridge)
4
Moderate
Strength design with prequalified capacity design
provisions for beamcolumn connections and member
slenderness.
S1MF3 2005 AISC – IMF 5
High
Seismically designed and detailed with prequalified
capacity design provisions. Wshape columns
shallower than W18 or box columns
S1MF4a 2005 AISC  SMF 6
S1MF4a with Wshape columns deeper than W18 S1MF4b 7
Steel Braced Frame (S2)
Limited
Concentric or eccentrically braced system. Basic
strength design w/o special detailing
S2BF1
2005 AISC Specification
(w/o seismic detailing)
8
Low
Concentrically braced system; strength design with
capacity design procedures for connections, compact
bracing member compactness.
S2CBF2a 2005 AISC – OCBF 9
S2CBF2a with braces other than HSS members S2CBF2b 10
S2CBF2a with Kbrace configuration S2CBF2c 11
Moderate
Concentrically braced system. Seismically designed
and detailed with slenderness/compactness limits on
braces and capacity design requirements for
connections, beams, and columns.
S2CBF3a
2005 AISC – SCBF
12
S2CBF3a with braces other than HSS members S2CBF3b 13
April 15, 2009
C6 C: Default Structural Fragility Data ATC58
Table C1 Classification System for Vertical Seismic Framing Systems
Material/System
Designation
Seismic
Ductility
Seismic Behavior Characteristics
Numerical
Designation
Code Equivalency No.
S2CBF3a with chevron configuration braces S2CBF3c 14
S2CBF3a with Xconfiguration braces S2CBF3d 15
S2CBF3a with beambracecolumn connections
designed as moment resisting
S2CBF3e 16
High
Concentrically braced system with buckling restrained
braces and capacity design of connections, beams, and
columns.
S2CBF4 2005 AISC – BRBF 17
Steel Eccentrically Braced
Frame (S3)
Moderate
Eccentrically braced system designed and detailed with
capacity design provisions; flexural links remote from
column
S3EBF3a 2005 AISC – EBF 18
Eccentrically braced system designed and detailed with
capacity design provisions; flexural links adjacent to
column
S3EBF3b 2005 AISC – EBF 19
High
Eccentrically braced system designed and detailed with
capacity design provisions. Shear links adjacent to
column
S3EBF4a 2005 AISC – EBF 20
S3EBF1a but with shear links away from column S3EBF4b 2005 AISC – EBF 21
S3EBF1a but with shear links away from column
with beambracecolumn connections designed as
moment resisting
S3EBF4c 2005 AISC – EBF 22
Steel Light Frame (S4)
Limited
Steel moment frame structure with light gravity load
(i.e., a metal building). Basic strength design w/o
special detailing
S4LMF1
2005 AISC Specification
(w/o seismic detailing)
23
Low
Steel framed structure with light gravity load (i.e., a
metal building). Basic strength design with capacity
design of connections.
S4LMF2
2005 AISC – OMF for
systems with low gravity
load
24
Concrete Moment Frame
(C1)
Limited Basic strength design w/o special detailing C1MF1
2005 ACI318 (w/o
seismic detailing)
27
April 15, 2009
ATC58 C: Default Structural Fragility Data C7
Table C1 Classification System for Vertical Seismic Framing Systems
Material/System
Designation
Seismic
Ductility
Seismic Behavior Characteristics
Numerical
Designation
Code Equivalency No.
Low
Basic strength design with continuous rebar at beam
column connections.
C1MF2 2005 ACI318, OMF 28
Moderate
Strength design with continuous rebar at beamcolumn
connections and column confinement rebar.
C1MF3 2005 ACI318, IMF 29
High
Seismically designed and detailed with capacity design
provisions, including precast systems that utilize pre
qualified connections.
C1MF4 2005 ACI318, SMF 30
Concrete Shear Wall (C2)
Limited
Basic strength design with deficient (nonconforming)
reinforcement details
C2W1a
Pre1976 UBC (strength
design)
31
Low
Low gravity load, basic strength design w/o seismic
detailing
C2W2a
2005 ACI318 (non
seismic)
32
High gravity load, basic strength design w/o seismic
detailing
C2W2b 33
Precast shear wall (including tiltup), basic strength
design w/o seismic detailing
C2W2c 34
Moderate
Low gravity load, squat shear wall, strength design. C2W3a
2005 ACI318, including
Chapter 21 provisions
35
Low gravity load, slender shear wall, strength design
with seismic detailing for shear and boundary zone
confinement
C2W3b 36
Same as C2W3b but with coupling beams C2W3c 37
High gravity load, squat shear wall, strength design. C2W3d 38
High gravity load, slender shear wall, strength design
with seismic detailing for shear and boundary zone
confinement
C2W3e 39
Same as C2W3e but with coupling beams C2W3f 40
April 15, 2009
C8 C: Default Structural Fragility Data ATC58
Table C1 Classification System for Vertical Seismic Framing Systems
Material/System
Designation
Seismic
Ductility
Seismic Behavior Characteristics
Numerical
Designation
Code Equivalency No.
Precast shear wall, strength design with seismic
detailing for wall anchorage, shear and boundary zone
confinement
C2W3g 41
Reinforced Masonry
Bearing Walls (RM)
Limited Basic strength design w/o special detailing RM1a 43
Moderate Strength design with seismic detailing RM2a 44
Unreinforced Masonry
Bearing Walls (URM)
Limited
Basic strength design w/o special detailing URM1a
Existing URM
construction
45
Same as URM1a, but with tiebacks to diaphragm to
help avoid out of plane failures
URM1b
Retrofitted URM
construction in high
seismic regions, or new
URM in low seismic
regions
46
Light Wood Frame Shear
Wall (W1)
Limited
Structural panel sheathing (plywood or OSB) shear
walls with interior gypsum wall board, basic strength
design
WSW1a 47
Structural panel sheathing with stucco exterior and
gypsum wall board interior, basic strength design
WSW1b 48
Stucco on gypsum wallboard WSW1c 49
Low
Structural panel sheathing (plywood or OSB) shear
walls with interior gypsum wall board, strength design
with seismic tie downs and nail/screw details
WSW2a 50
Structural panel sheathing (plywood or OSB) shear
walls with stucco exterior and gypsum wall board
interior, strength design with seismic tie downs and
nail/screw details
WSW2b 51
April 15, 2009
ATC58 C: Default Structural Fragility Data C9
Table C2 Classification System for Horizontal Seismic Framing Systems (Diaphragms)
Material/System
Designation
Seismic
Toughness
Seismic Behavioral Characteristics
Numerical
Designation
No.
Concrete Slab on
Ribbed Steel Deck
Moderate
Welded wire fabric reinforcement; decking attached to
framing by puddle welds
CSD2a 1
High
Connected to supporting framing through welded shear
studs, reinforcing bars at critical locations for shear and
inplane flexure in slab
CSD3a 2
Ribbed Steel Deck
Low
Connected by intermittent plates and screws to seismic
framing system.
SD1a 3
Moderate
Connected by studs or puddle welds to provide
diaphragm continuity and connection to the seismic
framing system.
SD2a 4
CastInPlace
Concrete Slab
Moderate
Connected to supporting framing by nominal
reinforcement.
CS2a 5
High
Connected to supporting framing by reinforcing bars or
other anchorage, slab reinforcing bars at critical
locations for shear and inplane flexure in slab
CS3a 6
Precast Concrete Slab
or Slab/Joists
Low
Connections between slab/joists and at boundary
connections designed for strength only.
PC1a 7
Moderate
Detailed for ductile connections between slab/joists and
at boundary connections
PC2a 8
Same as PC3a, but with cast in place topping slab with
provisions made for bond/shear transfer between
topping and precast panels with welded wire fabric.
PC2b 9
High
Same as PC3a, but with cast in place topping slab with
provisions made for bond/shear transfer between
topping and precast units using rebar.
PC3a 10
Wood Sheathing with
Structural Panels
Low
Structural sheathing panels with minimum nailing to
supporting framing.
SPA1a 11
Moderate
Structural sheathing panels with close nailing to
supporting framing to provide continuity and shear
transfer.
SPA2a 12
Structural sheathing panels with close nailing to
supporting framing to provide continuity and shear
transfer with concrete topping slab
SPA2b 13
Wood Sheathing with
Structural Planks
Low
Diagonal or straight plank sheathing with nailing
designed for shear transfer.
SPL1a 14
Moderate
Diagonal or straight plank sheathing with nailing
designed for shear transfer and with a concrete topping
slab
SPL2a 15
Steel Bracing
Low Diagonal rod or other steel bracing SR1a 16
Moderate
Diagonal rod or other steel member bracing designed
using capacity design principles to prevent connection
failure prior to significant yielding either in bars or the
seismic lateral system
SR2a 17
April 15, 2009
C10 C: Default Structural Fragility Data ATC58
Table C3 Classification System for Gravity Framing Systems
Material/System
Designation
Deformation
Class
Seismic Behavior Characteristics
Numerical
Designation
No.
Composite Metal Deck
Supported by Steel Beams,
Girders and Columns
Moderate
Concrete over metal deck or metal deck
supported by rolled steel beams or openweb
joists and HSS section columns
MD2a 1
High
Concrete over metal deck or metal deck
supported by rolled steel beams or open web
joists and rolled steel columns
MD3a 2
CastinPlace Concrete
TwoWay, Flat Slab
Supported by Concrete
Columns
Low
Lightly reinforced and posttensioned slabs not
designed to transfer shear/flexural loads into
columns
CIP1a 3
Moderate
Lightly reinforced and posttensioned slabs not
specifically designed to transfer shear/flexural
loads into columns, but include structural
integrity reinforcement
CIP22a 4
High
Lightly reinforced and posttensioned slabs
specifically designed to transfer shear/flexural
loads into columns, including structural
integrity reinforcement
CIP23a 5
CastinPlace Concrete
OneWay Slab Supported
by Concrete Joists, Beams,
and Columns
Low
Oneway slab with reinforcement in one
direction supported by concrete joists, beams,
and columns
CIP11a 6
Moderate
Oneway slab with reinforcement in both
directions supported by concrete joists, beams,
and columns
CIP12a 7
Precast Concrete Planks or
Beams Supported by
Concrete Beams and
Columns
Low
Precast concrete planks without topping slab
supported by concrete beams and columns
PC1a 8
Moderate
Precast concrete planks with topping slab
integrating the planks and the supporting
beams, supported by concrete columns
PC2a 9
Wood Joists Supported by
Beams, Columns, and
Bearing Walls
Moderate
Wood structural flooring supported by wood
joists beams, columns, and bearing walls
W2a 10
High
Wood structural flooring supported by wood
joists, beams, and columns
W3a 11
April 15, 2009
ATC58 C: Default Structural Fragility Data C11
Figure C1 Fragility specification for post1994 welded steel moment frame
April 15, 2009
C12 C: Default Structural Fragility Data ATC58
Figure C2 Fragility specification for exterior wall with structural sheathing and cement plaster
April 15, 2009
ATC58 C: Default Structural Fragility Data C13
DS1 DS2 DS3
cracked GWB Major cracks, buckling of gypsum wallboards at
corners of walls
Wood stud failure, sill plate splitting and failure,
gypboard wall panel failure
0.0020 0.010 0.030
0.40 0.40 0.40
Taping, sanding, and painting. Replacing gypboard panels, and then taping,
sanding and painting
Replace entire partition walls
Includes no mechanical repairs
Includes no door repairs
Cost per sq ft of curtain wall
Max. cost up to lower quantity $ 3.33 $ 5.56 13.22
100 100 100
Min cost over upper quantity $ 2.94 $ 4.76 $ 11.22
10,000 10,000 10,000
Beta (cost) 0.20 0.20 0.20
Minor repairs can be done in a couple days
Demo, new gyp, taping and painting will take at least
one week
New framing, possible mechanical inspection and/or
repair, new gyp, taping and painting will take at least
2 weeks.
The picture above for DS3 shows a 2x6 wall. Repair
costs are for 2x4 per the damage state info given to
ARG
The picture above for DS3 shows a 2x6 wall. Repair
costs are for 2x4 per the damage state info given to
ARG
The picture above for DS3 shows a 2x6 wall. Repair
costs are for 2x4 per the damage state info given to
ARG
Fragility, damage measures, and consequence functions for
Interior Walls GWB on Wood studs C1011.001a
BASIC COMPOSITION
GypsumWall Board constructed as follows; ½in GWB, laid horizontally, #6 1 ¼in standard drywall screws 16 in OC vertically to 2x4 framing (i.e., not fixed at
T&B), 1st level has sillplate anchored to slab w ½in Abs 32 in OC, bottomplate of upperlevel ptns are nailed to diaphragm w 16d @ 16 in OC, Top connections
at bottom level: A35 @ 32 in OC
NORMATIVE QUANTITIES Square feet of wall area in a given level in a specific direction
DAMAGES STATES, FRAGILIITES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS
MEDIAN EDP
(interstory drift in specified direction)
BETA
CORRELATION (%)
REPAIR MEASURES
Figure C3 Fragility specification for interior wall with wood studs and gypsum board sheathing
April 15, 2009
ATC58 D: Nonstructural Fragility Specifications D1
Appendix D
Nonstructural Fragility
Specifications
D.1 Default Nonstructural Fragility Data Tables
Figures D1 through D13 present the default fragilities contained within the
PACT database.
DS1 DS2 DS3
Visible damage and small cracks in gypsumwallboard
that can be repaired with taping, patching and
painting. No window and door damage
Extensive cracking or crushing in gypsumwallboard
and minimal or no damage to metal studs. Rehang
door.
Severe damage to gypsumwallboard and enough
damage to metal studs and runners to require
replacement
0.0025 0.0060 0.014
0.70 0.50 0.40
low low low
Taping, patching and painting Replacing the gypsumboards, and then taping,
and painting
Remove damaged materials
Reframe walls
Repair damaged electrical
Install new gypsumwallboard
Tape, sand and paint
Rehang doors Includes some door repairs
Includes some minor mechanical repairs
Cost per sq ft of interior partition
Max. cost up to lower quantity $3.33 $5.56 $12.92
100 100 100
Min cost over upper quantity $2.94 $4.76 $10.92
10,000 10,000 10,000
Beta (cost) 0.20 0.30 0.30
Minor repairs can be done in a couple of days,
including repainting patches
Larger repairs take min. 5 days to remove damage,
patch and paint
Major reframing of walls and mechanical damage
repairs will take a minimumof 10 days, and possible
longer depending on crew size, amount of damage,
and availability of mechanical parts for repairs
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS
MEDIAN EDP
(interstory drift in specified direction)
BETA
CORRELATION (%)
REPAIR MEASURES
DAMAGES STATES, FRAGILIITES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
Fragility, damage measures, and consequence functions for
Interior Partitions Type 9a C1011.009a
BASIC COMPOSITION Full height 5/8 inch gypsumboard screwed on metal studs. No slip track or window panels.
NORMATIVE QUANTITIES Square feet of wall area at given level in a specfied direction
Figure D1 Fragility for interior partitions
April 15, 2009
D2 D: Nonstructural Fragility Specifications ATC58
DS1 DS2
Glass cracking Glass fallout
0.031 0.034
0.3 0.3
low low
Replace the cracked glass panel(s) Replace the glass panel(s) where glass has fallen
out or cracked
Set up scaffold to access façade
Repair damaged framework
Cost per sq ft of curtain wall
Max. cost $ 54.00 $ 59.74
lower quantity 250 250
Min cost $ 29.00 $ 32.74
upper quantity 10,000 10,000
Beta (cost) 0.20 0.20
Lead time on framework is min. 15 days. Tempured
glass is min. 10 days
Lead time on framework is min. 15 days. Tempured
glass is min. 10 days
Fragility, damage measures, and consequence functions for
Exterior SkinGlass CurtainwallType 1 B2022.001
BASIC COMPOSITION
Full height glass curtain wall made of annealed glass with square glazed corners, thicknesses in the range
of 6 mm with height to width ratio of about 1.2 and with 11 mm clearances.
NORMATIVE QUANTITIES Square feet of curtain wall oriented in a specified direction per floor
DAMAGES STATES, FRAGILIITES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS
MEDIAN EDP
(interstory drift in specified direction)
BETA
CORRELATION (%)
REPAIR MEASURES
Figure D2 Fragility for unitized glazed curtainwall
April 15, 2009
ATC58 D: Nonstructural Fragility Specifications D3
DS1 DS2
Some tiles displaced and fallen, usually in the
perimeter of the room.
Extensive tile fallout and T bar grid failure.
Mechanical systems damaged
Photo courtesy National Information Service for
Earthquake Engineering, University of California,
Berkeley
Photo courtesy National Information Service for
Earthquake Engineering, University of California,
Berkeley
0.55g 1.00g
0.40 0.40
ASSUMPTIONS
Assumed leakage of sprinklers associated with
ceiling collapse, but at a higher level.
Remove all damaged materials
Inspect structure for damage
Inspect all mechanical systems
Reinstall new Tbar
Repair damaged mechanical system
Install new ceiling tile
Replace displaced tiles, and replace damaged tiles Remove all damaged tile and grid
Repair mechanical damage
Replace grid, tile and mechanical trims as
necessary.
Cost per sq ceiling system
Max. cost up to lower quantity $5.26 $26.81
100 100
Min cost over upper quantity $3.89 25
10,000 10,000
Beta (cost) 0.40 0.40
Minimum3 days for tile and grid
Fragility, damage measures, and consequence functions for
Ceiling Systems Suspended acoustical tile type1 C3032.001
BASIC COMPOSITION
Suspended ceiling systemwith T bars and acoustical ceiling tiles (2x4), 16x16 ft in plan, ceiling tiles not
exceeding 2 lbs/sf, installed in accordance with ICBO standard 252 or IBC (CISCA zones 3 and 4 and
ASTM 635 and 636), no tile retainer clip.
NORMATIVE QUANTITIES Square feet of floor area at a specified level.
DAMAGES STATES, FRAGILIITES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
MEDIAN EDP
(peak floor accelerartion)
BETA
CORRELATION (%)
Assumed leakage of sprinklers associated with
ceiling collapse, but at a higher level.
REPAIR MEASURES
Turn off all mechanical systems. Remove fallen
debris. Remove any residual water. Verify sound
condition of ducting. Verify that structure above is
sound. Install new ceiling tiles.
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS Minimum1 day for tile replacement only
Figure D3 Fragility for suspended acoustic ceiling systems
April 15, 2009
D4 D: Nonstructural Fragility Specifications ATC58
DS1 DS2
Cracking of ceiling to wall joints Total ceiling Collapse
0.010 0.050
0.30 0.60
Total ceiling replacement
Cost per sq ft of curtain wall
Max. cost up to lower quantity $ 2.86 $ 4.99
Min cost over upper quantity $ 2.33 $ 4.05
Beta (cost) 1 2
One week to patch and repaint entire ceiling area
One two weeks for reframing and new sheetrock.
Indeterminate time for mechanical repairs. If
mechanical repairs are minor they could be done in 5
days ready for inspection and close up.
Taping and painting
HVAC, lighting, and sprinkler drops will need repair
measures, and inspection.
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS
MEDIAN EDP
(peak story drift in specified direction)
BETA
CORRELATION (%)
REPAIR MEASURES
DAMAGES STATES, FRAGILIITES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
Fragility, damage measures, and consequence functions for
Ceiling GWB on wood joists C3033.001
BASIC COMPOSITION
½in gypsumwallboard ceilings installed same as walls.
#6 1 ¼in standard drywall screws 11 in OC parallel to joists or rafters
NORMATIVE QUANTITIES Square feet of ceiling area per floor
Figure D4 Fragility for gypsum ceiling on wood joists
April 15, 2009
ATC58 D: Nonstructural Fragility Specifications D5
DS1 DS2
Tile dislodged – requires repair Major portion of tiles dislodged, requires full roof
replacement
Image fromXiao, Y. and H.W Yun, 1998. Dynamic
testing of fullscale concrete and claytile roof
models. ASCE J ournal of Structural Engineering,
124 (5), May1998, 482489
Image fromXiao, Y. and H.W Yun, 1998. Dynamic
testing of fullscale concrete and claytile roof
models. ASCE Journal of Structural Engineering,
124 (5), May1998, 482489
1.50g 1.90g
0.40 0.40
Replace full roof
Cost per sq ft of repair/replacement
Max. cost $ 5.06 $ 10.85
up to lower quantity 1,000 1,000
Min cost $ 4.03 $ 9.85
over upper quantity 10,000 10,000
Beta (cost) 0.30 0.2
Assume miunimum 1 dayfor tile replacement.
Procurement could take 26 weeks depending on
pattern and availability
Scaffold installation and roof demo could take 10
days.
Assume miunimum5 days for tile replacement.
Procurement could take 26 weeks depending on
pattern and availability
TIMEFRAME FOR REPAIRS
Fragility, damage measures, and consequence functions for
Exterior Roofing Concrete tile type 2 B3011.002
BASIC COMPOSITION Concrete tile weighing 10 psf attached per mfg. recommendations, each tile nailed to roof
NORMATIVE QUANTITIES Square feet of roof area
DAMAGES STATES, FRAGILIITES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
Repair or replace dislodged tiles. Assume repair to
20% 0f roof area.
Assumes that plywood underlayment is intact and
needs no repairs
CONSEQUENCE FUNCTION
MEDIAN EDP
(peak horizontal roof acceleration)
BETA
CORRELATION (%)
REPAIR MEASURES
Figure D5 Fragility for concrete tile roofs
April 15, 2009
D6 D: Nonstructural Fragility Specifications ATC58
DS1 DS2 DS3
Elevator does not work (because of various reasons,
see related table). Major repairs needed
TBD
0.40 g
0.30
Repair elevator (depends on the type of damage,
see related table).
Cost per unit of conveyance
Max. cost $ 56,000.00
up to lower quantity 3
Min cost $ 33,600.00
over upper quantity 5
Beta (cost) 0.20
Cost per unit of conveyance
Fragility, damage measures, and consequence functions for
Conveying Hydraulic elevator D1011.002
BASIC COMPOSITION Single Cab Hydraulic 3 stop passenger elevator 3500 lb capacityw/ standard finishes
NORMATIVE QUANTITIES Number of Elevators per Building
DAMAGES STATES, FRAGILIITES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
MEDIAN EDP
(Peak Ground Acceleration)
BETA
CORRELATION (%)
REPAIR MEASURES
CONSEQUENCE FUNCTION
Timeframe for Repairs Assume that parts will take a minimumof 10 days.
Some parts could take up to 12 weeks
Figure D6 Fragility for hydraulic elevators
April 15, 2009
ATC58 D: Nonstructural Fragility Specifications D7
DS1
Inoperative
Photo credit: Mason Industries (2002)
1.60g
0.50
ASSUMPTIONS
Cost per each
Max. cost up to lower quantity $ 220,000.00
2
Min cost over upper quantity $ 150,000.00
8
Beta (cost) 0.6
NORMATIVE QUANTITIES
CORRELATION (%)
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
REPAIR MEASURES
Fragility, damage measures, and consequence functions for
D3063.000 Roof Mounted Equipment
BASIC COMPOSITION
Chillers, fans, air handlers on vibration isolators with restraints and anchorage designed and installed per ASCE 7 requirements for normal occupancy. Flexible
utility lines provided.
Number of HVAC Units
DAMAGES STATES, FRAGILITIES, AND CONSEQUENCE FUNCTIONS
MEDIAN EDP
(peak horizontal acceleration of roof)
BETA
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS
Figure D7 Fragility for roofmounted mechanical equipment
April 15, 2009
D8 D: Nonstructural Fragility Specifications ATC58
DS1
Fall off shelf, shelf over turns, objects break
Filiatrault et al. (2001)
0.20g
0.50
low
ASSUMPTIONS
Cost per each
Max. cost up to lower quantity 20k
Min cost over upper quantity 20k
Beta (cost) 0.4
Fragility, damage measures, and consequence functions for
E2022.000 Miscellaneous housewares and art objects
BASIC COMPOSITION Miscellaneous houswares,'China and art objects, includes China cabinet – assumed unanchored
NORMATIVE QUANTITIES Cost of Replacement per 1200 square feet
DAMAGES STATES, FRAGILITIES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
MEDIAN EDP
(peak horizontal floor acceleration)
BETA
CORRELATION (%)
REPAIR MEASURES
Replace
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS
Figure D8 Fragility for miscellaneous housewares and art objects
April 15, 2009
ATC58 D: Nonstructural Fragility Specifications D9
DS1
Falls off wall, slides off shelf, does not function
Filiatrault et al. (2001)
0.20g
0.50
ASSUMPTIONS
Cost per each
Max. cost up to lower quantity $2,500.00
Min cost over upper quantity $2,500.00
Beta (cost) 0.4
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS
REPAIR MEASURES
Replace
MEDIAN EDP
(peak horizontal floor acceleration)
BETA
CORRELATION (%)
DAMAGES STATES, FRAGILITIES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
Fragility, damage measures, and consequence functions for
E2022.004 Home Entertainment Equipment
BASIC COMPOSITION Speakers or televisions resting on shelves, stereos, etc.  assume unanchored.
NORMATIVE QUANTITIES 1 set per 600 square of floor area
Figure D9 Fragility for home entertainment equipment
April 15, 2009
D10 D: Nonstructural Fragility Specifications ATC58
DS1
Computer falls, becomes inoperative
Photo courtesy National Information Service for
Earthquake Engineering, University of California,
Berkeley
1.20g
0.60
ASSUMPTIONS
Cost per each
Max. cost $2,500.00
up to lower quantity 10
Min cost over upper quantity $1,000.00
over upper quantity 100
Beta (cost) 0.4
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS
REPAIR MEASURES
Repair/replace
MEDIAN EDP
(peak horizontal floor acceleration)
BETA
CORRELATION (%)
Computer hard drive parks when the unit falls (as in
Mac Powerbooks) or when power is lost (as in most
other computers), so systemunit is rugged
(nondamageable).
DAMAGES STATES, FRAGILITIES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
Fragility, damage measures, and consequence functions for
E2022.011 Desktop Computers
BASIC COMPOSITION Monitors and computers  assume unsecured and computer are on desk and not on floors.
NORMATIVE QUANTITIES Number of sets per floor
Figure D10 Fragility for desktop computer equipment
April 15, 2009
ATC58 D: Nonstructural Fragility Specifications D11
Figure D11 Fragility for servers and network equipment
April 15, 2009
D12 D: Nonstructural Fragility Specifications ATC58
Figure D12 Fragility for tall filing cabinets
April 15, 2009
ATC58 D: Nonstructural Fragility Specifications D13
DS1
Overturns, contents fall out, fragile contents break.
Image courtesy of National Information Service for
Earthquake Engineering, University of California,
Berkeley
0.40g
0.30
ASSUMPTIONS
Cost per each
Max. cost $150.00
up to lower quantity 6
Min cost $75.00
over upper quantity 50
Beta (cost) 0.20
Fragility, damage measures, and consequence functions for
E2022.029 Unanchored Bookcase
BASIC COMPOSITION
Includes 6 foot and 8 foot high book cases. Fully loaded, not secured to wall or floor, ½ inch gap between back of bookcase and gypboardmetal stud wall.
NORMATIVE QUANTITIES Number of Bookcases per floor
DAMAGES STATES, FRAGILITIES, AND CONSEQUENCE FUNCTIONS
DESCRIPTION
ILLUSTRATION
(example photo or drawing)
MEDIAN EDP
(peak horizontal floor acceleration)
BETA
CORRELATION (%)
REPAIR MEASURES
CONSEQUENCE FUNCTION
TIMEFRAME FOR REPAIRS
Figure D13 Fragility for unanchored bookcases
April 15, 2009
ATC58 E: Ground Shaking Hazards E1
Appendix E
Ground Shaking Hazards
E.1 Scope
This appendix presents supplemental information on the characterization of
ground shaking hazards and an alternate procedure for scaling ground
motions for scenariobased assessments.
E.2 Attenuation Relationships
Attenuation relationships relate ground motion parameters to the magnitude
of an earthquake and the distance away from the fault rupture. Relationships
have been established for many ground motion parameters including
 Peak horizontal ground acceleration, velocity, displacement and
corresponding spectral terms
 Peak vertical ground acceleration, velocity, displacement and
corresponding spectral terms
Attenuation relationships are developed by statistical evaluation of large sets
of ground motion data. Relationships have been developed for different
regions of the United States (and other countries) and different fault types
(i.e., strikeslip, dipslip and subduction). These relationships are only as
good as the dataset from which the relationships were derived; the greater the
size of the data set, the more robust the relationship.
The basic construction of a ground motion attenuation relationship is
presented in Chapter 5 and is not repeated here. Attenuation relationships
generally return a geometric mean
1
of two horizontal spectral ordinates.
Selected North American ground motion attenuation relationships are
presented in Table E1. These attenuation relationships use moment
magnitude
W
M to define earthquake magnitude. The attenuation
1
For a given pair of horizontal earthquake histories, the geometric mean
(
g
S of the spectral ordinates of the two components (
x
S and
y
S ) is generally
used to characterize the pair of histories:
g x y
S S S = . Since the functional
form of the attenuation relationship involves the natural log of the ground
motion parameter, the geometric mean of the ordinates (which is equivalent
to the arithmetic mean of the logs of the ordinates) is used instead of the
arithmetic mean.
April 15, 2009
E2 E: Ground Shaking Hazards ATC58
relationships of Table E1 use different definitions of sitetosource distance;
some of the definitions are illustrated in Figure E1 that is adapted from
Table E1 Ground Motion Attenuation Relationships
Model
Calculated
1
Site Conditions
2
Variables
3
Ranges
n
T
(secs)
r (km)
W
M
4
Western North America (WNA)
Abrahamson and Silva,
2008
PHA, PHV,
Sah
30
180 2000
S
V s s
5
M,
rup
r ,
jb
r ,
30 S
V ,
F, H, B
010 0200
58.5
Boore and Atkinson,
2008
PHA, PHV,
Sah
30
180 1300
S
V s s
M,
jb
r ,
30 S
V , F
58
Campbell and
Bozorgnia, 2008
PHA, PHV,
Sah
30
150 1500
S
V s s
M,
rup
r ,
jb
r ,
30 S
V ,
F, H, B
48.5 (s)
48 (r)
47.5 (n)
Chiou and Youngs,
2008
PHA, PHV,
Sah
30
150 1500
S
V s s
M,
rup
r ,
jb
r ,
30 S
V ,
F, H, B
48.5 (s)
48 (r&n)
Idriss, 2008 PHA, Sah
30
900
S
V > ,
30
450 900
S
V s s
M,
rup
r , F 4.58
5
Central and Eastern North America (CENA)
Atkinson & Boore,
1997
PHA, Sah Rock M,
hypo
r 02 10300 49.5
Campbell, 2003 PHA, Sah Hard rock M,
rup
r 04 11000 58.2
Toro et al., 1997 PHA, Sah Rock M,
jb
r 02 1100 58
Subduction Zones
Atkinson & Boore,
2003
PHA, Sah Rock to poor soil M,
hypo
r , + 03 10300 5.58.3
Youngs et al., 1997 PHA, Sah Rock, Soil
M,
rup
r , F,
H
04 0100 49.5
1. PHA = peak horizontal ground acceleration, PHV = peak horizontal ground velocity, Sah = horizontal spectral
acceleration
2.
30 S
V = Average shear wave velocity in the upper 30 meters of sediments, unit: m/s
3.
rup
r = closest distance to the rupture surface,
jb
r = closest horizontal distance to the vertical projection of the
rupture,
hypo
r = hypocentral distance, M = magnitude, F = fault type, H = hanging wall, B = 3D basin
response.
4. s = strikeslip faulting, r = reverse faulting, n = normal faulting.
5. Ranges are not specified (Abrahamson and Silva 2008; Idriss 2008) but are estimated based on the datasets.
April 15, 2009
ATC58 E: Ground Shaking Hazards E3
Abrahamson and Shedlock (1997). The seismogenic depth is defined here as
the depth of the surface materials.
Figure E1 Sitetosource distance definitions (Abrahamson and Shedlock,
1997)
E.3 Fault Rupture Directivity and Maximum Direction
Shaking
Rupture directivity causes spatial variations in the amplitude and duration of
ground motions around faults. Propagation of rupture towards a site produces
larger amplitudes of shaking at periods longer than 0.6 second and shorter
strongmotion durations than for average directivity conditions.
Somerville et al. (1997) developed modifications to the empirical attenuation
relations of Abrahamson and Silva (1997) to account for these variations.
This and other attenuation functions of that time were constructed from
strong motion databases that included few nearfault records and so could not
adequately represent the spectral demands associated with nearfault shaking.
Somerville (1997) identified fault rupture directivity parameters u and X for
strikeslip faults and  and Y for dip slip faults as shown in Figure E2.
Somerville developed three ground motion parameters to characterize
directivity: (1) Amplitude factor: bias in average horizontal response
spectrum acceleration with respect to Abrahamson and Silva (1997); (2)
Duration factor: bias in duration of acceleration with respect to Abrahamson
and Silva (1997); and (3) Faultnormal/Average amplitude: ratio of fault
April 15, 2009
E4 E: Ground Shaking Hazards ATC58
normal to average (directivity) horizontal response spectrum acceleration.
Bounds were set on the range of applicability of the directivity model.
Figure E2 Fault rupture directivity parameters (Somerville et al., 1997)
The recently developed Next Generation Attenuation (NGA) relationships
for the Western United States were developed using a database that included
many nearfault ground motion recordings. These relationships predict
geomean spectral demand of recorded nearfault motions (i.e., average
directivity) well with little bias (Huang et al. 2008a, 2008b). Huang et al. also
report the median and 84th percentile ratios of maximum and minimum
direction spectral demand to NGAcalculated geomean spectral demand for
average directivity (i.e., all nearfault records) and the forward directivity
region.
E.4 Probabilistic Seismic Hazard Assessment
E.4.1 Introduction
Performancebased seismic design will often utilize a sitespecific
characterization of the ground shaking associated with different probabilities
of exceedance or return periods. Such characterizations are routinely
performed using Probabilistic Seismic Hazard Assessment (PSHA). The
following subsections provide introductory information on PSHA, and draw
April 15, 2009
ATC58 E: Ground Shaking Hazards E5
substantially from Kramer (1996), McGuire (2004) and Bozorgnia and
Bertero (2004).
E.4.2 PSHA Calculations
General
For Probabilistic Seismic Hazard Assessment (PSHA), probability
distributions are determined for the magnitude of each earthquake on each
source, ( )
M
f m , the location of the earthquake in or along each source,
( )
R
f r , and the prediction of the response parameter of interest
P(pga pga , ) m r ' > . Kramer describes PHSA as a fourstep process that is
enumerated below and depicted in Figure E3.
1. Identify and characterize (geometry and potential
W
M ) all earthquake
sources capable of generating significant shaking (say 4.5
W
M > ) at the
site. Develop the probability distribution of rupture locations within each
source. Combine this distribution with the source geometry to obtain the
probability distribution of sourcetosite distance for each source.
2. Develop a seismicity or temporal distribution of earthquake occurrence
for each source using a recurrence relationship.
3. Using predictive (attenuation) relationships, determine the ground
motion produced at the building site (including the uncertainty) by
earthquakes of any possible size or magnitude occurring at any possible
point in each source zone.
4. Combine the uncertainties in earthquake location, size, and ground
motion prediction to obtain the probability that the chosen ground motion
parameter (e.g., peak horizontal ground acceleration, spectral
acceleration at a specified frequency) will be exceeded in a particular
time period (say 10% in 50 years).
April 15, 2009
E6 E: Ground Shaking Hazards ATC58
Figure E3 Steps in probabilistic seismic hazard assessment (Kramer, 1996)
Summary information on parts of each step in the process described above is
presented below. The reader is referred to Kramer (1996) and McGuire
(2004) for much additional information.
Earthquake Source Characterization
The characterization of an earthquake source (and there might be a number
of sources for a given site) requires consideration of the spatial
characteristics of the source, the distribution of earthquakes within that
source, of the distribution of earthquake size within that source, and of the
distribution of earthquakes with time. Each of these characteristics involves
some degree of uncertainty and such uncertainty is addressed explicitly by
PSHA.
Spatial Uncertainty
The geometries of earthquake sources are typically characterized as point
sources (e.g., volcanoes), twodimensional areal sources (e.g., a welldefined
fault plane) and threedimensional volumetric sources (e.g., areas where
earthquake mechanisms are poorly defined such as the Central and Eastern
USA). Source zones might be similar to or different from the actual source,
depending on the relative geometry of the source and the site of interest, as
shown in Figure E4 below from Kramer (1996).
April 15, 2009
ATC58 E: Ground Shaking Hazards E7
Figure E4 Source zone geometries (Kramer, 1996)
Since attenuation relationships express ground motion parameters in terms of
a measure of the sourcetosite distance, the spatial uncertainty must be
described with respect to the appropriate distance parameter. The uncertainty
in sourcetosite distance can be described by a probability density function
as shown in Figure E5.
Figure E5 Variations in sitetosource distance for three source zone
geometries (Kramer, 1996)
For the point source above, the distance R is
s
r and the probability that
s
R r = is 1.0 and
s
R r = is 0. For more complex source zones, it is easier to
evaluate ( )
R
f r by numerical integration. For example, the source zone of
part c. of the figure above is broken up into a large number of discrete
elements of the same area. A histogram that approximates ( )
R
f r can be
constructed by tabulating the values of R that correspond to the center of
each element.
Size Uncertainty
The distribution of earthquake sizes in a given period is described by a
recurrence law. One basic assumption of PSHA is that the recurrence law
obtained on the basis of past seismicity is appropriate for the prediction of
future seismicity. The best known recurrence law is that of Gutenberg and
April 15, 2009
E8 E: Ground Shaking Hazards ATC58
Richter (1944), who collected data from Southern Californian earthquakes
over a period of years and plotted the data according to the number of
earthquakes that exceeded different magnitudes during that period. The
number of exceedances of each magnitude was divided by the length of the
time period used to assemble the data to define a mean annual rate of
exceedance
m
ì of an earthquake of magnitude m. The reciprocal of the mean
annual rate of exceedance of a particular magnitude is termed the return
period of earthquakes exceeding that magnitude. GuttenbergRichter plotted
the logarithm of the annual rate of exceedance (of earthquakes in Southern
California) against earthquake magnitude and the resulting relationship was
linear, namely,
log
m
a bm ì = ÷ (E1)
where
m
ì is defined above, 10
a
is the mean yearly number of earthquakes of
magnitude greater than or equal to 0, and b describes the relative likelihood
of large and small earthquakes. As the value of b increases, the number of
larger magnitude earthquakes relative to smaller magnitude earthquakes
decreases. The values of a and b are generally obtained by regression
analysis on a database of seismicity from the source zone of interest. The
mean rate of small earthquakes is often underpredicted because historical
records are often used to supplement the instrumental records and only the
larger magnitude events from part of the historical record.
The GuttenbergRichter recurrence law can also be expressed as
10 exp(2.303 2.303 )
a bm
m
a b ì
÷
= = ÷ (E2)
which shows that the law implies that earthquake magnitudes are
exponentially distributed and that the range of magnitude is from ÷· to · .
Small magnitude earthquakes are of little significance to the built
environment and can be ignored in terms of hazard assessment. The law also
predicts nonzero mean rates of exceedance from magnitudes up to · , which
is not physically possible. Bounded (lower and upper) recurrence laws have
been proposed to deal with these practical bounds on magnitude,
The GuttenbergRichter law was originally developed from regional data and
not for specific source zones. Paleoseismic studies over the past 30 years
have indicated that individual points on faults (or fault segments) tend to
move by approximately the same distance in each earthquakes, suggesting
that individual faults repeatedly generate earthquakes of a similar (with 0.5
magnitude unit) size, known as characteristic earthquakes at or near their
maximum magnitude. Geological evidence indicates that characteristic
earthquakes occur more frequently than would be implied by extrapolation of
April 15, 2009
ATC58 E: Ground Shaking Hazards E9
the law from high exceedance rates (of low magnitude events) to low
exceedance rates (of high magnitude), resulting in a more complex
recurrence law than that given by equation E2.
Attenuation Relationships
Predictive (or attenuation) relationships are generally obtained empirically by
leastsquares regression on a strongmotion dataset. Scatter or randomness in
the results is inevitable because of differences in site conditions, travel path
and fault rupture mechanics. The scatter can be characterized by confidence
limits or by the standard deviation of the predicted parameter.
The probability that a ground motion parameter Y exceeds a certain value y
for an earthquake of magnitude m, occurring at a distance r is given by
[ , ] 1 ( )
y
P Y y m r F y > = ÷ (E3)
where ( )
Y
F y is the value of the CDF of Y at m and r. The value of ( ) F y
Y
depends on the probability distribution used to describe Y. As noted
previously, ground motion parameters are generally assumed to be
lognormally distributed. Figure E6 illustrates the conditional probability of
exceeding a particular value of the ground motion parameter for a given
combination of m and r.
Figure E6 Conditional probability calculation (Kramer, 1996)
Temporal Uncertainty
The distribution of earthquake occurrence with time must be computed or
assumed to calculate the probabilities of different earthquake magnitudes
occurring in a given time period. Earthquakes are assumed to occur randomly
with time and the assumption of random occurrence permits the use of
simple probability models.
The temporal occurrence of earthquakes is commonly described as a Poisson
process: one that yields values of a random variable describing the number of
occurrences of a particular event during a given time interval (or spatial
region). In a Poisson process: (a) the number of occurrences in one time
April 15, 2009
E10 E: Ground Shaking Hazards ATC58
interval are independent of the number that occur in any other time interval;
(b) the probability of occurrence during a very short time interval is
proportional to the length of the time interval; and (c) the probability of more
than one occurrence in a very short time interval is negligible. Events in a
Poisson process occur randomly, with no memory of the time, size or
location of any preceding events. Cornell showed that the Poisson model is
useful for PSHA unless the hazard is dominated by a single source zone that
produces characteristic earthquakes and the time period since the previous
significant event exceeds the average interevent time.
For a Poisson process, the probability of a random variable N, representing
the number of occurrences of a particular event in a given time period is
given by
[ ]
!
n
e
P N n
n
µ
µ
÷
= = (E4)
where µ is the average number of occurrences of the event in the time
period. To characterize the temporal distribution of earthquake recurrence for
PSHA, the Poisson probability is normally expressed as
( )
[ ]
!
n t
t e
P N n
n
ì
ì
÷
= = (E5)
where ì is the average rate of recurrence of the event and t is the time
period. When the event is the exceedance of a particular earthquake
magnitude, the Poisson model can be combined with a suitable recurrence
law to predict the probability of at least one exceedance in a period of t years
by
[ 1] 1
m
t
P N e
ì ÷
> = ÷ (E6)
Probability Computations and Seismic Hazard Curves
The development of seismic hazard curves, which identify the annual
probability of exceedance of different values of a selected ground motion
parameter, involves probabilistic calculations that combine the uncertainties
in earthquake size, location and frequency for each potential earthquake
source that could impact shaking at the site under study. The seismic hazard
curves can then be used to compute the probability of exceeding the chosen
ground motion parameter in a specified period of time.
The seismic hazard curve calculations are (somewhat) straightforward once
the uncertainties in earthquake size, location and frequency are established
but much bookkeeping is involved. The probability of exceeding a particular
value y of a ground motion parameter Y is calculated for one possible source
April 15, 2009
ATC58 E: Ground Shaking Hazards E11
location and then multiplied by the probability of that magnitude earthquake
occurring at that particular location. The calculation is then repeated for all
possible magnitudes and locations and the probabilities of each are summed
to compute the [ ] P Y y > at the site. A summary of the presentation of
Kramer (1996) is presented below. The reader is referred to Kramer (1996)
and McGuire (2004) for much additional information.
For a given earthquake occurrence, the probability that a ground motion
parameter Y will exceed a particular value y can be computed using the total
probability theorem (Cornell and Benjamin, 1968), namely,
[ ] [ ] [ ] [ ] ( )
x
P Y y P Y y P P Y y f dx > = > = >
}
X X X X (E7)
where X is a vector of random variables that influence Y. In most cases, the
quantities in X are limited to the magnitude M and distance R. Assuming
that M and R are independent, the probability of exceedance can be written as
[ ] [ , ] ( ) ( )
M R
P Y y P Y y m r f m f r dmdr > = >
}}
(E8)
where [ , ] P Y y m r > is obtained from the predictive relationship and ( )
M
f m
and ( )
R
f r are the probability density functions for magnitude and distance,
respectively.
If the building site is in a region of
s
N potential earthquake sources, each of
which has an average rate of threshold exceedance exp( )
i i i
m v o  = ÷ , the
total average exceedance rate for the region is given by the equation below,
which is typically solved by numerical integration.
1
[ , ] ( ) ( )
s
N
y Mi Ri
i
P Y y m r f m f r dmdr ì
=
= >
¿
(E9)
One approach that is described by Kramer as simple rather than efficient is to
divide the possible ranges of magnitude and distance into
M
N and
R
N
segments, respectively. The average exceedance rate can then be calculated
using a multilevel summation as follows:
1 1 1
1 1 1
[ , ] ( ) ( )
[ , ] [ ] [ ]
s m R
s m R
N N N
y i j k Mi j Ri k
i j k
N N N
i j k j k
i j k
P Y y m r f m f r m r
P Y y m r P M m P R r
ì v
v
= = =
= = =
= > A A
= > = =
¿¿¿
¿¿¿
(E10)
where the terms are
0 max 0
( 0.5)( )/
j M
m m j m m N = + ÷ ÷ ,
max min
( )/
r
r r r N A = ÷ ,
min max min
( 0.5)( )/
k R
r r k r r N = + ÷ ÷ and
max 0
( )/
M
m m m N A = ÷ . The above statement is equivalent to assuming that
each source is capable of generating only
M
N different earthquakes of
April 15, 2009
E12 E: Ground Shaking Hazards ATC58
magnitude
j
m at only
R
N different sourcetosite distances of
k
r . The
accuracy of the above method increases with smaller segments and thus
larger values of
M
N and
R
N .
Figure E7 presents a sample seismic hazard curve for peak ground
acceleration at a site in Berkeley, California (McGuire, 2004). The
contributions to the annual frequency of exceedance from 9 seismic sources
are shown.
Figure E7 Seismic hazard curve for Berkeley, California (McGuire, 2004)
Probabilities of exceedance in a selected time period can be computed using
seismic hazard curves combined with the Poisson model. The probability of
exceedance of y in a time period T is
[ ] 1
y
T
T
P Y y e
ì ÷
> = ÷ (E11)
As an example, the probability that a peak horizontal ground acceleration of
0.10 g will be exceeded in a 50year time period for the site characterized by
the hazard curve above:
(0.060)(50)
[PHA 0.1 in50 years] 1 1 0.950 95%
y
T
P g e e
ì ÷
÷
> = = ÷ = = (E12)
An alternate, often made, computation is the value of the ground motion
parameter corresponding to a particular probability of exceedance in a given
time period. For example, the acceleration that has a 10% probability of
exceedance in a 50year period would be that with an annual rate of
exceedance, calculated by rearranging the secondtolast equation, namely
In(1 [ ]) In(1 0.10)
0.00211
50
T
y
P Y y
T
ì
÷ > ÷
= ÷ = ÷ = (E13)
April 15, 2009
ATC58 E: Ground Shaking Hazards E13
Using the hazard curve from Figure B13 for all nine sources, the
corresponding zeroperiod spectral acceleration is approximately 0.75 g.
Hazard curves can be developed for specific oscillator periods (0.2 second
and 1.0 second are widely used) and the above calculations of probability of
exceedance can be extended to the spectral domain, namely, to develop
probabilistic estimates of spectral demands for different probabilities of
exceedance or return periods. Hazard curves have been developed for many
different response quantities and characterizations of earthquake ground
motion.
E.4.3 Inclusion of Rupture Directivity Effects
Rupture directivity effects should be considered in PSHA if the building site
is located within 20 km of an active fault capable of generating a moment
magnitude 6.5 earthquake or greater.
Average rupture directivity effects can be included directly in PSHA if the
attenuation relationships are either a) constructed using a database of
recorded ground motions that include directivity effects (i.e., the NGA
relationships described previously), or b) corrected to account for rupture
directivity in a manner similar to that described by Somerville et al. (1997)
and Abrahamson (2000). Note that the ratio of median maximum demand to
geomean demand in the nearfault region is perioddependant but might be as
great as 2 (Huang et al. 2008a, 2008b).
E.4.4 Deaggregation of Seismic Hazard Curves and Epsilon
The probabilistic seismic hazard analysis procedures described previously
enable the calculation of annual rates of exceedance of ground motion
parameters (e.g., spectral acceleration at a selected period) at a particular site
based on aggregating the risks from all possible source zones. The rate of
exceedance computed in the probabilistic seismic hazard analysis is therefore
not associated with any particular earthquake magnitude m or distance r.
For a given building site and hazard curve, the combinations of magnitude,
distance and source that contribute most to the hazard curve can be
established. This process is termed deaggregation (or disaggregation).
Hazard deaggregation requires that the mean annual rate of exceedance be
expressed as a function of magnitude and distance as follows:
1
( , ) [ ] [ ] [ , ]
s
N
y j k j k i j k
i
m r P M m P R r P Y y m r ì v
=
= = = >
¿
(E14)
April 15, 2009
E14 E: Ground Shaking Hazards ATC58
Detailed information on the deaggregation of seismic hazard curves is
provided in McGuire (2004).
Sample deaggregation results for horizontal spectral acceleration at periods
of 0.2 second and 1.0 second for a site in San Francisco, Latitude =37.8º,
Longitude =122.4 º , for a 2% probability of exceedance in 50 years, are
shown in Figure E8 below. This site is identical to that used in Chapter 5 to
illustrate the use of the USGS ground motion calculator. To generate similar
results, visit the USGS website below and click on the Interactive
deaggregation button under the headings of 2008 and Hazard Values, and
enter the Lat and Long for the building site and the type of deaggregation
required.
Consider the 1second deaggregation data of part b of Figure E8. The plot
shows the contribution to the 1second uniform hazard spectral ordinate for a
2% probability of exceedance in 50 years as a function of moment magnitude
and distance. The figure shows that approximately 50% of the total 1second
seismic demand can be ascribed to a moment magnitude 7.8 earthquake at a
distance of 14 km–the [M, r] pair that dominates the 1second spectral
demand at this site is [7.8, 14].
Importantly, this figure introduces another important ground motion variable,
epsilon, c . One straightforward definition of c is:
a
S u
c

÷
= (E15)
where all variables vary as a function of period;
a
S is the spectral
acceleration computed by PSHA for a given probability (e.g., 2%) of
exceedance in a specified time period (e.g., 50 years) and equal to 0.829g in
this instance at a period of 1 second; u is the median value of spectral
acceleration computed by an appropriate attenuation relationship(s) for the
dominant [M, r] pair (equal to [7.8, 14] in this instance), and  is the
dispersion in the attenuation relationship. In this example, and using the
modal [M, r, c ] triple, we see that c ranges between 1 and 2, meaning that
less than 15% of moment magnitude 7.8 earthquakes at a distance of 14 km
would produce geomean spectral demands in excess of 0.829 g.
Baker and Cornell (2005, 2006) observed that c is an indicator of spectral
shape, with a positive (negative) value of c at a given period tending to
indicate a relative peak (valley) in the acceleration response spectrum at that
period. Figure E9 illustrates these trends using six bins of 20 ground
motions defined by 2.25 c > , 0.06 0.06 c ÷ s s and 2.25 c s ÷ , respectively,
at periods of 0.8 (Figure E9a) and 0.3 second (Figure E9b). The relative
April 15, 2009
ATC58 E: Ground Shaking Hazards E15
a. 0.2 second deaggregation
b. 1.0 second deaggregation
Figure E8 Sample deaggregation of a hazard curve (from www.usgs.gov)
April 15, 2009
E16 E: Ground Shaking Hazards ATC58
peak (valley) in the spectra for the c + ( c ÷ ) bin of motions at the two
periods is clearly evident. No relative peak or valley at the reference periods
is associated with the c neutral bin.
E.4.5 Conditional Mean Spectrum and Spectral Shape
Chapter 5 introduced Probabilistic Seismic Hazard Assessment (PSHA) and
the development of a Uniform Hazard Spectrum (UHS) and noted that the
ordinates in a UHS are each associated with the same annual frequency of
exceedance.
Codes of practice and seismic design standards recommend scaling ground
motions to match a UHS over a wide period range. Although convenient for
design and likely conservative, such scaling procedures ignore the fact that
1. If the spectral ordinates of a UHS are governed by multiple scenario
events, the spectral shape of the UHS will not represent the spectral
shape for any of the governing events, regardless of the return period of
the UHS (Cornell 2006, Baker and Cornell 2006).
2. For longreturnperiod earthquake shaking, the spectral ordinates of a
UHS are associated typically with a high value (say greater than 1) of c
across a wide range of period (Harmsen 2001). For the case where the
geomean spectral ordinate of a ground motion pair attains the UHS
ordinate at a given period, the geomean spectrum of the pair is unlikely
to have ordinates as large as those of the UHS at other periods. The
geomean spectra for the c + record sets in Figure E9 support this
observation.
Figure E9 Sample geometricmean response spectra for negative, zero and positive c
record sets with each record in the sets scaled to a) (0.8 )
a
S s = 0.5 g and b)
(0.3 )
a
S s = 0.5 g (Baker and Cornell 2006)
April 15, 2009
ATC58 E: Ground Shaking Hazards E17
Baker and Cornell (2005, 2006) introduced the Conditional Mean Spectrum
(CMS), which considers the correlation of spectral demands (represented by
values of c ) at different periods, to address the two issues identified above.
A CMS estimates the median geomean spectral acceleration response of a
pair of ground motions given an [M, r] pair and a target spectral ordinate,
1
( )
a
S T (from which
1
T
c is backcalculated using an appropriate attenuation
relationship). A CMS is generated using the following equation (Baker and
Cornell 2005):
1 2 1 1
2 2 2 ( ), ( ) ( )
( ) ( , , ) exp ( , , )
a
T T T S T
CMS T M r T M r T
c c
u  µ c ( = · · ·
¸ ¸
(E15)
where
1
2 ( )
( )
a
S T
CMS T is the ordinate of the CMS at period
2
T given that the
spectral demand at
1
T is
1
( )
a
S T ;
2
( , , ) M r T u and
2
( , , ) M r T  are the median
and logarithmic standard deviation of spectral acceleration at
2
T computed
using a ground motion attenuation relationship for the [M, r] pair of interest;
1
T
c is the value of c associated with
1
( )
a
S T ; and
1 2
( ), ( ) T T c c
µ is the correlation
coefficient for c between
1
T and
2
T .
Baker and J ayaram (2008) developed a model for
1 2
( ), ( ) T T c c
µ using the NGA
ground motion database (www.peer.berkeley.edu) :
1 2
1 2
1 2
1 2
max (T ), (T ) 2
min (T ), (T ) 1
max (T ), (T ) 2 4
(T ), (T ) 4
if <0.109
else if >0.109
else if <0.109 min( , )
else
T C
T C
T C C
C
c c
c c
c c
c c
µ
µ
µ
µ
=
=
=
=
(E16)
where
min 1 2
min( , ) T T T = ,
max 1 2
max( , ) T T T = and
max
1
min
1 cos 0.366ln
2 max( ,0.11)
T
C
T
t
 
 
= ÷ ÷
 

\ . \ .
(E17)
max
max min
max 100 5
2 max
1
1 0.105 1 if 0.2sec
0.01 1
0 otherwise
T
T T
T
C T e
÷
¦
  ÷  
÷ ÷ <
¦ 

= ÷ +
\ . ´
\ .
¦
¹
(E18)
2 max
3
1 max
if 0.11sec
if 0.11sec
C T
C
C T
< ¦
=
´
>
¹
(E19)
min
4 1 3 3
0.5( )(1 cos
0.11
T
C C C C
t  
= + ÷ +

\ .
(E20)
April 15, 2009
E18 E: Ground Shaking Hazards ATC58
A CMS can be constructed as follows:
1. Determine the value of
1
( )
a
S T from the UHS at the desired mean annual
frequency of exceedance.
2. Deaggregate the hazard at
1
( )
a
S T and identify the modal [M, r, c ]
triple.
3. Select an appropriate ground motion attenuation relationship.
4. Generate the CMS using (E15) through (E20) and the attenuation
relationship of Step 3 for the [M, r, c ] triple of Step 2.
5. Amplitude scale the CMS of Step 4 to recover
1
( )
a
S T .
To illustrate the procedure, consider the deaggregation results of Figure B8b
for a site in San Francisco. As presented in Figure E8b, the value of
1
( )
a
S T
is 0.83 g for a return period of 2475 years and the modal [M, r, c ] triple for
the hazard is [7.8, 14 km, 1 to 2]. A value of 1.5 c = (midpoint of 1 and 2)
was used for this example. The attenuation relationship selected for Step 3
was the ChiouYoungs NGA relationship (Chiou and Youngs 2008). The
results of Steps 4 and 5 are presented in Figure E10 using dotted and solid
lines, respectively. (If 2 c = is used in Step 4, the scale factor for Step 5 is 1.)
Figure E10 also presents the UHS for the site for a return period of 2475
years. The spectral ordinates of the UHS are greater than those of the scaled
CMS except at
1
T .
0 0.5 1 1.5 2
Period (sec)
0
0.5
1
1.5
2
S
p
e
c
t
r
a
l
a
c
c
e
l
e
r
a
t
i
o
n
(
g
)
Original CMS
Scaled CMS
UHS
Figure E10 UHS for a 2% probability of exceedance in 50 years and
original and scaled CMS for a rock site in San Francisco.
April 15, 2009
ATC58 E: Ground Shaking Hazards E19
E.5 Selection of Earthquake Ground Motions
E.5.1 Introduction
Procedures for selecting and scaling pairs of ground motions for response
history analysis are presented in Chapters 5 and 7 of the Guidelines and are
not repeated here. Additional information related to c , the Conditional Mean
Spectrum, and spectral shapes are presented in Section E.4.
E.5.2 Earthquake Histories for Response Analysis
Geomean spectral shape is the key variable to be considered in the selection
of seed pairs of ground motions that will be amplitude scaled for response
history analysis.
E.6 Vertical Earthquake Shaking
E.6.1 Introduction
For most buildings, where important vertical periods of response are less
than or equal to 1.0 second, vertical earthquake spectra can be derived from
the horizontal spectra using the procedures of this section. For buildings with
important natural periods of vertical response in excess of 1.0 second, a site
specific analysis should be conducted.
Sections E.6.2 and E.6.3 present procedures for Site Classes A, B and C, and
Site Classes D and E, per ASCE705, respectively.
E.6.2 Procedure for Site Classes A, B and C
For structural periods less than or equal to 1.0 second, the vertical response
spectrum can be constructed by scaling the corresponding ordinates of the
horizontal response spectrum,
a
S , as follows.
For periods less than or equal to 0.1 second, the vertical spectral acceleration,
v
S , can be taken as
v a
S S = (E21)
For periods between 0.1 and 0.3 second, the vertical spectral acceleration,
v
S , can be taken as
(1 1.048[log( ) 1])
v a
S T S = ÷ + (E22)
For periods between 0.3 and 1.0 second, the vertical spectral acceleration,
v
S , can be taken as
0.5
v a
S S = (E23)
April 15, 2009
E20 E: Ground Shaking Hazards ATC58
E.6.3 Procedure for Site Classes D and E
For structural periods less than or equal to 1.0 second, the vertical response
spectrum can be constructed by scaling the corresponding ordinates of the
horizontal response spectrum,
a
S , as follows.
For periods less than or equal to 0.1 second, the vertical spectral acceleration,
v
S , can be taken as
v a
S S q = (E24)
For periods between 0.1 and 0.3 second, the vertical spectral acceleration,
v
S , can be taken as
( 2.1( 0.5)[log( ) 1])
v a
S T S q q = ÷ ÷ + (E25)
For periods between 0.3 and 1.0 second, the vertical spectral acceleration,
v
S , can be taken as
0.5
v a
S S = (E26)
In equations (E24) through (E26), q can be taken as 1.0 for 0.5
s
S g s ; 1.5
for 1.5
s
S g s ; and (1 0.5( 0.5))
s
S + ÷ for 0.5 1.5
s
g S g s s .
E.7 SoilFoundationStructure Interaction
E.7.1 General
The response of a structure to earthquake shaking is affected by the response
of the dynamic properties of the three components of the soilfoundation
structure system and the interactions between the three components. A soil
foundationstructure interaction (SFSI) analysis involves the direct or
indirect analysis of the threecomponent system to prescribed freefield
inputs, which are typically imposed as bedrock motions.
In traditional structural analysis, the base of the structure is assumed to be
rigid (components fixed at their base) and the freefield motions at the base
of the foundation are derived assuming one of the soil categories described
previously. SFSI effects are absent from such an analysis. Figure E11a
illustrates the traditional structural model. The fixed base model is
inappropriate for structural framing systems that incorporate stiff vertical
components for lateral resistance (e.g., reinforced concrete shear walls, steel
braced frames) because the response of such systems can be sensitive to
small base rotations and translations that are neglected with a fixed base.
Relatively flexible lateral framing systems such as steel momentresisting
frames are often not significantly affected by SFSI.
April 15, 2009
ATC58 E: Ground Shaking Hazards E21
E.7.2 Direct SoilFoundationStructureInteraction Analysis
SFSI can be analyzed directly using the finiteelement method and response
history solutions, where the soil is discretized with solid finite elements. Full
nonlinear analysis is also possible using this method although typical
analysis is performed using equivalent linear soil properties. Direct analysis
can address the three key effects of SFSI, namely, foundation flexibility,
kinematic effects, and foundation damping, although solution of the
kinematic interaction problem is often difficult without customized finite
element codes because typical finite element codes cannot account for wave
inclination and wave incoherence effects.
u
g
= free field motion (FFM) with
conventional damping
Infinitely rigid foundation and soil
u
g
= free field motion (FFM) with
conventional damping
structural components of
foundation
geotechnical components
of foundation
a) Rigid base model b) Flexible base model
u
g
= foundation input motion (FIM)
with conventional damping
Kinematic interaction
(high Tpass filter)
free field motion (FFM) with
conventional damping
u
g
= foundation input motion (FIM)
with system damping including
foundation damping
Adjust for foundation
damping
foundation input motion (FIM) with
conventional damping
Kinematic interaction
(high Tpass filter)
free field motion (FFM) with
conventional damping
c) Kinematic interaction d) Foundation damping
u
g
= free field motion (FFM) with
conventional damping
Infinitely rigid foundation and soil
u
g
= free field motion (FFM) with
conventional damping
structural components of
foundation
geotechnical components
of foundation
a) Rigid base model b) Flexible base model
u
g
= foundation input motion (FIM)
with conventional damping
Kinematic interaction
(high Tpass filter)
free field motion (FFM) with
conventional damping
u
g
= foundation input motion (FIM)
with conventional damping
Kinematic interaction
(high Tpass filter)
Kinematic interaction
(high Tpass filter)
free field motion (FFM) with
conventional damping
u
g
= foundation input motion (FIM)
with system damping including
foundation damping
Adjust for foundation
damping
Adjust for foundation
damping
foundation input motion (FIM) with
conventional damping
Kinematic interaction
(high Tpass filter)
Kinematic interaction
(high Tpass filter)
free field motion (FFM) with
conventional damping
c) Kinematic interaction d) Foundation damping
Figure E11 Analysis for soil foundation structure interaction (FEMA, 2005)
Responsehistory analysis will require the development of a three
dimensional numerical model of the soilfoundationstructure system. Stress
strain or constitutive models (linear, nonlinear or equivalent linear) for the
soils in the model should be developed based on test data. Appropriate and
consistent frequencydependant stiffness and damping matrices should be
developed for the edges or boundaries of the soil in the numerical model.
April 15, 2009
E22 E: Ground Shaking Hazards ATC58
E.7.3 Simplified SoilFoundationStructureInteraction Analysis
Simplified Procedures
Simplified procedures for including the effects of SFSI in response analysis
are presented in this section. The procedures are based on those presented in
FEMA 440 (FEMA, 2005). More rigorous procedures are available in
Appendix A of FEMA 440.
For the discussion below, soilfoundationstructure interaction is parsed into
three key effects:
 foundation flexibility
 kinematic effects (filtering of the ground shaking to the building)
 foundation damping (dissipation of energy from the soil structure system
through radiation and hysteretic soil damping).
Figure E11b illustrates the incorporation of foundation flexibility into the
numerical model of a building frame. Current analysis procedures in
guidelines such as ASCE 41 (ASCE, 2006) and ATC 40 (ATC, 1996)
partially address the flexible foundation effect by providing guidance on the
stiffness and strength of the geotechnical (soil) components of the foundation
in the structural analysis model. However, these analysis procedures do not
characterize the reduction of the shaking demand on the structural framing
system relative to the freefield motion due to kinematic interaction or the
foundation damping effect, both of which are described below. Guidance on
including these effects in a simplified manner for nonlinear dynamic
response analysis is provided below. The product of numerical simulation
using the model of Figure E11b is a global response that includes elastic and
inelastic deformations in the structural and geotechnical parts of the
foundation system. These deformations are sometimes referred to as an
inertial soilstructureinteraction effect. The inclusion of foundation
flexibility in the numerical model can lead to both significant changes to the
responses computed assuming a fixed base (Figure E11a) and more accurate
representation of probable structural response. Foundation responses and
failures (e.g., rocking, soil bearing failure, pier/pile slip) can be explicitly
evaluated using a numerical model that explicitly includes foundation
flexibility.
Kinematic soilstructure interaction (Figure E11c) results from the presence
of relatively stiff foundation elements atop or in soil that causes the
foundation motions to deviate from freefield motions. Base slab averaging
and embedment are two kinematic effects. Ground motion shaking is
April 15, 2009
ATC58 E: Ground Shaking Hazards E23
spatially variable (i.e., not each point beneath a building footprint, in the
absence of the foundation, would experience identical shaking at the same
instant in time). Placement of a structure and foundation across these
spatially variable motions produces an averaging effect and the weighted or
overall motion experienced by the building is less than the localized maxima
that would have occurred in the freefield. The embedment effect is
associated with the reduction of ground motion with depth into a soil deposit.
Both base slab averaging
2
and embedment
3
affect the characteristics of the
foundationlevel motion (sometimes called the foundation input motion) in a
manner that is independent of the superstructure (i.e., the portion of the
structure above the foundation) with one exception. The effects are strongly
period dependent, being maximized at small periods. The effects can be
visualized as a filter applied to the highfrequency components of the free
field ground motion. The impact of those effects on superstructure response
will tend to be greatest for shortperiod buildings. A description of a
simplified procedure to reduce the free field motion to a foundation input
motion is presented below. The foundation input motion can be applied to a
fixed base model or combined with a flexible base model. Kinematic effects
tend to be important for buildings with relatively short fundamental periods
(e.g., less than 0.5 second), large plan dimensions and basements embedded
10 feet or more in soil materials.
Figure E11d illustrates foundation damping effects that are another result of
inertial soilstructure interaction. Foundation damping results from the
relative movements of the foundation and freefield soil. It is associated with
radiation of energy away from the foundation and hysteretic damping within
the soil. The result is an effective decrease in the spectral ordinates of ground
motion experienced by the structure. Foundation damping effects tend to be
greatest for stiff structural framing systems (e.g., reinforced concrete shear
walls, steel braced frames) and relatively soft foundation soils (e.g., Site
Classes D and E per ASCE705). Simplified procedures for incorporating
2
Base slab averaging occurs to some extent in virtually all buildings. The
slab averaging effect occurs at the foundation level for mats or spread
footings interconnected by either grade beams or reinforced concrete slabs.
Even if a laterally stiff foundation system is not present, the averaging can
occur at the first elevated level of buildings with rigid diaphragms. The
only case where base slab averaging effects should be neglected is for
buildings without a laterally connected foundation system and with flexible
floor and roof diaphragms.
3
Embedment effects should not be considered for buildings without
basements, even if the footings are embedded. Embedment effects tend to
be significant when the depth of basements is greater than about 10 feet.
April 15, 2009
E24 E: Ground Shaking Hazards ATC58
foundation damping in a numerical model for nonlinear dynamic response
analysis are described below.
Simplified Procedure for Kinematic Interaction
A ratioofresponsespectra (RRS) factor can be used to represent kinematic
interaction effects. An RRS is the ratio of the response spectral ordinates
imposed on the foundation (i.e., the foundation input motion, FIM) to the
freefield spectral ordinates. Two phenomena should be considered in
evaluating the RRS: baseslab averaging and foundation embedment.
Foundation embedment effects should be considered for buildings with
basements. The simplified procedure of Chapter 8 of FEMA 440 (FEMA,
2005) can be used for assessment of kinematic interaction. The following
information is needed for such an assessment: dominant building period (as
measured by % mass participating in the elastic base shear computation),
building foundation plan dimension, foundation embedment, peak horizontal
ground acceleration, and shear wave velocity. Limitations on the use of the
simplified procedure are presented in Section 8 of FEMA 440. Figures E12a
and E12b (from FEMA 440) provide summary information on the
reductions in spectral demand due to baseslab averaging and embedment.
The 5% damped freefield spectrum is multiplied by the product of the
perioddependant reduction factors for baseslab averaging and embedment
to derive the foundation input motion spectrum. Since the embedment
computation is dependant on the peak horizontal ground acceleration, the
computation must be repeated for every level of hazard considered for the
performance and loss assessment.
0 0.2 0.4 0.6 0.8 1 1.2
Period (s)
0.4
0.5
0.6
0.7
0.8
0.9
1
F
o
u
n
d
a
t
i
o
n
/
f
r
e
e

f
i
e
l
d
R
R
S
f
r
o
m
b
a
s
e
s
l
a
b
a
v
e
r
a
g
i
n
g
(
R
R
S
b
s
a
)
Simplified Model
b
e
=65 ft
b
e
=130 ft
b
e
=200 ft
b
e
=330 ft
0 0.4 0.8 1.2 1.6 2
Period (s)
0
0.2
0.4
0.6
0.8
1
1.2
F
o
u
n
d
a
t
i
o
n
/
F
r
e
e

F
i
e
l
d
R
R
S
e =30 ft
V
s
=2500 ft/s
V
s
=1200 ft/s
V
s
=600 ft/s
a. base slab averaging b. embedment
Figure E12 Reductions in spectral demand due to kinematic interaction
April 15, 2009
ATC58 E: Ground Shaking Hazards E25
Simplified Procedure for Foundation Damping
Damping related to soilfoundation interaction due to hysteretic soil damping
and radiation damping can significantly increase the effective damping in a
structural frame.
Herein, the effects of foundation radiation damping serve to reduce the
ordinates of a 5% damped acceleration response spectrum, which is generally
used as the basis for generating earthquake histories for responsehistory
analysis. The simplified procedure of FEMA 440 can be used for assessment
of foundation damping. Hysteretic damping in the soil should be captured
directly through the use of nonlinear soil springs. The following information
is needed for such an assessment: dominant (first mode) building period (as
measured by % mass participating in the elastic base shear computation),
1
T ;
first modal mass; building foundation plan dimension; building height;
foundation embedment; straindegraded soil shear modulus and soil
Poisson’s ratio. Limitations on the use of the simplified procedure are
presented in Section 8 of FEMA 440. The 5% damped freefield spectrum is
modified in the period range
1
0.75 T T > to account for the additional
damping provided by the foundation.
E.8 Alternate Procedure for Hazard Characterization for
ScenarioBased Assessment Using Nonlinear
ResponseHistory Analysis
Scenariobased assessments utilize distributions in spectral demand for the
purpose of scaling earthquake ground motions for responsehistory analysis.
Ground motion intensity is assumed to be lognormally distributed with
median value u , and dispersion,  .
Eleven values of spectral acceleration at each userselected period are used to
characterize the distribution of seismic demand for scenarioassessments as
follows:
1,11
i
ai
S e i
q
u = = (E27)
where
ai
S is the
th
i target spectral acceleration at a given period and values
of
i
q are listed in Table E2.
April 15, 2009
E26 E: Ground Shaking Hazards ATC58
Table E2 Values of
i
q for Generating a Distribution of ( )
ai
S T
i
i
q i
i
q i
i
q
1 1.69 5 0.23 9 0.75
2 1.10 6 0 10 1.10
3 0.75 7 0.23 11 1.69
4 0.47 8 0.47
Figure E13 illustrates this process for a scenario earthquake having median
spectral acceleration of 0.3 g and dispersion equal to 0.4. The figure shows
the cumulative probability distribution represented by this median and
dispersion. Horizontal lines across the plot divide the distribution into eleven
stripedregions, each having a probability of occurrence of 9.09%. For each
stripe, the central or midpoint value of the probability of exceedance is
shown by with values of 4.55%, 13.64%, 22.73%, 31.82%, 40.91%, 50%,
59.09%, 68.18%, 77.27%, 86.36% and 95.45%. A dashed horizontal line is
drawn across the plot from the vertical axis to intersect the cumulative
distribution function and dropped vertically to the horizontal axis, where
spectral acceleration values of .153 g, .193 g, .222 g, 0.248 g, 0.274 g, 0.300
g, 0.329 g, 0.362 g, 0.405 g, 0.465 g and 0.590 g, respectively, are read off.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Spectral acceleration (g)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
P
r
o
b
a
b
i
l
i
t
y
o
f
n
o
n

e
x
c
e
e
d
a
n
c
e
u=0.3g; =0.4
Figure E13 Calculation of spectral accelerations given a lognormal
distribution
April 15, 2009
ATC58 E: Ground Shaking Hazards E27
Sidebar: Calculation of hazard values for scenariobased assessments.
The values of spectral acceleration,
1 a
S to
11 a
S , characterize the distribution
of spectral acceleration shown in the figure and can be obtained from the
following equation:
1
( )
i
P
ai
S e

u
÷
·u
= ·
where
1 ÷
u is the inverse standardized normal distribution and
i
P is the
midpoint cumulative probability for region i. The values for
i
q in Table E2
are calculated from
1
( )
i
P
÷
u for 11 target spectral accelerations.
Alternatively, these values can be generated using an Excel spreadsheet and
the function LOGINV (p, Inu ,  ), where values of p =(0.04545 …..
0.9545), Inu (In(0.3)=1.204) is the mean, and  (=0.4) is the dispersion.
For scenariobased assessment, seismic performance and loss are computed
for a userspecified combination of earthquake moment magnitude and site
tosource distance. Attenuation relationships are used to predict median
spectral demands and dispersions as a function of period.
Identical to the procedures of Section 5.6 involving the Uniform Hazard
Spectrum, the user must specify a period range (
max min
, T T ) for scaling pairs
of ground motions. The characterization of the earthquake hazard can involve
the following steps:
1. select the magnitude and sitetosource distance for the scenario event
2. select an appropriate attenuation relationship(s) for the region, site soil
type and source characteristics
3. establish the period range (
max min
, T T ) for scaling pairs of ground motions
4. determine the median spectral acceleration demands, u , and the
corresponding dispersions,  , in the period range (
max min
, T T ) using the
attenuation relationship(s) of step 2
5. establish 11 target spectra, ( ), 1,11
ai
S T i = , using the medians and
dispersions from step 4 and equation (57) in the period range (
max min
, T T )
6. select 11 pairs of ground motions
7. construct a geomean spectrum for each pair of ground motions
8. for each target spectrum of step 5, scale a pair of motions such that for
each period in the range (
max min
, T T ), the geomean spectrum matches the
response spectrum ( )
ai
S T
9. repeat step 8 for the remaining 10 spectra.
April 15, 2009
E28 E: Ground Shaking Hazards ATC58
The product of step 9 will be 11 pairs of earthquake histories for response
history analysis per Chapter 6.
Step 8 in this procedure requires the selection of pairs of motions in step 6
with geomean spectra in the period range (
max min
, T T ) that are similar in shape
to the target spectra of step 5. If appropriate pairs of motions are unavailable,
an alternative to steps 7, 8 and 9 is to spectrally match pairs of motions to the
target spectra of step 5, with each component in a pair matched to the same
target spectrum.
A refinement to the above 9step procedure involves the use of Conditional
Mean Spectra and the effective fundamental period T as defined in Section
5.6.1. Steps 4 through 9 above are altered as follows:
4. determine the median spectral acceleration demand, u , and the
corresponding dispersion,  , at effective fundamental period T , using
the attenuation relationship(s) of step 2
5. compute 11 target values of spectral acceleration, ( ), 1,11
ai
S T i = , using
the medians and dispersions from step 3 and (57)
6. generate Conditional Mean Spectra (CMS
i
) for the 11 target values of
( )
ai
S T given magnitude and distance from step 1 and q from Section
5.6.3
7. select 11 pairs of ground motions, , 1,11
i
p i = such that the shape of the
geomean spectrum for
i
p is similar to that of the CMS
i
of step 5 over a
userspecified range of period, which must include
1
X
T and
1
Y
T
8. amplitude scale the geomean spectral acceleration of the two components
of
i
p to ( )
ai
S T
9. repeat step 8 for the remaining 10 spectra.
Similar to the first procedure presented in this section, spectral matching of
pairs of motions can be used in lieu of selecting pairs of ground motions with
geomean spectral shapes that are similar to CMS
i
.
April 15, 2009
ATC58 F: Fragility Development F1
Appendix F
Fragility Development
F.1 Introduction
F.1.1 Purpose
This appendix provides guidelines for development of fragility functions for
individual building components for use in building performance assessment.
It may be used to set fragility functions for either structural or nonstructural
components, elements or systems.
F.1.2 Fragility Function Definition
Fragility functions are probability distributions that are used to indicate the
probability that a component, element or system will be damaged to a given
or more severe damage state as a function of a single predictive demand
parameter such as story drift or floor acceleration. Here, fragility functions
take the form of lognormal cumulative distribution functions, having a
median value u and logarithmic standard deviation, or dispersion, . The
mathematical form for such a fragility function is:
( )


.

\

u =
i
i
i
D
D F

u ln
) ( (F1)
where: F
i
(D) is the conditional probability that the component will be
damaged to damage state “i” or a more severe damage state as a function of
demand parameter, D; u denotes the standard normal (Gaussian) cumulative
distribution function, u
i
denotes the median value of the probability
distribution, and β
i
denotes the logarithmic standard deviation. Both u and β
are established for each component type and damage state using one of the
methods presented in Section F.2.
The conditional probability that a component will be damaged to damage
state “i” and not to a more or less severe state, given that it experiences
demand, D is given by:
  ( ) ( ) D F D F D i P
i i
÷ =
+1
(F2)
where F
i+1
(D) is the conditional probability that the component will be
damaged to damage state “i+1” or a more severe state and F
i
(D) is as
previously defined. Note that, when 
i+1
is unequal to 
i
, Equation F2 can
April 15, 2009
F2 F: Fragility Development ATC58
produce a meaningless negative probability at some levels of D. This case is
addressed in Section F.3.4.
Figure F1 (a) shows the form of a typical fragility function when plotted in
the form of a cumulative distribution function and (b) the calculation of the
probability that a component will be in damage state “i” at a particular level
of demand, d.
Figure F1 Illustration of (a) fragility function, and (b) evaluating individual
damagestate probabilities
The dispersion, , represents uncertainty in the actual value of demand, D, at
which a damage state is likely to initiate in a component . This uncertainty is
a result of variability in the quality of construction and installation of the
April 15, 2009
ATC58 F: Fragility Development F3
components in a building, as well as variability in the loading history that the
component may experience before it fails. When fragility parameters are
determined on the basis of a limited set of test data, two components of the
dispersion should be considered. The first of these, termed herein 
r
,
represents the random variability that is observed in the available test data
from which the fragility parameters are determined. The second portion, 
u
,
represents uncertainty that the tests represent the actual conditions of
installation and loading that a real component in a building will experience,
or that the available test data is an inadequate sample to accurately represent
the true random variability. The dispersion parameter , is computed as:
2 2
u r
   + = (F3)
In these guidelines, the following minimum values of the uncertainty
parameter 
u
are recommended for use: A minimum value of 0.25 should be
used if any of the following apply:
 Test data are available for five (5) or fewer specimens
 In an actual building, the component can be installed in a number of
different configurations, however, all specimens tested had the same
configuration.
 All specimens were subjected to the same loading protocol
 Actual behavior of the component is expected to be dependent on two or
more demand parameters, e.g. simultaneous drift in two orthogonal
directions, however, specimens were loaded with only one of these
parameters.
If none of the above conditions apply, a value of 
u
of 0.10 may be used.
F.1.3 Derivation Methods
Fragility functions can best be derived when there is a large quantity of
appropriate test data available on the behavior of the component of interest at
varying levels of demand. FEMA 461 provides recommended protocols for
performing such tests and recording the data obtained. Since testing is
expensive and time consuming, there is not a great body of test data presently
available to serve as the basis for determining fragility functions for many
building components. Therefore, these guidelines provide procedures for
developing the median (u) and dispersion () values for a fragility under five
different conditions of data. These are:
April 15, 2009
F4 F: Fragility Development ATC58
a. Actual Demand Data: When test data is available from M number of
specimens and each tested component actually experienced the damage
state of interest at a known value of demand, D.
b. Bounding Demand Data: When test data or earthquake experience data
are available from M number of specimens, however, the damage state of
interest only occurred in some specimens. For the other specimens,
testing was terminated before the damage state occurred or the
earthquake did not damage the specimens. The value of the maximum
demand, D
i
, to which each specimen was subjected is known for each
specimen. This maximum demand need not necessarily be the demand at
which the damage state initiated.
c. Capable Demand Data: When test data or earthquake experience data
are available from M number of specimens, however, the damage state of
interest did not occur in any of the specimens. The maximum value of
demand, D
i
, to which each specimen was subjected is known.
d. Derivation (analysis): When no test data are available, however, it is
possible to model the behavior and estimate the level of demand at which
the damage state of interest will occur.
e. Expert Opinion: When no data are available and analysis of the behavior
is not feasible, however, one or more knowledgeable individuals can
offer an opinion as to the level of demand at which damage is likely to
occur, based either on experience or judgment.
In addition, a procedure is presented for updating existing fragility functions
as more data become available. Section C2.6 provides this guidance.
F.1.4 Documentation
This section provides recommendations for documenting the basis for
fragility functions. Each fragility function should be accompanied by
documentation of the sources of data and procedures used to establish the
fragility parameters in sufficient detail that others may evaluate the adequacy
of the process and findings. As a minimum, it is recommended that the
documentation include the following:
1. Description of applicability. Describe the type of component that the
fragility function addresses including any limitations on the type of
installation to which the fragility applies.
2. Description of specimens. Describe the specimens used to establish the
fragility including identifying the number of specimens examined, their
April 15, 2009
ATC58 F: Fragility Development F5
locations, and the specific details of the specimen
fabrication/construction, mounting and installation.
3. Demands and Load Application. Detail the loading protocol or
characteristics of earthquake motion applied to each specimen. Identify
the demand parameters examined that might be most closely related to
failure probability and define how demand is calculated or inferred from
the loading protocol or excitation. Indicate whether the reported demand
quantities are the value at which damage occurred (Method A data) or
the maximum to which each specimen was subjected.
4. Damage state. Fully describe each damage state for which fragilities are
developed including the kinds of physical damage observed and any
forcedeformation quantities recorded. Define damage states
quantitatively in terms of the repairs required or potential downtime or
casualty consequences.
5. Observation summary, analysis method, and results. Present a tabular or
graphical listing of specimens, demand parameters, and damage states.
Identify the method(s) used to derive the fragility parameters per Section
F.2 of this Appendix. Present resulting fragility function parameters u
and β and results of tests to establish fragility function quality (discussed
below). Provide sample calculations.
F.2 Fragility Parameter Derivation
F.2.1 Actual Demand Data
This section defines the procedures for deriving fragility parameters (u, )
when data is available from a suitable series of tests and in each specimen,
the damage state of interest was initiated at a known value of the demand. In
this case, the median value of the demand at which the damage state is likely
to initiate, u, is given by the equation:


.

\

¿
=
=
M
i
i
d
M
e
1
ln
1
u (F3)
where:
M = total number of specimens tested to at least the initiation of the
damage state
d
i
= demand in test “i” at which the damage state was first observed
to occur.
The value of the random dispersion, 
r
, is given by:
April 15, 2009
F6 F: Fragility Development ATC58
2
1
2
ln
1
1


.

\


.

\


.

\

÷
=
¿
=
M
i
i
r
d
M
u
 (F4)
where M, r
i
and u are as defined above.
If one or more of the r
i
data appear to lie far from the bulk of the data, either
above or below, apply the procedure specified in Section C3.2. Finally, test
the resulting fragility parameters using the Lilliefors goodnessoffit test
(Section C3.3). If it passes at the 5% significance level, the fragility function
may be deemed acceptable.
Example: Determine the parameters u and , from a series of 10 tests, all of
which produced the damage state of interest. Demands at which the damage
state initiated are respectively story drifts of: 0.9, 0.9, 1.0, 1.1, 1.1, 1.2, 1.3,
1.4, 1.7, and 2 percent.
Test # Demand di ln(di) ln(di/ui) ln(di/ui)^2
1 0.9 0.10536 0.30384 0.092321
2 0.9 0.10536 0.30384 0.092321
3 1 0 0.19848 0.039396
4 1.1 0.09531 0.10317 0.010645
5 1.1 0.09531 0.10317 0.010645
6 1.2 0.182322 0.01616 0.000261
7 1.3 0.262364 0.063881 0.004081
8 1.4 0.336472 0.137989 0.019041
9 1.7 0.530628 0.332145 0.11032
10 2 0.693147 0.494664 0.244692
E 1.984833 0.623723
( )
22 . 1
9848 . 1
10
1
ln
1
1
= =
¿
=

.

\



.

\

=
e e
M
i
i
d
M
u
( )
( ) 26 . 0 6237 . 0
1 10
1
ln
1
1
1
2
=
÷
=

.

\


.

\

÷
=
¿
=
M
i
i
r
d
M
u

F.2.2 Bounding Demand Data
This section defines the procedures for deriving fragility parameters (u, )
when data are available from a suitable series of tests or earthquake
experience records; however, the damage state of interest was initiated in
only some of the specimens. For the other specimens, loading applied
during the testing or earthquake shaking was insufficient to initiate the
damage state of interest occurred. For each specimen “i”, it is necessary to
know the maximum value of the demand, d
i
to which the specimen was
subjected, and whether or not the damage state did occur in the specimen.
April 15, 2009
ATC58 F: Fragility Development F7
Divide the data into a series of N bins. It is suggested, but not essential, that
N be taken as the largest integer that is less than or equal to the square root of
M, where M is the total number of specimens available, as this will usually
result in an appropriate number of approximately equally sized sets.
In order to divide the specimens into the several bins, sort the specimen data
in order of ascending maximum demand value, d
i
, for each test, then divide
the list into N groups of approximately equal size. Each group “j” will have
M
j
specimens, where:
M M
N
i
j
=
¿
=1
(F5)
Next, determine the average value of the maximum demand for each bin of
specimens:
¿
=
=
j
M
k
k
j
j
d
M
d
1
1
(F6)
and x
j
, the natural logarithm of d
j
(i.e., ln(d
j
)). Also determine the number of
specimens within each bin, m
j
, in which the damage state of interest was
achieved and the inverse standard normal distribution, y
j
, of the failed
fraction specimens in the bin:
1
1
1
j
j
j
m
y
M
÷
  +
= u


+
\ .
(F7)
That is, determine the number of standard deviations, above the mean that
the stated fraction lies, assuming a mean value, µ=0 and a standard deviation,
o=1. This can easily be determined using the “normsinv” function on a
Microsoft Excel spreadsheet or by referring to standard tables of the normal
distribution. Next, fit a straight line to the data points, x
j
, y
j
, using a least
squares approach. The straight line will have the form:
c bx y + = (F8)
where, b is the slope of the line and c is the y intercept. The slope b is given
by:
( )( )
( )
¿
¿
=
=
÷
÷ ÷
=
M
i
j
M
i
j j
x x
y y x x
b
1
2
1
(F9)
April 15, 2009
F8 F: Fragility Development ATC58
¿
=
=
M
j
j
x
M
x
1
1
(F10)
1
1
M
j
j
y y
M
=
=
¿
(F11)
Determine the value of the random dispersion, 
r
as:
( )
( )( )
1
1
1
M
j
i
r
M
j j
i
x x
b
x x y y

=
=
÷
= =
÷ ÷
¿
¿
(F12)
The value of the median, u, is taken as:
( )
r
r
x y
c
e e


u
÷
÷
= =
(F13)
Example: Consider the damage statistics shown in the figure below. The figure depicts the hypothetical
performance of motor control centers (MCCs) observed after various earthquakes in 45 facilities. Each box
represents one specimen. Several damage states are represented. Crosshatched boxes represent MCCs that
experienced a noticeable earthquake effect such as shifting but that remained operable. Black boxes represent
those that were found to be inoperable following the earthquake. Each stack of boxes represents one facility.
Calculate the fragility function using PGA as the demand parameter, binning between halfway points between
PGA values shown in the figure.
Undamaged
Affected but operable
Not operable
0.20 g 0.30 g 0.40 g 0.50 g 0.60 g
1
9
T
o
t
a
l
2
0
T
o
t
a
l
Figure Hypothetical observed earthquake damage data for motor control centers
The number of bins, N, and the lower demand bounds a
j
, are dictated by the available data: N is taken as 5
with lower bounds, a
j
of 0.15g, 0.25g, 0.35g, 0.45g, and 0.55g respectively. The damage state of interest is
loss of postearthquake functionality (black boxes in figure). The values of M
j
and m
j
are found by counting
all boxes and black boxes, respectively, in the figure in each bin, and are shown in the table below. The value
of M is found by summing: M = ΣM
j
= 260. Values x
j
and y
j
are calculated as x
j
= ln(
j
r ), and y
j
= u

1
((m
j
+1)/(M
j
+1)). Average values are calculated as shown: x = –0.99, y = –1.05, according to Equations F10
and F11. For each bin, the values of
j
x x ÷ and
j
y y ÷ are calculated as shown.
April 15, 2009
ATC58 F: Fragility Development F9
Table Example solution data
j a
j
(g)
j
r (g)
M
j
m
j
x
j
y
j
j
x x ÷
j
y y ÷
( )
2
j
x x ÷ ( )( )
j j
x x y y ÷ ÷
1 0.15 0.2 52 0 1.61 2.08 0.623 1.031 0.388 0.642
2 0.25 0.3 48 4 1.20 1.27 0.217 0.223 0.047 0.049
3 0.35 0.4 84 8 0.92 1.25 0.070 0.202 0.005 0.014
4 0.45 0.5 35 15 0.69 0.14 0.294 0.907 0.086 0.266
5 0.55 0.6 41 12 0.51 0.50 0.476 0.549 0.226 0.261
Σ = 260 4.93 5.23 0.753 1.204
Average = 0.99 1.05
Then, β and u are calculated as:
( )
( )( )
2
1
0.753
0.63
1.204
N
j
j
r
j j
x x
x x y y

=
÷
= = =
÷ ÷
¿
( ) ( ) ( ) 0.99 1.05 0.63
( 0.329)
0.72
r
x y
e e e g

u
÷ + ÷ ÷
= = = =
F.2.3 Capable Demand Data
This section defines the procedures for deriving fragility parameters (u, )
when data is available from a suitable series of specimens, however, the
damage state of interest was not initiated in any of the specimens. For each
available specimen, ”i,” the maximum demand at which the specimen was
loaded, “d
i
”, and whether or not the specimen experienced any distress or
damage must be known.
From the data for M specimens, determine the maximum demand
experienced by each specimen, d
max
, and the minimum demand for any of the
specimens that exhibited any distress or damage, d
min
. Determine d
a
as the
smaller of d
min
or 0.7d
max
. Determine M
A
as the number of specimens that did
not exhibit distress or damage, but that were loaded with demands, d
i
> d
a
;
M
B
as the number of specimens that exhibited distress or damage, but which
did not appear to be initiating or on the verge of initiating the damage state of
interest; and M
C
as the number of specimens appeared to be on the verge of
initiating the damage state of interest.
April 15, 2009
F10 F: Fragility Development ATC58
If none of the specimens in any of the tests exhibited any sign of distress or
damage, take the value of d
m
as d
max
. If one or more of the specimens
exhibited distress or damage of some type, take d
m
as:
2
max a
m
d d
d
+
= (F14)
Determine the subjective failure probability S at d
m
as:
C B A
B C
M M M
M M
S
+ +
+
=
1 . 0 5 . 0
(F15)
Take the logarithmic standard deviation, , as having a value of 0.4.
Determine the median, u, as:
0.4z
m
d e u
÷
=
(F16)
where, z is determined from Table F1 based on the value of M
A
and S.
Table F1 Values of z
Conditions Z
M
A
≥ 3 and S = 0 2.326
M
A
< 3 and S ≤ 0.075 1.645
0.075 < S ≤ 0.15 1.282
0.15 < S ≤ 0.3 0.842
S > 0.3 0.253
Example: Determine the parameters u and , from tests of 10 specimens.
Five of the specimens had maximum imposed drift demands of 1% with no
observable signs of distress. Three of the specimens had maximum imposed
drift demands of 1.5% and exhibited minor distress, but did not appear to be
at or near the initiation of the damage state of interest. Two of the specimens
had maximum imposed drift demands of 2%, did not enter the damage state
of interest during the test, but appeared to be about to sustain such damage.
From the given data determine: d
max
= 2%, d
min
= 1.5%. d
a
is the smaller of
0.7d
max
or d
min
and therefore, is 0.7(2%) = 1.4%. M
A
= 0, M
B
=3, and M
C
=2.
% 7 . 1
2
% 4 . 1 % 2
2
max
=
+
=
+
=
a
m
d d
d
26 . 0
2 3 0
) 3 ( 1 . 0 ) 2 ( 5 . 0 1 . 0 5 . 0
=
+ +
+
=
+ +
+
=
C B A
B C
M M M
M M
S
From Table 21, z is taken as 0.842. Therefore,
0.4 0.4 ( 0.842 )
1.7% 2.4% u
÷ ÷ ÷
= = =
z
m
d e e
and  is taken as 0.4.
April 15, 2009
ATC58 F: Fragility Development F11
F.2.4 Derivation
There are two methods available for analytical derivation of fragility
parameters. The first of these uses a single calculation of the probable
capacity and a default value of the logarithmic standard deviation. The
second method uses Monte Carlo analysis to explore the effect of variation in
material strength, construction quality and other random variables.
Single calculation
Calculate the capacity of the component, Q in terms of a demand parameter,
d, using average material properties and dimensions and estimates of
workmanship. Resistance factors should be taken as unity and any
conservative bias in code equations, if such equations are used, should be
removed. The logarithmic standard deviation, , is taken as having a value
of 0.4. The median capacity u is taken as:
Q 92 . 0 = u (F17)
Monte Carlo simulation
Identify all those factors, important to predicting the capacity that are
uncertain including material strength, cross section dimensions, member
straightness, workmanship. Estimate a median value and dispersion for each
of these random variables. Conduct sufficient analyses, randomly selecting
the values of each of these random variables in accordance with their
estimated distribution properties, each time calculating the capacity.
Determine the median value of the capacity as that capacity exceeded in 50%
of the calculations. Determine the random logarithmic standard deviation,

r
, as the standard deviation of the natural logarithm of the calculated
capacity values. Use equation 13 to determine the total logarithmic standard
deviation, , assuming a value of 
u
of 0.25.
F.2.5 Expert Opinion
Select one or more experts with professional experience in the design or
postearthquake damage observation of the component of interest. Solicit
their advice using the format shown in
April 15, 2009
F12 F: Fragility Development ATC58
. Note the suggested inclusion of representative images, which should be
recorded with the responses. If an expert refuses to provide estimates or
limits them to certain conditions, either narrow the component definition
accordingly and iterate, or ignore that expert’s response and analyze the
remaining ones.
Calculate the median value, u, as:
¿
¿
=
=
=
N
i
i
N
i
i i
w
w
1
5 . 1
1
5 . 1
u
u (F18)
where, N is the number of experts providing an opinion; u
i
is expert “i’s”
opinion as to the median value, and w
i
is expert “i’s” level of expertise, on a
15 scale.
Calculate the lower bound value for the capacity as:
¿
¿
=
=
=
N
i
i
li
N
i
i
l
w
d w
d
1
5 . 1
1
5 . 1
(F19)
where, d
li
is expert “i’s” opinion as to the lower bound value and other terms
are as previously defined. The value of the logarithmic standard deviation,
, is taken as:
( )
28 . 1
/ ln
l
d u
 = (F20)
If this calculation produces an estimate of β that is less than 0.4, either
justify the β, or take  as having a value of 0.4 and recalculate u as:
l
d 67 . 1 = u (F21)
April 15, 2009
ATC58 F: Fragility Development F13
Objective. This form solicits your judgment about the values of a demand parameter (D) at which a
particular damage state occurs to a particular building component. Judgment is needed because the
component may contribute significantly to the future earthquake repair cost, fatality risk, or post
earthquake operability of a building, and because relevant empirical and analytical data are currently
impractical to acquire. Your judgment is solicited because you have professional experience in the design
or postearthquake damage observation of the component of interest.
Definitions. Please provide judgment on the damageability of the following component and damage
state. Images of a representative sample of the component and damage state may be attached. It is
recognized that other demand parameters may correlate better with damage, but please consider only the
one specified here.
Component name:
Component definition:
Damage state name:
Damage state definition:
Relevant Demand Parameter
Definition of Demand Parameter
Uncertainty; no personal stake. Please provide judgment about this general class of components, not any
particular instance, and not one that you personally designed, constructed, checked, or otherwise have
any stake in. There is probably no precise threshold level of demand that causes damage, because of
variability in design, construction, installation, inspection, age, maintenance, interaction with nearby
components, etc. Even if there were such a precise level, nobody might know it with certainty. To
account for these uncertainties, please provide two values of demand at which damage occurs:
median and lower bound.
Estimated median capacity Definition. Damage would occur at this
level of demand in 5 cases out of 10, or in a single instance, you judge there to be an equal chance that
your median estimate is too low or too high.
Estimated lowerbound capacity Definition. Damage would occur at this
level of demand in 1 case in 10. In a single case, you judge there to be a 10% chance that your estimate is
too high. Judge the lower bound carefully. Make an initial guess, then imagine all the conditions that
might make the actual threshold demand lower, such as errors in design, construction or installation,
substantial deterioration, poor maintenance, more interaction with nearby components, etc. Revise
accordingly and record your revised estimate. Research shows that without careful thought, expert
judgment of the lower bound tends to be too close to the median estimate, so think twice and do not be
afraid of showing uncertainty.
On a 1to5 scale, please judge your expertise with this component and damage state, where 1 means “no
experience or expertise” and 5 means “very familiar or highly experienced.”
Your level of expertise:
Your name: Date:
Figure F2 Form for soliciting expert judgment on component fragility
F.2.6 Updating
This section addresses procedures for reevaluating fragility parameters for a
building component as additional data become available. The preexisting
and updated fragility parameters are respectively termed u, , u’, and ’. The
additional data are assumed to be a set of M specimens with known
April 15, 2009
F14 F: Fragility Development ATC58
maximum demand and damage states. It is not necessary that any of the
specimens experienced damage.
Calculate the revised median, u’ and logarithmic standard deviation ’ as
follows:
5
1
ln( )
j j
j
w d
e u
=
'
¿
'
=
(F21)
¿
=
' = '
5
1 j
j j
w   (F22)
where:
( )
( )
1
5
1 1
,
,
M
j
i
j M
j
j i
w L i j
w
w L i j
=
= =
' =
[
¿ [
(F23)
where Π denotes the product of the terms that come after it, and
( )
( )


.

\

u ÷ =
j
j i
s
x d
j i L
 707 . 0
ln
1 , if specimen “i” did not experience the damage
state of interest, and
( )
( )


.

\

u =
j
j i
s
x d
j i L
 707 . 0
ln
, if specimen “i” did experience the damage state of
interest, or a more severe state. and
6 / 1
3 / 1
36 . 1
64 . 0
5 4 3 2
1
5
4
3 2 1
22 . 1
3
22 . 1
2
5 4 1
= = = =
=
=
=
= = =
=
=
= = =
÷
w w w w
w
e x
e x
x x x
 
 
   
u
u
u


F.3 Assessing Fragility Function Quality
This section provides procedures that can be used to assess the quality of
fragility parameters.
April 15, 2009
ATC58 F: Fragility Development F15
F.3.1 Competing Demand Parameters
The behavior of some components may be dependent on several types of
demands, for example inplane and outofplane drift, or both drift and
acceleration. It may not be clear which demand is the best single predictor of
component damage. Assuming that data are available to create fragility
functions for each possibly relevant demand, do so. Choose the fragility
function that has the lowest β.
F.3.2 Dealing with Outliers using Pierce’s Criterion
When fragilities are determined on the basis of actual demand data (Section
C2.1), it is possible that one or more tests reported spurious values of
demand, d
i
, and reflect experimental errors rather than the true demands at
which the specimens failed. In cases where one or more d
i
values in the data
set are obvious outliers from the bulk of the data, investigate whether the
data reflects real issues in the damage process that may recur, especially
where d
i
<< u for these outliers. If there is no indication that these data
reflect a real recurring issue in the damage process, apply the following
procedure (Peirce’s criterion) to test and eliminate doubtful observations of
d
i
.
1. Calculate ln(u) and β of the complete data set.
2. Let D denote the number of doubtful observations, and let R denote the
maximum distance of an observation from the body of the data, defined
as:

u ln ) ln( ÷
=
d
R (F24)
where u, β, and M are as previously defined, d is a measured demand value,
and R is as shown in Table F. Assume D = 1 first, even if there appears to be
more than one doubtful observation.
3. Calculate the maximum allowable deviation:  ln(d) – ln(u) 
max
. Note that
this can include d >>u and d << u.
4. For any suspicious measurement d
i
, obtain  ln(d
i
) – ln(u) .
5. Eliminate the suspicious measurements if:
max
ln( ) ln( ) ln( ) ln( )
i
d d u u ÷ > ÷
6. If this results in the rejection of one measurement, assume D=2, keeping
the original values of u and β, and go to step 8.
7. If more than one measurement is rejected in the above test, assume the
next highest value of doubtful observations. For example, if two
April 15, 2009
F16 F: Fragility Development ATC58
measurements are rejected in step 5, assume the case of D = 3, keeping
the original values of u, and β, as the process is continued.
8. Repeat steps 2 – 5, sequentially increasing D until no more data
measurements are eliminated.
9. Obtain u and β of the reduced data set as for the original data.
Table F2 Parameters for Applying Peirce's Criterion
M D=1 D=2 D=3 D=4 D=5 D=6 D=7 D=8 D=9
3 1.1960
4 1.3830 1.0780
5 1.5090 1.2000
6 1.6100 1.2990 1.0990
7 1.6930 1.3820 1.1870 1.0220
8 1.7630 1.4530 1.2610 1.1090
9 1.8240 1.5150 1.3240 1.1780 1.0450
10 1.8780 1.5700 1.3800 1.2370 1.1140
11 1.9250 1.6190 1.4300 1.2890 1.1720 1.0590
12 1.9690 1.6630 1.4750 1.3360 1.2210 1.1180 1.0090
13 2.0070 1.7040 1.5160 1.3790 1.2660 1.1670 1.0700
14 2.0430 1.7410 1.5540 1.4170 1.3070 1.2100 1.1200 1.0260
15 2.0760 1.7750 1.5890 1.4530 1.3440 1.2490 1.1640 1.0780
16 2.1060 1.8070 1.6220 1.4860 1.3780 1.2850 1.2020 1.1220 1.0390
17 2.1340 1.8360 1.6520 1.5170 1.4090 1.3180 1.2370 1.1610 1.0840
18 2.1610 1.8640 1.6800 1.5460 1.4380 1.3480 1.2680 1.1950 1.1230
19 2.1850 1.8900 1.7070 1.5730 1.4660 1.3770 1.2980 1.2260 1.1580
20 2.2090 1.9140 1.7320 1.5990 1.4920 1.4040 1.3260 1.2550 1.1900
>20 alnM + b
a 0.4094 0.4393 0.4565 0.4680 0.4770 0.4842 0.4905 0.4973 0.5046
b 0.9910 0.6069 0.3725 0.2036 0.0701 0.0401 0.1358 0.2242 0.3079
F.3.3 Goodness of Fit Testing
Fragility parameters that are developed based on actual demand data (Section
C2.1) should be tested for goodness of fit in accordance with this section.
Calculate
April 15, 2009
ATC58 F: Fragility Development F17
) ( ) ( max d S d F D
M i x
÷ = (F25)
where S
M
(d) denotes the sample cumulative distribution function
( )
1
1
( )
M
M i
i
S d H d d
M
=
= ÷
¿
(F26)
and H is taken as:
 1.0 if d
i
– d is positive
 ½ if d
i
– d is zero
 0 if d
i
– d is negative.
If D > D
crit
from Table F, the fragility function fails the goodness of fit test.
This result is used in assigning a quality level to the fragility function. Use α
= 0.05.
Table F3 Critical Values for the Lilliefors Test
Significance Level D
crit
o = 0.15 0.775 / (M
0.5
– 0.01 + 0.85M
–0.5
)
o = 0.10 0.819 / (M
0.5
– 0.01 + 0.85M
–0.5
)
o = 0.05 0.895 / (M
0.5
– 0.01 + 0.85M
–0.5
)
o = 0.025 0.995 / (M
0.5
– 0.01 + 0.85M
–0.5
)
F.3.4 Fragility Functions that Cross
Some components will have two or more possible damage states, with a
defined fragility function for each. For any two (cumulative lognormal)
fragility functions i and j with medians u
j
> u
i
and logarithmic standard
deviations β
i
≠ β
j
, the fragility functions will cross at extreme values. In such
a case, adjust the fragility functions by one of the following two methods;
Method 1: adjust the fragility functions such that.
( )
)
`
¹
¹
´
¦


.

\

u =
i
i
j i
D
D F

u ln
max ) ( for all j ≥ i (F27)
This has the effect that for the damage state with the higher median value, the
probability of failure, F
i
(D) is never taken as less than the probability of
failure for a damage state with a lower median value.
April 15, 2009
F18 F: Fragility Development ATC58
Method 2: First establish u and β values for the various damage states
independently. Next calculate the average of the dispersion values for each
of the damage states with crossing fragility curves as:
¿
=
= '
N
i
i i
N
1
1
  (F28)
This average logarithmic standard deviation is used as a replacement for the
independently calculated values. An adjusted median value must be
calculated for each of the crossing fragilities as:
( ) ( )
1.28 ln
i i
i
e
  u
u
'÷ +
' =
(F29)
F.3.5 Assigning a Single Quality Level to a Fragility Function
Assign each fragility function a quality level of high, medium, or low, as
shown in Table F4.
April 15, 2009
ATC58 F: Fragility Development F19
Table F4 Fragility Function Quality Level
Quality Method Peer
reviewed*
Number of
specimens
Other
High A Yes ≥ 5 Passes Lilliefors test at 5%
significance level. Examine
and justify (a) differences of
greater than 20% in θ or β,
compared with past
estimates, and (b) any case of
β < 0.2 or β > 0.6.
B Yes ≥ 20
Examine and justify (a)
differences of greater than
20% in θ or β, compared
with past estimates, and (b)
any case of β < 0.2 or β >
0.6.
U Yes ≥ 6 Prior was at least moderate
quality
Moderate A ≥ 3 Examine and justify any case
of β < 0.2 or β > 0.6.
B ≥ 16 Examine and justify any case
of β < 0.2 or β > 0.6.
C Yes ≥ 6
D Yes
E Yes At least 3 experts with w ≥ 3
U ≥ 6 or prior was moderate quality
Low All other cases
* Data and derivation published in a peerreviewed archival journal.
April 15, 2009
ATC58 G: Generation of Realizations for Loss Computations G1
Appendix G
Generation of Realizations for
Loss Computations
G.1 Loss Computations
The losscomputation algorithm in PACT generates a large number of
realizations (or vectors) of demand per intensity level to develop a loss curve.
This appendix describes the two algorithms in PACT that are used to
transform the demands computed by nonlinear responsehistory analysis
(multiple demand vectors per intensity level) or simplified analysis (1 vector
per intensity level), into a large number of realizations
1
. Section G.2 presents
the algorithm for assessment using nonlinear responsehistory analysis.
Section G.3 presents the algorithm for assessment using the simplified
analysis procedure. For both types of assessment, the input vector(s) of
demand parameters is (are) not included in the generated realizations used for
loss computations.
G.2 Realizations for Assessment Using Nonlinear
ResponseHistory Analysis
G.2.1 Introduction
Section G.2.2 introduces the algorithm implemented in PACT to transform a
limited number of vectors of demand parameters established by response
history analysis to the large number of demand vectors used for loss
calculations. Section G.2.3 presents sample calculations to illustrate the
1
In these Guidelines, nonlinear responsehistory analysis is recommended
for typical assessment, a linear static procedure is provided for simplified
analysis, and 500 realizations are generated per intensity level for loss
computations. A minimum of 11 simulations per intensity level are
recommended for nonlinear analysis and one simulation is used for linear
analysis. These procedures, numbers of analyses and number of
realizations are not mandatory and others can be used. This appendix uses
the default procedures, number of realizations and numbers of demand
vectors per intensity level, 11 for nonlinear responsehistory analysis (for
basic assessment) and 1 for linear analysis (for simplified assessment), to
illustrate the algorithms included in PACT to generate a large number of
realizations for loss computations.
April 15, 2009
G2 G: Generation of Realizations for Loss Computations ATC58
procedure. A Matlab script is presented in G.2.4 to enable the reader to
generate correlated vectors outside of PACT.
G.2.2 Algorithm
The technical basis for the algorithm described herein was developed by
Yang, Moehle and DerKiureghian (see Yang, 2006). The algorithm was
updated by Huang et al. (2008) to address modeling uncertainty
m
 , record
torecord variability
a
 and ground motion variability
gm
 .
For typical assessment, a limited number, m, of response analyses (the
minimum number is 11) are performed at each intensity level. For each
analysis, the peak absolute value of each demand parameter (e.g., third story
drift, roof acceleration) is assembled into a row vector with n entries, where n
is the number of demand parameters. The m row vectors are catenated to
form an m×n matrix (rows ×columns; simulations ×demand parameters)–
each column presents m values of one demand parameter.
Denote the matrix of demand parameters X. The entries in X are assumed
to be jointly lognormal (see Appendix A for a discussion of joint probability
distributions). The natural logarithm of each entry in X is computed to form
an 11×n matrix Y. The entries in Y are assumed to be jointly normal and
can be characterized by 1×n mean vector,
Y
M , a n×n correlation coefficient
matrix,
YY
R , and a n×n diagonal matrix of standard deviations,
Y
D .
A vector of demand parameters, Z, can be generated with the same
statistical distribution as Y using a vector of uncorrelated standard normal
random variables, U, with a mean of 0 and a standard deviation of 1 for each
random variable. For this case
=
Y Y Y
Z = AU+ B D L U+ M (G1)
where A is a matrix of constant coefficients that linearly transforms U to Z
and B is a vector of constant coefficients that translates U to Z. Matrix A
and vector B are derived in Yang (2006). The matrix
Y
L is the transposed
Cholesky decomposition
2
of
YY
R , all other terms are defined above.
One realization of demand is generated by 1) computing
Y
M ,
Y
D and
Y
L
by sampling Y; 2) updating
Y
D for modeling uncertainty per Section G.2.4,
3) populating Uby random sampling each demand parameter on a
2
If matrix K is symmetric and positivedefinite (e.g., a stiffness matrix), it
can be decomposed into a lower triangular matrix, L, the Cholesky
triangle, and the transpose of the lower triangular matrix, such that
T
K = LL . To solve = Ku R , one solves first = Ly R for y and then
=
T
L u y for u .
April, 2009
ATC58 G: Generation of Realizations for Loss Computations G3
distribution with a mean of 0 and a standard deviation of 1.0; 4) computing
Z per (G1); and 5) taking the exponential of every entry in Z to recover the
demand parameters. Five hundred vectors of U are required for the 500
realizations: one vector per realization. The process is computationally
efficient because
Y
M ,
Y
D and
Y
L are computed just once for each intensity
of shaking. The process is illustrated in Figure G1 (from Yang 2006) below.
Figure G1 Generation of vectors of correlated demand parameters (Yang
2006)
G.2.3 Sample Application of the Algorithm
G.2.3.1 Description
The sample building is the three story steel moment frame. For this sample
application, analysis is performed along one axis and only 200 (rather than
500) realizations were generated. Seven demand parameters are used for
performance assessment for this building: 1
st
, 2
nd
and 3
rd
story drift
(expressed as a percentage of the story height); and 1
st
, 2
nd
and 3
rd
floor and
roof acceleration:
1
o ,
2
o ,
3
o ,
1
a ,
2
a ,
3
a and
r
a , respectively. Residual story
drift is not considered. The matrix X of demand parameters, 11 rows (one
per simulation) ×7 columns (one per demand parameter), is presented in
Table G1. Modeling uncertainty, recordtorecord variability and ground
motion variability are set aside here to simplify the presentation.
G2.3.2 Probability distributions
Consider the matrix
1 2 7
[ , ...... ] = X X X X above: 7 column vectors of demand,
with 11 entries per vector. For each vector, the mean (expected) value,
x
µ and the variance,
2
x
s , can be established as follows:
1
1
i n
x i
i
x
n
µ
=
=
=
¿
(G2)
2 2
1
1
( )
1
i n
x i x
i
s x
n
µ
=
=
= ÷
÷
¿
(G3)
The mean and variance of
1 2 7
[ , ...... ] = X X X X are presented in Table G2.
April 15, 2009
G4 G: Generation of Realizations for Loss Computations ATC58
Table G1 Matrix of Demand Parameters, X
1
o (%)
2
o (%)
3
o (%)
1
a (g)
2
a (g)
3
a (g)
r
a (g)
1 1.26 1.45 1.71 0.54 0.87 0.88 0.65
2 1.41 2.05 2.43 0.55 0.87 0.77 0.78
3 1.37 1.96 2.63 0.75 1.04 0.89 0.81
4 0.97 1.87 2.74 0.55 0.92 1.12 0.75
5 0.94 1.8 2.02 0.40 0.77 0.74 0.64
6 1.73 2.55 2.46 0.45 0.57 0.45 0.59
7 1.05 2.15 2.26 0.38 0.59 0.49 0.52
8 1.40 1.67 2.1 0.73 1.50 1.34 0.83
9 1.59 1.76 2.01 0.59 0.94 0.81 0.72
10 0.83 1.68 2.25 0.53 1.00 0.9 0.74
11 0.96 1.83 2.25 0.49 0.90 0.81 0.64
Table G2 Mean and Variance of X
1
o (%)
2
o (%)
3
o (%)
1
a (g)
2
a (g)
3
a (g)
r
a (g)
x
µ 1.2282 1.8882 2.2600 0.5418 0.9064 0.8364 0.6973
2
x
s 0.0878 0.0849 0.0885 0.0139 0.0615 0.0627 0.0094
J oint probability distributions can be developed for pairs of vectors, for
examples, between a)
1
X (first story drift) and
3
X (third story drift), and b)
4
X (ground floor acceleration) and
7
X (roof acceleration). The black solid
circles in Figure G2 below presents the relationships between
1
X and
3
X
(part a) and
4
X and
7
X (part b) constructed using the data of Table G1:
1
X and
3
X are weakly correlated and
4
X and
7
X are strongly correlated.
0 0.5 1 1.5 2 2.5
o
1
(%)
0
1
2
3
4
o
3
(
%
)
original data
simulated data
0 0.2 0.4 0.6 0.8 1 1.2
a
1
(g)
0
0.2
0.4
0.6
0.8
1
1.2
a
r
(
g
)
original data
simulated data
a.
1
X and
3
X b.
4
X and
7
X
Figure G2 Relationships between demand parameters
April, 2009
ATC58 G: Generation of Realizations for Loss Computations G5
A joint probability density function of the type shown in Figure G3 can be
constructed for any pair of vectors given two means, two variances, a
covariance and type of distribution (normal, lognormal). This figure presents
the joint probability density functions for
1
X and
3
X (part a) and
4
X and
7
X (part c) constructed using the data of Table G2—termed the original
data hereafter.
a. joint pdf for
1
X and
3
X , original data b. joint pdf for
1
X and
3
X , simulated data
c. joint pdf for
4
X and
7
X , original data d. joint pdf for
4
X and
7
X , simulated data
Figure G3 Joint probability density functions
G2.3.3 Analysis
The entries in matrix X are assumed to be jointly lognormal. The log of each
entry in X is used to construct the entries in matrixY, which are now
assumed to be jointly normal. Table G3 below presents the entries in Y; the
last two rows in the table present the mean and variance of each column
vector in Y. (Note that the natural logarithm of any number less than 1.00
will be negative.)
April 15, 2009
G6 G: Generation of Realizations for Loss Computations ATC58
The row vector
y
µ in Table G3 is the mean vector
Y
M of (G1). The
diagonal matrix of standard deviations
Y
D in (F1) is formed by first taking
the square root of each entry in the last row of Table G3 (standard deviation
y
s is the square root of the variance
2
y
s ) and placing the first entry
(0.2431 0.059 = ) in cell (1,1), the second entry in cell (2,2) and so on.
Matrix
Y
D for this example is presented in Table G4. Sources of
uncertainty that are not included in the responsehistory analysis, including
modeling uncertainty herein because a bestestimate model was used for
nonlinear analysis, can be included indirectly by adjusting each diagonal
entry of Table G4 by
2 2
ii
s  + , where
ii
s is the ith diagonal entry and  is
the dispersion to be included.
Table G3 Demand Parameters, Y
1
o
2
o
3
o
1
a
2
a
3
a
r
a
1 0.231 0.372 0.536 0.616 0.139 0.128 0.431
2 0.344 0.718 0.888 0.598 0.139 0.261 0.248
3 0.315 0.673 0.967 0.288 0.039 0.117 0.211
4 0.030 0.626 1.008 0.598 0.083 0.113 0.288
5 0.062 0.588 0.703 0.916 0.261 0.301 0.446
6 0.548 0.936 0.900 0.799 0.562 0.799 0.528
7 0.049 0.765 0.815 0.968 0.528 0.713 0.654
8 0.336 0.513 0.742 0.315 0.405 0.293 0.186
9 0.464 0.565 0.698 0.528 0.062 0.211 0.329
10 0.186 0.519 0.811 0.635 0.000 0.105 0.301
11 0.041 0.604 0.811 0.713 0.105 0.211 0.446
y
µ
0.179 0.625 0.807 0.634 0.131 0.222 0.370
2
y
s 0.059 0.022 0.018 0.046 0.070 0.0993 0.021
Table G4 Matrix
Y
D for the Sample Problem
0.243 0.000 0.000 0.000 0.000 0.000 0.000
0.000 0.149 0.000 0.000 0.000 0.000 0.000
0.000 0.000 0.135 0.000 0.000 0.000 0.000
0.000 0.000 0.000 0.215 0.000 0.000 0.000
0.000 0.000 0.000 0.000 0.265 0.000 0.000
0.000 0.000 0.000 0.000 0.000 0.315 0.000
0.000 0.000 0.000 0.000 0.000 0.000 0.144
April, 2009
ATC58 G: Generation of Realizations for Loss Computations G7
The correlation coefficient matrix,
YY
R , is a symmetric matrix with entries
, i j
µ , which is the correlation coefficient between random variables
i
Y and
j
Y . The range for
, i j
µ is 1 to +1, where values close to a) 1.0 indicate a
positive linear relationship between the random variables, b) 1 indicate a
negative linear relationship between the random variables, and c) values
close to or equal to 0 indicate no linear relationship between the random
variables. Table G5 presents the correlation coefficient matrix for this
sample problem. The values on the leading diagonal are equal to 1.000,
which is an expected result because
, i i
µ is the correlation of a given random
variable (7 in this example) with itself. Correlation coefficients are
dimensionless.
Table G5 Matrix
YY
R for the Sample Problem
1.000 0.339 0.019 0.375 0.022 0.193 0.145
0.339 1.000 0.656 0.353 0.646 0.723 0.376
0.019 0.656 1.000 0.136 0.094 0.066 0.220
0.375 0.353 0.136 1.000 0.839 0.731 0.881
0.022 0.646 0.094 0.839 1.000 0.934 0.863
0.193 0.723 0.066 0.731 0.934 1.000 0.820
0.145 0.376 0.220 0.881 0.863 0.820 1.000
From this table one can judge the linear dependence of one random variable
on another. Consider
1,3
µ , the correlation coefficient between first story drift
and third story drift, in the log space. In this example,
1,3
0.019 µ = ÷ , which
indicates negligible dependence of
3
Y on
1
Y (and viceversa). Consider now
the linear dependence of the second and third floor and roof accelerations on
the ground acceleration,
4,5 4,6 4,7
, , µ µ µ , respectively. All values are close to
1.0, which indicates strong linear dependence.
Reader Note: It is not known at this time how many pairs of ground motions
are required for analysis to obtain reliable estimates of the entries in the
correlation coefficient matrix.
The correlation coefficient is closely related to the covariance, which is a
measure of how much two random variables vary together. (Variance is a
measure of how much a single random variable varies—see the last row of
Table G3 that presents the variance in
i
Y for all seven demand parameters).
The unit of covariance is the product of the units of the random variables.
The covariance of random variables
i
Y and
j
Y , cov( )
i j
Y , Y , is computed as
cov( ) ( )
i
E µ µ = · ÷ ·
j
i j i j Y Y
Y , Y Y Y (G4)
April 15, 2009
G8 G: Generation of Realizations for Loss Computations ATC58
where E is the expectation value operator and all other terms have been
defined previously. For a sample calculation, return to Table G3 and
compute the
1 2
cov( ) Y , Y as
1 2
1 2
1 2 1 2
1, 2,
1
cov( ) (( )( ))
1
( )( )
1
1
((0.231 0.179) (0.372 0.625)..
10
( 0.041 0.179) (0.604 0.625))
0.0123
n
i i
i
E
n
µ µ
µ µ
=
= ÷ ÷
= ÷ ÷
÷
= ÷ × ÷
+ ÷ ÷ × ÷
=
¿
Y Y
Y Y
Y , Y Y Y
Y Y
(G5)
The correlation coefficient,
1,2
µ , is calculated as
1,2
1 2 2
cov( ) 0.01226
0.339
cov( ) cov( ) 0.05910 0.02209
µ = = =
× ×
1 2
1
Y , Y
Y , Y Y , Y
(G6)
The matrix
Y
L is the Cholesky decomposition of
YY
R , namely,
T
YY Y Y
R = L L . Information on the Cholesky algorithm can be found in
textbooks. The lower triangular matrix
Y
L for the sample problem is given
in Table G6. (Note that the Matlab macro chol returns the upper triangular
matrix, the transpose of which is presented in Table G6.)
Table G6 Matrix
Y
L for the Sample Problem
1.000 0.000 0.000 0.000 0.000 0.000 0.000
0.339 0.941 0.000 0.000 0.000 0.000 0.000
0.019 0.704 0.710 0.000 0.000 0.000 0.000
0.375 0.510 0.708 0.314 0.000 0.000 0.000
0.022 0.678 0.540 0.377 0.325 0.000 0.000
0.193 0.699 0.594 0.085 0.315 0.120 0.000
0.145 0.451 0.761 0.185 0.242 0.101 0.305
All of the matrices required to generate the correlated vectors have been
constructed. The thirdtolast step in the process is to compute the 7×1 vector
U of uncorrelated standard normal random variables, with a mean of 0 and a
standard deviation of 1 for each random variable. The randn function in
Matlab is used for this purpose. This process is repeated 199 times to
construct the 200 realizations for loss assessment. The next step involves
taking the exponential of each value in the 7×200 matrix to recover the
demand parameters. Table G7 presents the first 10 realizations of the 200.
As a last step, the statistics of the simulated demands should be compared
with those of the demands calculated by responsehistory analysis. Figures
April, 2009
ATC58 G: Generation of Realizations for Loss Computations G9
G2 and G3 enable a partial comparison of demands. A rigorous comparison
of the original and simulated demands is achieved by computing the ratios of
the mean vectors and the correlation coefficient matrices. Table G8 presents
ratios of mean simulated response to mean responsehistory response—all
values are close to 1.0. Table G9 presents ratios of the simulated to
responsehistory correlation coefficients—all values are close to 1.0 for those
correlation coefficients in Table G5 with an absolute value of greater than
0.5. The data of Tables G8 and G9 indicate that the simulated vectors of
demand have the same underlying statistics as those of the original demand
vectors.
Table G7 Matrix of Simulated Demand Parameters (first 10 vectors)
1
o (%)
2
o (%)
3
o (%)
1
a (g)
2
a (g)
3
a (g)
r
a (g)
1 1.361 2.150 2.361 0.469 0.716 0.681 0.620
2 2.009 2.725 2.640 0.467 0.607 0.504 0.633
3 1.303 2.011 2.080 0.474 0.716 0.564 0.623
4 1.265 1.676 2.060 0.695 1.435 1.136 0.826
5 1.635 2.802 2.985 0.577 0.744 0.500 0.684
6 1.192 1.661 2.474 0.901 1.593 1.347 0.955
7 0.865 1.817 2.328 0.460 0.970 0.965 0.661
8 1.045 2.038 2.830 0.612 0.926 0.801 0.852
9 1.437 2.305 2.424 0.461 0.634 0.494 0.609
10 0.958 1.769 1.973 0.408 0.747 0.686 0.648
Table G8 Ratio of Simulated to Original Logarithmic Means
1
o
2
o
3
o
1
a
2
a
3
a
r
a
0.9462 0.9870 0.9822 1.0231 1.0715 1.0559 1.0220
Table G9 Ratio Of Entries in Simulated and Original
YY
R Matrices
1.000 0.987 2.128 0.761 2.205 1.161 1.265
0.987 1.000 0.908 1.338 1.077 1.084 1.146
2.128 0.908 1.000 0.916 0.783 1.119 0.926
0.761 1.338 0.916 1.000 1.047 1.069 1.045
2.205 1.077 0.783 1.047 1.000 1.007 1.000
1.161 1.084 1.119 1.069 1.007 1.000 0.964
1.265 1.146 0.926 1.045 1.000 0.964 1.000
April 15, 2009
G10 G: Generation of Realizations for Loss Computations ATC58
G.2.4 Matlab Code
The Matlab macro below can be used to generate correlated vectors of
demand parameters. The vectors of demand parameters established by
responsehistory analysis, DP.txt file, should be constructed per Table G1,
namely, one demand parameter per column, one simulation per row. The file
beta.txt includes a 3×n matrix where the values in each of the 3 rows are for
a
 (
a

A
or
aa
 ),
m
 and
gm
 , respectively, and those in each column are
for each demand parameter in the same order the demand parameters are
input to PACT. The file n_rows.txt includes the number of the simulated row
vectors, which is taken as 200 for the example of Section G.2.3.3. This
Matlab code assumes that dispersions input by the analyst are used in lieu of
those computed using the results of nonlinear analysis.
% Develop underlying statistics of the responsehistory analysis
X=load(DP.txt');
B=load(‘beta.txt’);%matrix for dispersions
nr=load(‘n_rows.txt’);%number of the generated row vectors
Y=log(X);
My=(mean(Y))';
n =length(My);
Dy=diag(sqrt(sum(B.*B)));
Ryy=corrcoef(Y);
Ly=(chol(Ryy))';
% Generate correlated demand vectors using a MonteCarlo technique
U=randn(nr,n);
My=diag(My)*ones(n,nr);
Z=Dy*Ly*U'+My;
W=exp(Z');
save correlated_DP.txt W ascii double tabs;
%Check results
Mz=(mean(Z'))';
Rzz=corrcoef(Z');
Mz./(mean(Y))';% for Table G8
Rzz./Ryy; % for Table G9
% end
For intensity and timebased assessments and for scenariobased
assessments using Appendix E.8, the user should input for each demand
parameter dispersions for analysis and modeling and set the dispersion for
ground motion equal to zero. For scenariobased assessment using the
April, 2009
ATC58 G: Generation of Realizations for Loss Computations G11
procedures of Chapter 5, the user should input for each demand parameter
dispersions for analysis, modeling and ground motion.
G.3 Realizations for Assessment Using Simplified
Nonlinear Analysis
For assessment using the simplified method of analysis per Chapter 6, PACT
generates internally a set of 500 statistically consistent demand vectors, given
the column vector of demand parameters, { } u and a value of the dispersion,
 , using the following twostep procedure: 1) randomly generate 500 values
of z, where z is normally distributed with a mean of 0 and a standard
deviation of 1.0; and 2) compute { }exp( ) z u  for each value of z. Each
vector computed using step 2 is used to generate one value of loss.
April 15, 2009
ATC58 H: Residual Drift H1
Appendix H
Residual Drift
H.1 Introduction
Experience from past earthquakes suggests that residual (permanent) drift is
one of the determining factors as to whether a structure can be economically
repaired and whether it is deemed unsafe for occupancy (or redtagged) after
an earthquake. Modest amounts of residual drift may be acceptable from the
standpoint of structural performance but may require adjustments to
nonstructural components (e.g., replumbing of elevator rails, adjustments to
building facades) and may trigger a redtag during postearthquake
inspections, which will have a significant impact on the building downtime.
Larger residual drifts may require either straightening of the structural frame
or alternative measures to strengthen the frame for stability. Finally, at some
stage, residual drifts may be large enough to seriously jeopardize structural
stability to earthquake aftershocks and uneconomical to repair, in which case
the cost of repair is on par with complete building replacement. Therefore,
residual drift should be considered in a building performance assessment.
Research has shown that residual drifts predicted by nonlinear analysis are
highly variable and sensitive to the assumed modeling features including the
postyield hardening/softening slope and unloading response and that
residual drift may be larger under nearfault records with forward directivity
pulses or under longduration records that cause many cycles of significant
response. Accurate statistical simulation of residual drifts therefore requires
advanced nonlinear responsehistory analyses, with a large number of ground
motions and with careful attention paid to cyclic hysteretic response of the
models and numerical accuracy of the solution. Since the requirements for
direct simulation of residual drifts are likely computationally expensive,
equations are provided to estimate residual drifts as a function of the peak
transient response.
In addition to providing methods to estimate the residual drifts, it is equally
important to have damage and loss models to quantify the implications of
residual drifts. One of the key considerations in this regard will be to
distinguish between repairs (or other consequences) associated with residual
drifts that are not otherwise accounted for damage and loss assessment in
April 15, 2009
H2 H: Residual Drift ATC58
other components. For example, consider the case of a steel braced frame
that experiences brace buckling or fracture. Costs associated with repair to
the building are likely to vary tremendously whether or not there are residual
drifts present.
Past studies on the residual story drifts are presented in Section H.2. Section
H.3 presents a simplified procedure for computing residual story drifts.
Damage states associated with residual story drift are proposed in Section
H.4.
H.2 Prediction of Residual Story Drifts
Analytical studies by Riddell and Newmark (1979) and Mahin and Bertero
(1981) first identified some of the key behavioral aspects associated with
calculating residual drifts. More recent studies (including MacRae and
Kawashima 1997, MacRae 1994, Christopoulos et al. 2003, Christopoulos
and Pampanin 2004, Pampanin et al. 2003, and RuizGarcia and Miranda
2005, 2006ac) have confirmed that residual drifts are largely dependent on
the following parameters: (1) magnitude of peak transient inelastic drifts, (2)
lateral strength of the structure relative to the earthquake demand, (3)
inelastic postyield stiffness, (4) cyclic unloading response,(5) pulses in the
ground motions, and (6) duration of ground shaking. In many respects these
parameters are all interrelated and affect transient and residual drifts.
Assuming that the calculation of residual story drifts will be conditioned on
peak transient story drifts, some of these factors will be implicitly accounted
for in the peak transient drift assessment. In terms of implications for
performance assessment, it is desirable that the models account for variations
in postyield stiffness and the cyclic unloading response since these
parameters are structural parameters that can be influenced by design. For
example, the residual drift assessment should differentiate between
conventional structures and ones that employ selfcentering features that are
designed to minimize residual drifts.
Shown in Figure H1 is a schematic plot relating the peak transient and
residual story drift under increasing earthquake intensity. Point “a”
corresponds to the onset of inelastic response, prior to which residual drifts
cannot occur. Point “c” corresponds to the collapse point, where residual and
transient response converges and point “b” corresponds to regions between
these two limits, where the residual drifts represent some fraction of the peak
transient drift. As this figure clearly indicates, the ratio of residual to
transient drifts is not constant but it may be reasonable to treat the ratio as a
constant. Much of the existing literature has evaluated residual drifts at
intermediate levels of response, sometimes distinguishing between the degree
April 15, 2009
ATC58 H: Residual Drift H3
of inelasticity by varying the strength of the structure, either by adjusting the
strength using a response modification factor or by adjusting the transient
story ductility demand.
Figure H1 Idealized incremental dynamic analysis presenting transient and
residual story drift ratios
One of the significant behavioral effects that impact residual displacements is
the unloading characteristics of a structure. The load versus deflection plots
of Figure H2 represents the basic aspects of unloading response that have
been identified as important in prior research. These studies have
consistently shown that residual displacements are largest in elastic plastic
(EP) systems, due to their tendency to preserve the inelastic displacements
upon unloading. This is in contrast to general inelastic (GI) response, where
pinching or other effects lead to smaller unloading stiffness. Most recently,
researchers have been examining the benefits of socalled selfcentering (SC)
systems that are specifically designed with unloading characteristics that are
intended to minimize (or potentially eliminate) residual displacements.
There are also behavioral effects, such as foundation rocking, which have
natural selfcentering tendencies.
Figure H2 Idealized response characteristics for elasticplastic (EP), general
inelastic (GI) and selfcentering (SC) systems
Peak Interstory Drift Ratio
G
M
I
n
t
e
n
s
i
t
y
Transient
Residual
a
b
c
Interstory Drift
S
h
e
a
r
F
o
r
c
e
EP GI
SC
April 15, 2009
H4 H: Residual Drift ATC58
Due to the variety of approaches and parameters addressed by various
researchers, it is difficult to generalize the results of previous research.
However, in the interest of developing a simplified model, the following
discussion summarizes some overall results and observations from published
analytical studies.
MacRae and Kawashima (1997) investigated residual displacements in single
degreeoffreedom bilinear oscillators with a focus on bridge design, where
the presence of residual displacements was cited as a major factor in bridge
closures and replacement. They proposed a model to estimate residual
displacements as a function of the peak inelastic displacement, unloading
properties of the inelastic oscillator, and a statistical coefficient based on
inelastic time history analyses. In an illustrative example, they predicted
residual displacements equal to about 0.16 times the maximum transient
displacement. According to their proposed model, the residual displacement
ratio could be disaggregated into the following product, 0.16 =0.67 x 0.9 x
0.27, where 0.67 is the ratio of plastic to total drift (equal to the sum of the
elastic and plastic drift), 0.9 reflects the effect of the reduced unloading
stiffness, and 0.27 is the statistical parameter to take account of inelastic
dynamic response. The specific value of 0.27 represents the mean plus
standard deviation from a range of analyses for system with modest strain
hardening and a ductility demand ratio of 4.
Christopoulos et al. (2003) conducted analyses of singledegree of freedom
models with the primary emphasis on characterizing the effect of postyield
hardening stiffness and hysteretic unloading behavior on response. The
models were generally representative of modern ductile moment frame
buildings, ranging in height from 4 to 20 stories designed with base shear
yield ratios ( /
y
V W ) ranging from 0.26 to 0.11, respectively. They applied
ground motions to represent both design and maximum earthquakes.
Maximum drift ratios observed in their analyses were in the range of 2% to
3%. Analyses of elastic plastic (EP) models with zero hardening had residual
displacements ranging from 0.1 to 0.4 times the peak inelastic transient
displacements. These ratios reduced to 0.1 to 0.2 for systems with 5% strain
hardening and increased up to 0.7 for systems with 5% strain softening (5%
hardening). The largest residual displacements and peak transient ductility
(up to 5 times yield) were observed in the 4story (period approximately
equal to1 second) building with strain softening. Comparable analyses run
with a general inelastic model (the Takeda model) had significantly smaller
residual drifts, ranging between 7 to 12% of the peak transient drifts for
systems with zero or positive strain hardening and 10 to 15% for systems
with strain softening. In related studies of multidegreeoffreedom models
April 15, 2009
ATC58 H: Residual Drift H5
(Pampanin et al. 2003) the researchers reported residual drifts equal to 0.05
to 0.25 times the total transient drift, depending on the hardening and
hysteretic model.
RuizGarcia and Miranda (2005, 2006ac) conducted an extensive parametric
study of single and multi degreeoffreedom systems with the goal to
characterize probabilistic residual drift hazard curves. They generally
expressed their results in terms of a parameter,
r
C , which related residual
drifts to elastic spectral displacement. Thus, for consistency with the
previous studies, this parameter would need to be modified to account for the
difference between peak inelastic transient displacement and the elastic
spectral displacement. Presumably, this adjustment is close to unity where
the equal displacement rule applies (long period single mode dominated
systems under small to moderate displacements). For elasticplastic models
with periods greater than 1 second and elastic demand to yield strength ratios
of R =1.5 to 6, they reported median values of
r
C =0.25 to 0.5. This ratio
increased to
r
C =0.8 for oscillators with period between 0.5 and 1 second,
and became even larger for smaller periods. Since the ratio is conditioned on
elastic spectral displacement, the large amplification at small periods
probably reflects the amplification between elastic and inelastic transient
displacements. For stiffness degrading systems, the median value coefficient
reduced to
r
C =0.1 to 0.3. They also reported dispersion in
r
C of about 0.8,
which was fairly constant across various periods and building strengths.
H.3 Model to Calculate Residual Story Drifts
Prior studies have indicated that for ductility demands of about 4, residual
drifts were on the order of 0.15 to 0.30 times the peak transient drifts. Using
this as a guide, a set of equations are proposed to relate residual story
drift,
r
, to peak transient story drift, :
0
0.3( ) 4
( 3.1 ) 4
r y
r y y y
r y y
(H1)
where
y
is the story drift at yield. Figure H3 illustrates these equations.
April 15, 2009
H6 H: Residual Drift ATC58
Figure H3 Idealized model to estimate residual story drift from peak
transient drift
These equations are calibrated such that at a story ductility ratio of 4, the
ratio of residual to peak transient drift is 0.23. At ductility ratios of 2 and 6,
the ratios of residual to peak transient drift are 0.15 and 0.48, respectively.
Ultimately, as the collapse point is reached the residual drift will approach
the peak transient drift. As an example, consider a steel moment frame with
a yield drift ratio of 1%. At 3% total drift, the predicted residual drift ratio
would be 0.6%, or about 0.20 times the peak transient. At 5% total drift, the
predicted residual would be about 1.9%, or about 0.38 times the peak
transient. At 7% total drift (approaching the collapse point), the residual drift
would be about 3.9%, or 0.56 times the peak transient.
The proposed equations apply to typical building systems and do not
distinguish between many of the behavioral effects that are known to affect
residual drifts. For example, whereas previous studies of idealized models
has shown a correlation between postyield stiffness and residual drifts, this
is not reflected in the equations except insofar as the postyield stiffness
affects the peak transient drift ratio. The equations are presented in this
simple format owing to the lack of physical data to validate modeling of
residual displacements and due to the complexity of obtaining significant
improvements in the residual drift estimates.
Per RuizGarcia and Miranda, the dispersion for residual drift estimates is
about 0.8. Two alternative ways to consider this in the assessment would be
to either (1) calculate the median residual drift demand fragility using
equation H1 based on the median transient and yield drift ratios, and then
apply a dispersion of 0.8, or (2) calculate the residual drift probability based
on the integration of the conditional probability of residual drift and the
Peak Interstory Drift Ratio
G
M
I
n
t
e
n
s
i
t
y
Transient
Residual
=1
IDR
y
=4
=inf
April 15, 2009
ATC58 H: Residual Drift H7
transient drift. In this case, the dispersion on the conditional probability of
residual drift should be reduced by the recordtorecord variability in the
transient drift.
H.4 Proposed Damage States for Residual Story Drifts
Four damage states are proposed in Table H1, ranging from the onset of
damage to nonstructural components to nearcollapse. The proposed median
story drift ratios (residual drift as a percentage of story height) are
approximate and based on a combination of judgment and limits that have
been previously suggested in FEMA 356 (Table C13). For damage state
DS4, the limits on the residual drifts that are expressed in terms of the design
shear force are intended to cover cases in regions of low seismicity where the
destabilizing effects of P may dominate at smaller drift ratios than the
values noted.
Further work remains to establish appropriate parameters to describe the
uncertainty in these limits and loss functions to define the direct economic
loss and downtime losses associated with these damage states. For example,
damage state DS1 may require repairs to nonstructural components that may
not otherwise be damaged, except for the need to realign components.
Similarly, depending on the type of system, damage state DS2 requires
extensive structural repairs beyond those evident from the assessment of
structural component damage.
The relationship between transient and residual drift demands, per H1, can
be combined with the information in Table H1 to relate these damage states
to the associated peak transient drifts. Through the yield drift ratio, the
relationship between the residual drift damage states and peak transient drift
will depend on the stiffness of the system. A few example relationships are
summarized in Table H2, which are based on the assumed yield drift ratios
noted in the table.
April 15, 2009
H8 H: Residual Drift ATC58
Table H1 Damage states for residual story drifts
Damage
state
Description
Residual story drift
ratio, /
r
h
1
DS1
No structural realignment is necessary for
structural stability; however, the building
may require adjustment and repairs to
nonstructural and mechanical
components that are sensitive to building
plumb (e.g., elevator rails, curtain wall,
doors in full height partitions).
0.2%
(equal to the maximum
out of plumb typically
permitted in new
construction)
DS2
Realignment of structural frame and
related structural repairs required to both
maintain permissible drift limits for
nonstructural and mechanical
components and to limit degradation in
structural stability (collapse safety)
0.5%
DS3
Major structural realignment is required
to restore safety margin for lateral
stability; however, the required
realignment and repair of the structure
may not be economically and practically
feasible (i.e., the structure might be at a
total loss from the standpoint of direct
economic losses).
1%
DS4
Residual drift is sufficiently large that the
structure is in danger of collapse from
earthquake aftershocks. For this reason
the performance point might be
considered as equal to collapse, albeit
with greater uncertainty.
High Ductility Systems
4% < 0.5V
design
/W
Moderate Ductility
Systems
2% < 0.5V
design
/W
Limited Ductility Systems
1% < 0.5V
design
/W
1. h is the story height
April 15, 2009
ATC58 H: Residual Drift H9
Table H2 Sample transient story drift ratios, / h , associated with the
residual story drift damage states of Table H1
/ h
Sample framing
system
/
y
h
DS1
1
DS2
1
DS3
1
DS4
2
Steel ductile
moment resisting
frame
1% 1.5% 2.7% 4.1% 7.1%
Reinforced
concrete wall
0.5% 1% 2.2% 2.6% 3.6%
Timber shear wall 1% 1.5% 2.7% 4.1% 5.1%
1. /
r
h =0.2%, 0.5% and 1.0% for DS1, DS2 and DS3, respectively.
2. /
r
h =4%, 2% and 2% for steel moment resisting frames, reinforced
concrete walls, and timber shear walls, respectively..
April 15, 2009
ATC58 I: References I1
References
ATC (2008), Interim Guidelines on Modeling and Acceptance Criteria for
Seismic Design and Analysis of Tall Buildings, ATC721 Project Report
– 90% Draft, Published by the Applied Technology Council for the
Pacific Earthquake Engineering Research Center Tall Buildings Initiative
– Task 7 Project Core Group, www.atcouncil.org.
Christopoulos, C., Pampanin, S., Priestley, M.J .N. (2003), “Performance
Based Seismic Response of Frame Structures Including Residual
Deformations. Part I: SingleDegree of Freedom Systems”, J ournal of
Earthquake Engineering, 7(1), pg. 97118.
Christopoulos, C. and Pampanin, S., 2004. “Towards PerformanceBased
Design of MDOF Structures with Explicit Consideration of Residual
Deformations”, Invited Paper, Indian Society for Earthquake
Technologies (ISET) Journal of Earthquake Technology, Special Issue
on PerformanceBased Design, Vol. 41, No. pp. 172193.
FEMA 445 (2006), NextGeneration PerformanceBased Seismic Design
Guidelines, Program Plan for New and Existing Buildings, published
by the Applied Technology Council for the Federal Emergency
Management Agency, Washington, D.C.
FEMA 461 (2007), Interim Testing Protocols for Determining the Seismic
Performance Characteristics of Structural and Nonstructural
Components, published by the Applied Technology Council for the
Federal Emergency Management Agency, Washington, D.C.
FEMA P695 (2008), Quantification of Building Seismic Performance
Factors, ATC63 Project Report – 90% Draft, published by the
Applied Technology Council for the Federal Emergency
Management Agency, www.atcouncil.org.
Haselton, C.B., Deierlein, G.G. (2007), Assessing Seismic Collapse Safety of
Modern Reinforced Concrete MomentFrame Buildings, PEER 2007/08,
www.peer.berkeley.edu.
Ibarra, L.F., Medina, R.A., and Krawinkler, H. (2005). “Hysteretic models
that incorporate strength and stiffness deterioration,” International
April 15, 2009
I2 I: References ATC58
Journal for Earthquake Engineering and Structural Dynamics, Vol. 34,
No.12, pp. 14891511.
Ibarra, L.F., and Krawinkler, H. (2005). Global Collapse of Frame Structures
Under Seismic Excitations, PEER TR 2005/06, www.peer.berkeley.edu.
Liel, A.B., Haselton, C.B., Deierlein, G.G., Baker, J .W. (2009),
“Incorporating modeling uncertainties in the assessment of seismic
collapse risk of buildings,” Structural Safety, 31(2), pp. 197211.
Lignos, D.G. (2008), Sidesway Collapse of Deteriorating Structural Systems,
Ph.D Dissertation, Department of Civil and Environmental Engineering,
Stanford University, Stanford California.
MacRae, G.A., Kawashima, K. (1997), “PostEarthquake Residual
Displacements of Bilinear Oscillators,” EESD, 26, pp. 701719.
MacRae G. A., (1994) "PD Effects on SingleDegreeofFreedom Structures
in Earthquakes", Earthquake Spectra, Vol. 10, No. 3, August 1994, pp.
539568
Mahin, S.A., Bertero, V.V. (1981), “Evaluation of Inelastic Seismic Design
Spectra,” J . Struct. Div., ASCE, 107(9), pp 17771795.
Moehle, J ., and Deierlein, G., 2004, A Framework Methodology for
PerformanceBased Earthquake Engineering, Proceedings, 13
th
World
Conference on Earthquake Engineering, Vancouver, British Columbia,
paper no. 679.
NIST (1999), UNIFORMAT II Elemental Classification for Building
Specifications, Cost Estimating, and Cost Analysis, NISTIR 6389,
National Institute of Standards and Technology, Gaithersberg, Maryland.
Pampanin, S., Christopoulos, C., Priestley, M.J .N. (2003), “Performance
Based Seismic Response of Frame Structures Including Residual
Deformations. Part II: MultiDegree of Freedom Systems”, J ournal of
Earthquake Engineering, 7(1), pg. 119147.
Riddell, R., Newmark, N. M. (1979), Statistic Analysis of the Response of
Nonlinear Systems Subjected to Earthquakes, Structural Report Series
468, Civil Engineering, University of Illinois.
RuizGarcia, J ., Miranda, E. (2005), “PerformanceBased Assessment of
Existing Structures Accounting for Residual Displacements”, Technical
Report No. 153, J ohn A. Blume Earthquake Engineering Center,
http://blume.stanford.edu, Stanford University, 444 pgs.
April 15, 2009
ATC58 I: References I3
RuizGarcia, J ., Miranda, E. (2006a), “Direct Estimation of Residual
Displacement From Displacement Spectral Ordinates”, Paper No. 1101,
Proceedings of 8NCEE, 10 pgs.
RuizGarcia J , Miranda E., (2006b), “Residual displacement ratios for
assessment of existing structures,” EESD, 35(3):315336.
RuizGarcia J , Miranda E., (2006c), “Evaluation of residual drift demands in
regular multistory frames for performancebased seismic assessment”.
EESD, 35(13)16091629.
Yang, T.Y., Moehle, J ., Stojadinovic, B., Der Kiureghian, A., 2006, An
Application of PEER PerformanceBased Earthquake Engineering
Methodology, Proceedings, 8
th
US National Conference on Earthquake
Engineering, San Francisco, California.
Zareian, F., Krawinkler, H. (2007) “Assessment of probability of collapse
and design for collapse safety”, Earthquake Engineering and Structural
Dynamics, 36(13), 19011914.
April 15, 2009
ATC58 J: Project Participants J1
Project Participants
Project Management
Christopher Rojahn
Project Executive Director
Applied Technology Council
201 Redwood Shores Parkway, Suite 240
Redwood City, CA 94065
Ronald O. Hamburger
Project Technical Director
Simpson Gumpertz & Heger
The Landmark @ One Market, Suite 600
San Francisco, CA 94105
FEMA Oversight
Mike Mahoney
Project Officer
Federal Emergency Management Agency
500 C Street, SW, Room 416
Washington, DC 20472
Robert D. Hanson
Technical Monitor
Federal Emergency Management Agency
2926 Saklan Indian Drive
Walnut Creek, CA 945953911
Project Management Committee
Christopher Rojahn, Chair
Ronald Hamburger, CoChair
J on A. Heintz (exofficio)
Applied Technology Council
201 Redwood Shores Parkway, Suite 240
Redwood City, CA 94065
William T. Holmes (exofficio)
Rutherford & Chekene
55 Second Street, Suite 600
San Francisco, CA 94105
J ohn Gillengerten
Office of Statewide Health
Planning and Development
1600 9th St., Room 420
Sacramento, CA 95814
Peter J . May
University of Washington
3630 Evergreen Point Road
Medina, WA 98039
J ack P. Moehle
University of California at Berkeley
3444 Echo Springs Road
Lafayette, CA 94549
Maryann T. Phipps (ATC Board Contact)
Estructure
8331 Kent Court, Suite 100
El Cerrito, CA 94530
April 15, 2009
J2 J: Project Participants ATC58
Steering Committee
William T. Holmes, Chair
Rutherford & Chekene
55 Second Street, Suite 600
San Francisco, CA 94105
Roger D. Borcherdt
U.S. Geological Survey
345 Middlefield Road, MS977
Menlo Park, CA 94025
Anne Bostrom
University of Washington
Parrington Hall, Room 327
Seattle, WA 981953055
Bruce Burr
Burr & Cole Consulting Engineers
3485 Poplar Avenue, Suite 200
Memphis, TN 38111
Kelly Cobeen
Wiss, J anney, Elstner Associates, Inc.
2200 Powell Street, Suite 925
Emeryville, CA 94608
Anthony B. Court
A. B. Court & Associates
4340 Hawk Street
San Diego, CA 92103
Terry Dooley
Morley Builders
2901 28th Street, Suite 100
Santa Monica, CA 90405
Dan Gramer
Turner Construction Company
830 4th Avenue South, Suite 400
Seattle, WA 98134
Michael Griffin
CCS Group, Inc.
1415 Elbridge Payne Road, Suite 265
Chesterfield, MO 63017
R. J ay Love
Degenkolb Engineers
300 Frank H Ogawa Plaza, Suite 450
Oakland, CA 94612
David Mar
TippingMar & Associates
1906 Shattuck Avenue
Berkeley, CA 94704
Steven McCabe
NEES Consortium, Inc.
400 F Street
Davis, CA 95616
Brian J . Meacham
Department of Fire Protection Engineering
Worcester Polytechnic Institute
100 Institute Road
Worcester, MA 016092280
William J . Petak
University of Southern California
School of Policy Planning and Development
MC 0626 Los Angeles, CA 90089062
April 15, 2009
ATC58 J: Project Participants J3
Nonstructural Performance Products Team
Robert E. Bachman, Team Leader
Consulting Structural Engineer
25152 La Estrada Drive
Laguna Niguel, CA 92677
Philip J . Caldwell
Square D/Schneider Electric
7 Sleepy Hollow Lane
Six Mile, SC 29682
Andre Filiatrault
MCEER, University at Buffalo
105 Red J acket Quadrangle
Buffalo, NY 142610025
Robert P. Kennedy
RPK Structural Mechanics Consulting, Inc.
28625 Mountain Meadow Road
Escondido, CA 92026
Helmut Krawinkler
Stanford University
Dept. of Civil & Environmental Eng.
Stanford, CA 943054020
Manos Maragakis
University of Nevada, Reno
Department of Civil Engineering
Mail Stop 258
Reno, NV 895570152
Eduardo Miranda
Stanford University
Civil & Environmental Engineering
Terman Room 293
Stanford, CA 943053707
Keith Porter
University of Colorado
Dept of Civil, Environmental & Architectural
Engineering
Engineering Center Office Tower 428
1111 Engineering Drive
Boulder, CO 8030904
Nonstructural Fragility Development Consultants
Richard Behr
Pennsylvania State University
104 Engineering Unit A
University Park, PA 168021416
J ohn Eidinger
G&E Engineering Systems
PO Box 3592
Olympic Valley, CA 96146
Paul Kremer
430 Canterbury Drive
State College, PA 16802
Ali M. Memari
Pennsylvania State University
104 Engineering Unit A
University Park, PA 168021416
William O'Brien
The Pointe
501 Vairo Boulevard
State College, PA 16803
J ohn Osteraas
Exponent
149 Commonwealth Drive
Menlo Park, CA 94025
Xin Xu
7 Xibei Xincun, Room 403
Wuxi, J iangsu 214062, P.R.China
April 15, 2009
J4 J: Project Participants ATC58
Risk Management Products Team
Craig D. Comartin, Team Leader
Comartin Engineers
7683 Andrea Avenue
Stockton, CA 952071705
Mary Comerio
University of California at Berkeley
Department of Architecture
232 Wurster Hall
Berkeley, CA 947201800
Gregory Fenves
University of Texas at Austin
Cockrell School of Engineering
Austin, TX 78758
Mahmoud Hachem
Skidmore Owings & Merrill LLP
One Front Street
San Francisco, CA 94111
Gee Heckscher
Architectural Resources Group
1301 53rd St
Port Townsend, WA 98368
J udith MitraniReiser
J ohns Hopkins University
Department of Civil Engineering
3400 N. Charles St., Latrobe 109
Baltimore, MD 21218
Farzad Naeim
J ohn A. Martin and Associates, Inc.
950 S. Grand Ave. 4
th
Fl.
Los Angeles, CA 90015
Keith Porter
University of Colorado
Dept of Civil, Environmental & Architectural
Engineering
Engineering Center Office Tower 428
1111 Engineering Drive
Boulder, CO 803090428
Hope Seligson
MMI Engineering
2100 Main Street, Suite 150
Huntington Beach, CA 92648
Risk Management Products Team Consultants
Peter Morris
Davis Langdon
1331 Garden Highway
Suite 310
Sacramento, CA 95833
Scott Shell
Esherick Homsey Dodge & Davis Architects
500 Treat Avenue, Suite 201
San Francisco, CA 94110
April 15, 2009
ATC58 J: Project Participants J5
Structural Performance Products Team
Andrew S. Whittaker, Team Leader
University at Buffalo
Department of Civil Engineering
230 Ketter Hall
Buffalo, NY 14260
Gregory Deierlein
Stanford University
Dept. of Civil & Environmental Eng.
240 Terman Engineering Center
Stanford, CA 943054020
J ohn D. Hooper
Magnusson Klemencic Associates
1301 Fifth Avenue, Suite 3200
Seattle, WA 98101
YinNan Huang
127 Paramount Pkwy
Buffalo, NY 14223
Nicolas Luco
U. S. Geological Survey
Box 25046 – DFC – MS 966
Denver, CO 80225
Andrew T. Merovich
A. T. Merovich & Associates, Inc.
1950 Addison Street, Suite 205
Berkeley, CA 94704
Structural Fragility Development Consultants
Charles Ekiert
12 Hillcrest Drive
Hamlin, NY 14464
Andre Filiatrault
MCEER, University at Buffalo
105 Red J acket Quadrangle
Buffalo, NY 142610025
Aysegul Gogus
Department of Civil Engineering
University of California, Los Angeles
Los Angeles, CA 900951593
Kerem Gulec
University at Buffalo
Department of Civil Engineering
Buffalo, NY 14260
Dawn Lehman
Department of Civil and Environmental
Engineering
214B More Hall
University of Washington
Seattle, WA 981952700
J ingjuan Li
4747 30th Ave NE B118
Seattle, WA, 98105
Laura Lowes
Department of Civil & Environmental Engineering
University of Washington
Box 352700
Seattle, WA 981952700
Eric Lumpkin
6353 NE Radford Drive Apt. #3917
Seattle, WA 98115
Hussein Okail
University of California San Diego
Department of Structural Engineering
409 University Center
9500 Gilman Dr. – MC0085
La J olla, CA 92093
April 15, 2009
J6 J: Project Participants ATC58
Charles Roeder
Department of Civil and Environmental
Engineering
233B More Hall
University of Washington
Seattle, WA 981952700
P. Benson Shing
University of California San Diego
Department of Structural Engineering
409 University Center
9500 Gilman Dr. – MC0085
La J olla, CA 92093
Christopher Smith
Department of Civil Engineering
MC:4020, Building 540
Stanford, CA 943054020
Victor Victorsson
1742 Sand Hill Road, Apt. 308
Palo Alto, CA 94304
J ohn Wallace
Department of Civil Engineering
University of California, Los Angeles
5731 Boelter Hall
Los Angeles, CA 900951593
Fragility Review Panel
Bruce R. Ellingwood
School of Civil and Environmental Eng.
Georgia Institute of Technology
790 Atlantic Drive
Atlanta, GA 303320355
Robert P. Kennedy
RPK Structural Mechanics Consulting, Inc.
28625 Mountain Meadow Road
Escondido, CA 92026
Stephen Mahin
Department of Civil Engineering
777 Davis Hall, MC 1710
University of California
Berkeley, CA 94720
Validation/Verification Team
J ack Baker
Stanford University
Terman Engineering Center, Room 240
Stanford, CA 943054020
David Bonneville
Degenkolb Engineers
235 Montgomery Street, Suite 500
San Francisco, CA 941042908
Charles Scawthorn
C. Scawthorn & Associates
744 Creston Road
Berkeley, CA 94708
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.