You are on page 1of 38

CHAPTER C32 TORNADO LOADS

2  C32.1 PROCEDURES

3  C32.1.1 Scope. Historically, tornadoes kill more people per year in the United States than
4  hurricanes and earthquakes combined (NIST 2014), and the average annual insured catastrophe
5  losses for events involving tornadoes exceed those for hurricanes and tropical storms combined
6  (Insurance Information Institute 2020). Despite these statistics, building codes and standards

LY
7  have not previously included tornado hazards in their scope except for specialized applications,
8  such as for storm shelters and nuclear facilities. Recent research on tornado climatology has

N
9  shown that tornadoes occur with much greater frequency and intensity than had previously been

O
10  quantified (Twisdale et al. 2021). The requirements introduced in Chapter 32 in the 2022 edition,
11  along with corresponding changes in Chapters 1, 2, and 26, provide for determination of tornado
12  n
loads on the main wind force resisting system (MWFRS) and components and cladding (C&C)
io
13  and incorporation of those loads into the design process.
at

14  While tornadoes are a type of windstorm, and many of the tornado load procedures are similar to
m

15  those in Chapters 26 through 31, there are significantly different characteristics between
tornadoes and other windstorms. For instance, tornadic winds have significant vertical
or

16 
17  components; atmospheric pressure changes so rapidly that it can induce loads; and load
nf

18  combinations including tornadoes are not always the same as those including wind. Hence,
tornado loads are treated separately from wind loads, not as a subset of wind loads. This is
rI

19 
20  analogous in some ways to the separate treatment of flood loads and tsunami loads.
Fo

21  Determining When Design for Tornado Loads Is Required. Under this standard, most
22  buildings and other structures are not required to be designed for tornado loads. A flowchart
23  describing the four-step process to determine whether tornado loads need to be checked is
24  provided in Figure 32.1-2. Risk Category I and II facilities are exempt from tornado load
25  requirements (Step 1), as are Risk Category III and IV buildings and other structures located
26  outside of the tornado-prone region shown in Figure 32.1-1 (Step 2). For buildings and structures
27  not already eliminated, the tornado speed, VT, is next determined in accordance with Section
28  32.5.1. If VT < 60 mi/h (26.8 m/s), design for tornado loads is not required per Section 32.5.2

5
1  (Step 3). Lastly, if VT is less than a certain percentage of V (as a function of exposure), where V
2  is the basic wind speed from Chapter 26, design for tornado loads is also not required (Step 4).
3  The rationales underlying these four criteria are described elsewhere in the Chapter 32
4  commentary: for Risk Category limitations, see Section C32.5.1; the Tornado-prone region is
5  covered later in this section; and limitations on VT in Steps 3 and 4 are addressed in Section
6  C35.5.2. While tornado loads must be checked for all buildings and other structures not meeting
7  any of these four exclusionary conditions, tornado loads will still not control in many cases. The
8  criteria used in Steps 3 and 4 represent the approximate threshold speeds at which tornado loads

LY
9  can begin to control some aspect of the wind load design, based on the worst-case combinations
10  of geographic location, exposure, effective plan area, mean roof height, enclosure classification,

N
11  building shape, and other parameters.

O
12  Tornado-Prone Region. The tornado-prone region (Figure 32.1-1) is approximately the same as
13  the portion of the conterminous United States that lies east of the Continental Divide. While
14  n
tornadoes can and do occur to the west of the tornado-prone region, tornado speeds associated
io
15  with Risk Category III and IV return periods in those locations are such that tornado loads would
at
16  not control over wind loads computed using Chapters 26 through 31. However, at longer return
17  periods, tornado speeds in the western United States do increase significantly (see Appendix G).
m

18  Tornadoes are rare in Alaska, Hawaii, and US territories, and when they do occur, are typically
or

19  weak. Therefore, those locations are not included in the tornado-prone region.
nf

20  The boundary of the tornado-prone region in Figure 32.1-1 corresponds to the 60 mi/h (26.8 m/s)
contour in Figure 32.5-2H, the longest return period (Risk Category IV, 3,000 years) map for the
rI

21 
22  largest effective plan area, 4 million ft2 (371,612 m2). This map has the greatest tornado speeds
Fo

23  extending over the largest area of the country; that is, the worst case is used to define the
24  tornado-prone region. The 60 mi/h (26.8 m/s) contour was chosen because it represents the
25  approximate threshold speed at which tornado loads can begin to control some aspect of the wind
26  load design.

27  Storm Shelters. Storm shelters and safe rooms are designed with the intent of providing an
28  enhanced level of life safety protection for a design storm event that is greater than typical for
29  conventional buildings and structures, including those designed to the tornado loads in this
30  chapter. A report (CDC 2012) on fatalities during the April 25–28, 2011, tornado outbreak

6
1  illustrates the importance of providing tornado protection for building occupants. There were 351
2  recorded tornadoes during the outbreak, where 338 fatalities were caused by 27 of the tornadoes.
3  According to the report, 90% of the people who died were in buildings at the time the tornadoes
4  struck. Similarly, during the May 22, 2011 tornado in Joplin Missouri, 135 (84%) of the 161
5  fatalities occurred inside buildings, with nearly all deaths caused by impact-related factors, such
6  as multiple blunt force trauma to the body (NIST 2014).

7  The design of storm shelters, including community and residential storm shelters for both
8  tornadoes and hurricanes, has been regulated by model building and residential codes since 2009.

LY
9  These codes all directly reference ICC 500, the ICC/NSSA Standard for the Design and
10  Construction of Storm Shelters (2020). The load provisions of ICC 500 directly reference ASCE

N
11  7, with some modifications including greater hazard levels. Beginning with the 2015

O
12  International Building Code (ICC 2014), certain types of critical facilities located in the 250 mi/h
13  (400 km/h) zone in Figure C32.1-1 have been required to include tornado shelters compliant with
14  n
ICC 500. This includes many Group E occupancies (i.e., K–12 schools), emergency operations
io
15  centers, and fire, rescue, ambulance, police, and 911 call stations. The 2018 International
at
16  Existing Building Code (ICC 2017) extended these requirements to include additions to certain
existing Group E occupancy buildings.
m

17 
or

18  FEMA also publishes guidance on the design and construction of structures for storm protection
19  in FEMA P-361, Safe Rooms for Tornadoes and Hurricanes: Guidance for Community and
nf

20  Residential Safe Rooms (2021a), and FEMA P-320, Taking Shelter from the Storm: Building or
rI

21  Installing a Safe Room for Your Home (2021b). FEMA P-361 provides comprehensive guidance
22  for the design of community and residential safe rooms, as well as operations and maintenance
Fo

23  guidance. While the criteria in FEMA P-361 and ICC 500 are quite similar, “safe room” is
24  FEMA terminology for storm shelters that also meet the criteria in FEMA P-361. All safe room
25  criteria in FEMA P-361 meet the storm shelter requirements in ICC 500, but a few design and
26  performance criteria in FEMA P-361 are more restrictive than those in ICC 500 and are required
27  for safe room projects to be eligible for federal grant funding. FEMA P-320 provides prescriptive
28  solutions for residential safe rooms.

7
1  Storm shelters and safe rooms are designed for more extreme hazard levels than conventional
2  buildings, including wind-borne debris. All elements of the tornado shelter envelope must resist
3  the impact of a 15 lb (6.8 kg) sawn lumber 2×4 missile traveling at speeds of 53 to 100 mi/h
4  (23.7 to 44.7 m/s), depending on the shelter design wind speed (Figure C32.1-1) and the
5  orientation of the shelter envelope surface (vertical or horizontal). In terms of tornado speed,
6  both ICC 500 and FEMA P-361 use the same map to determine design speeds for tornado
7  shelters and safe rooms. The map in Figure C32.1-1 was developed using a spatial analysis of
8  raw tornado climate data from 1950 to 2006. The 250 mi/h (400 km/h) zone represents the region

LY
9  of the country that most frequently experiences EF4 and EF5 tornadoes. The 200, 160, and 130
10  mi/h zones correspond to areas where the most intense tornadoes are usually EF3, EF2, and EF1,

N
11  respectively. The ICC 500 / FEMA P-361 tornado speeds and resultant tornado design pressures
12  and loads represent something approaching a worst-case scenario: “The design wind speeds

O
13  chosen by FEMA for safe room guidance were determined with the intent of specifying near-
14 
n
absolute protection with an emphasis on life safety” (FEMA 2021a). Given the deterministic
io
15  nature of the ICC 500 / FEMA P-361 tornado shelter design wind speed map, the annual
16  probability of exceedance (and its approximate reciprocal, the MRI) of these tornado speeds are
at

17  not uniform across the country and are therefore undefined. In contrast, the tornado speed maps
m

18  in Chapter 32 are probabilistic in nature (see Section C32.5.1). Tornado speeds for Risk
Category III and IV buildings are based on return periods of 1,700 and 3,000 years, respectively,
or

19 
20  two to three orders of magnitude smaller than the return periods estimated for the tornado shelter
nf

21  design speeds.


rI

22  The tornado loads specified in ICC 500 and FEMA P-361 are greater than ASCE 7 requirements
in Chapter 32, consistent with the differences in intended levels of life-safety protection and
Fo

23 
24  building performance. While ICC 500 does not explicitly state a performance objective for
25  shelters designed to that standard, safe rooms constructed to the almost identical FEMA criteria
26  are intended to “provide near-absolute protection from wind and wind-borne debris for
27  occupants” (FEMA 2021a). Since 1998, when FEMA first published guidance for residential and
28  community safe rooms, “thousands of safe rooms have been built, and a growing number of
29  these safe rooms have already saved lives in actual events. There has not been a single reported
30  failure of a safe room constructed to FEMA criteria” (FEMA 2021a). Similarly, failures of
31  tornado shelters constructed to ICC 500 criteria (first published in 2008) have not been reported.

8

LY
N
O
n
io
at

m

3  Figure C32.1-1. Tornado wind speeds for design of tornado shelters and tornado
4  safe rooms (ICC 2020).
or


nf

6  C32.1.2 Permitted Procedures. The basic framework for determination of tornado loads is
rI

7  adapted from the wind load procedures specified in Chapters 26 through 31, given the general
Fo

8  similarities between tornadic winds and other windstorm types and how they both interact with
9  buildings and other structures. There are a number of differences in the details of the procedures
10  between tornado loads and wind loads that are incorporated in this chapter. Most of the load
11  determination procedures from Chapters 27 through 31 have been adapted to work for tornadoes,
12  except for the Chapter 28 Envelope Procedure, which is not compatible with the tornado load
13  methodology. The MWFRS wind load coefficients given in Chapter 28 were developed by
14  combining influence coefficients for structural actions in a building by pneumatically averaging
15  and weighting the wind pressures measured on model buildings in a boundary layer wind tunnel.

9
1  The integration of the loads incorporates the effects of the correlation of the gusts over various
2  portions of the building associated with the size of the gusts in an atmospheric boundary layer.
3  The size of the gusts in a tornado are likely much smaller than those in a standard atmospheric
4  boundary layer, suggesting that the procedures in Chapter 28 are not appropriate for adaptation
5  to tornado loads. No changes introduced in Chapter 32 restrict the applicability of Chapter 28 for
6  use in determining nontornadic wind loads.

7  The organization of Chapter 32 follows that of the wind load chapters on which so many of the
8  tornado load provision are based. Sections 32.1 through 32.13 provide tornado-specific versions

LY
9  of those same sections in Chapter 26, as applicable. This structure provides parallel construction
10  with the familiar layout of the wind loading provisions. Section 32.14 introduces a new factor to

N
11  modify external pressure coefficients developed for boundary layer winds to work with tornadic

O
12  winds. Sections 32.15 through 32.18 provide the tornado loading procedures adapted from
13  Chapters 27, 29, 30, and 31, respectively. Last, Appendix G provides additional tornado hazard
14  n
maps for longer return periods than are used in Chapter 32, which can be of use for performance-
io
15  based tornado designs and other applications.
at

16  C32.1.3 Performance-Based Procedures. Performance-Based Procedures can quantify service


m

17  interruption or loss due to tornadoes, improve tornado resistance, or perform an enhanced
18  evaluation of the MWFRS and/or envelope. See Section C26.1.3 for discussion of performance-
or

19  based wind design, which could be adapted for use in tornado design. For nontornadic winds, the
nf

20  Prestandard for Performance-Based Wind Design (ASCE 2019) provides guidance on the
process to perform the design and provides examples of acceptance criteria. For a tornado
rI

21 
22  performance-based design, the acceptance criteria of the Prestandard might need modification
Fo

23  based on the facilities use and owner’s requirements. The determinations of the wind hazard
24  characteristics for tornado loading will require consultation with a wind engineering expert. For
25  applications where tornado hazard maps at longer return periods than those provided in Chapter
26  32 are needed, see Appendix CG.

27 

28 

10
1  C32.3 SYMBOLS AND NOTATION

2  The following notations apply to the Commentary for Chapter 32:

3  ABL = Atmospheric boundary layer

4  APC = Atmospheric pressure change

5  DI = Damage indicator

6  DOD = Degree of damage

LY
7  EF Scale = Enhanced Fujita Scale

N
8  FEMA = Federal Emergency Management Agency

O
9  FR12 = One-and-two-family residence DI in the EF Scale

10  ICC = International Code Council n


io
11  MRI = Mean recurrence interval
at

12  NWS = National Weather Service


m

PV = Photovoltaic
or

13 

14 
nf
rI

15  C32.5 TORNADO HAZARD MAPS


Fo

16  C32.5.1 Tornado Speed. The tornado hazard maps provided in Section 32.5.1 and Appendix G
17  are based on the first-ever engineering-derived tornado wind speed maps produced for the
18  contiguous United States, as reported by Twisdale et al. (2021). Tornado databases (for the years
19  of 1950 through 2016) and physiographic data were analyzed to identify large-scale spatial
20  patterns with similar tornado characteristics, using multivariate statistical analysis of 11 tornado
21  and physiographic variables. From these patterns, nine broad US regions with distinct tornado
22  climatologies were developed. Regional and national data were analyzed to produce probabilistic
23  models for population bias; Enhanced Fujita Scale (EF-Scale) distribution; tornado path length,

11
1  width, direction, and translational speed; radius of maximum winds; tornado path length intensity
2  variation; variable path widths within a tornado; mean to maximum path width ratios; and
3  maximum damage widths relative to local path width.

4  Since tornado wind speeds that result in an assigned EF rating for each tornado are estimated
5  based on observed damage in the field, development of probability distributions that relate
6  tornado wind speed to observed damage was an important element of the work. A probabilistic
7  load and resistance modeling framework was used to develop engineering-derived wind speeds
8  from the EF-Scale tornado intensity rating system (Texas Tech University 2006) by analyzing

LY
9  the most common Damage Indicator (DI) used to rate EF3 to EF5 tornadoes, namely that of one-
10  and two-family residences (FR12). Engineering models were developed to enable tornado wind

N
11  speed estimation for the EF-Scale Degrees of Damage (DOD) for FR12. This work used load

O
12  path quality considerations that the National Weather Service (NWS) employs in its damage
13  assessment process to estimate wind speeds and assign EF ratings to tornadoes. Field
14  n
investigations of tornadoes were conducted to gain a better understanding of the field process
io
15  used by the NWS, to investigate failure modes of various DIs, and to obtain data for validation of
at
16  the models used to estimate tornado wind speed from observed damage. The developed wind
speed distributions, based on 44 3D models of FR12 structures, are broad and encompass the
m

17 
18  original judgment-based EF-Scale wind speeds. These data were used to support development of
or

19  a probabilistic tornado wind field model. Monte Carlo methods were then used to simulate
nf

20  tornadoes, produce damage swaths, score wind speed exceedances numerically over a wind
21  range of wind speeds, and develop regional tornado hazard curves. Since tornadoes often have
rI

22  modest path widths relative to the size of the structure, tornado wind speed risk depends on
Fo

23  structure size (see Section C32.5.4). Therefore, the tornado hazard curves were developed as a
24  function of structure size.

25  The final step was the development of the tornado hazard maps, which reflect the spatial
26  variation in risk across the contiguous United States and within the context of the tornado
27  climatology regions. A one-degree grid was used to map the regional tornado wind speeds for a
28  given return period. Gaussian smoothing was applied to the grid to reflect epistemic uncertainties
29  in the location of the region boundaries. ArcGIS kriging routines were next used to produce the

12
1  tornado speed contours from the smoothed grid. These contours were further smoothed using the
2  Polynomial Approximation with Exponential Kernel (PAEK) method routine in ArcGIS.

3  Tornado hazard maps for return periods of 1,700, 3,000, 10,000, 100,000, 1,000,000, and
4  10,000,000 years are included in ASCE 7. These six return periods span the range of design
5  needs for conventional buildings and structures, Critical and Essential Facilities, and nuclear
6  power plants. The 1,700-year and 3,000-year maps are provided in Section 32.5 for use in
7  determining tornado loads on Risk Category III and IV buildings, other structures, and facilities.
8  These are the same mean recurrence intervals (MRIs) used for Risk Category III and IV basic

LY
9  wind speeds in Chapter 26 (see Return Periods later in this commentary section for information
10  on the reliability analysis underpinning selection of these MRIs). Tornado hazard maps for the

N
11  10,000-year and longer return periods are provided in Appendix G, which may be needed for

O
12  performance-based tornado designs and other applications.

13 
n
Maps were produced for eight structure sizes at each return period, ranging from a point target of
io
14  1 ft2 (0.9 m2) to 4 million ft2 (371,612 m2). An example of a very small target would be a
at
15  freestanding communications tower. The maps for this small structure size are also helpful for
16  interpolation between this and the next largest available mapped size of 2,000 ft2 (186 m2) (see
m

17  Section 32.5.4 for determination of structure size, referred to as Effective Plan Area).
or

18  Tornado speeds are contoured at 10 mi/h (4.5 m/s) intervals except for the innermost (greatest
nf

19  speed) and outermost (least speed) contours, which are shown to 1 mi/h (0.45 m/s) to aid with
interpolation. In South Texas there is often a narrow ridge of slightly higher tornado speeds that
rI

20 
21  occurs parallel to the coast, a few counties inland. This ridge is represented by either a short
Fo

22  contour or a point value, for situations where the ridge maximum value (near the border with
23  Mexico) is either a small plateau or a local peak, respectively.

24  The tornado speeds shown on the maps represent the maximum 3 s gust horizontal wind speed at
25  a height of 33 ft (10 m) anywhere within the effective plan area of the target building or other
26  structure produced by the translating tornado (Twisdale et al. 2021). The tornado speed is
27  defined independently of terrain.

13
1  The tornado hazard maps were developed through probabilistic models that are “best estimates”
2  rather than “conservatively based.” This approach follows the intent of ASCE 7 wind speed
3  maps for other wind hazards, including hurricanes (Vickery et al. 2009) and straight winds
4  (Pintar et al. 2015), as provided in Section 26.5. Epistemic (modeling) uncertainties were
5  considered in the tornado hazard map development process. As a result, the maps are intended to
6  provide results that are also applicable to the nuclear power industry in the United States, where
7  both aleatory (randomness) and epistemic uncertainties are required in the risk analysis of
8  nuclear power plants.

LY
9  Return Periods. An ad hoc working group composed of members from both the ASCE 7-22
10  Load Combinations and Wind Load Subcommittees conducted a study to identify appropriate

N
11  return periods for the tornado hazard maps in Chapter 32 (Levitan et al. 2021). The methodology

O
12  used to determine wind speed map return periods for each Risk Category in ASCE 7-16 [see
13  Section C26.5.1 and McAllister et al. (2018)] was adapted for use with tornado loads. A
14  n
reliability analysis was conducted for roof uplift, the primary mode of structural failure in
io
15  tornadic events, where roof framing members and/or connections fail due to significant wind
at
16  pressure resulting from the combination of internal and external pressures. The reliability
analysis procedures used Monte Carlo analysis, in which significant uncertainties for system
m

17 
18  demands and capacity were identified and quantified in the form of random variables with
or

19  defined probability distributions. Investigation over a range of return periods found that the
nf

20  1,700 and 3,000-year MRIs used for Risk Category III and IV wind hazard maps in Chapter 26
21  were also suitable for the tornado hazard maps. The results of a series of risk-informed analyses
rI

22  show that the tornadic load criteria of Chapter 32 provide reasonable consistency with the
Fo

23  reliability delivered by the existing criteria in Chapters 26 and 27 for MWFRSs (Levitan et al.
24  2021). Tornado speeds at return periods of 300 and 700 years are generally so low that tornado
25  loads would not control over Chapter 26 wind loads, so design for tornadoes is not required for
26  Risk Category I and II buildings and other structures. It should be recognized that the tornado
27  hazard maps and tornado load design methods are new, that the profession needs experience with
28  their application, and that changes may be needed in the future.

29  The tornado speeds required by Chapter 32 depend on the geographic location, Risk Category,
30  and Effective Plan Area of the building or other structure. For many Risk Category III buildings

14
1  and other structures located in the area having the greatest tornado hazard (roughly the area
2  between northern Texas, South Dakota/Minnesota, and northern Alabama, and extending father
3  east for large target sizes), the tornado speed will be in the EF1 intensity range (86 to 110 mi/h
4  [140 to 180 km/h]). For Risk Category IV, the tornado speeds will typically be of EF1 or EF2
5  intensity (86 to 135 mi/h [140 to 220 km/h]) in the same geographic area. While this clearly does
6  not include the most intense tornadoes, it does include the majority of tornadoes that occur in the
7  United States. From 1995 to 2016, 61.3% of the over 1,200 recorded tornadoes per year were
8  rated EF0, 27.8% were EF1, 8.0% EF2, 2.3% EF3, 0.52% EF4, and 0.05% EF5 (Twisdale et al.

LY
9  2021). In addition, most of the area impacted by a tornado does not experience the maximum
10  winds on which the tornado is rated. For example, in the 2011 EF5 Joplin, Missouri, Tornado, an

N
11  estimated 72% of the area swept by the tornado experienced EF0 to EF2 winds, while just 28%
12  experienced EF3 to EF5 winds (NIST 2014). While the aggregate number of fatalities caused by

O
13  lower-intensity tornadoes is clearly smaller than for the more intense tornadoes, property damage
14 
n
follows a different trend. Losses per individual tornado increase dramatically with increasing EF
io
15  rating; however, since there are so many more lower-intensity tornadoes, the aggregate losses
16  caused by all EF1 tornadoes are very similar in magnitude to aggregate losses for all EF2s, all
at

17  EF3s, all EF4s, and all EF5s (NIST 2014, Section 2.4.2.2).
m

18  Tornado Speed Determination. Tornado speeds for Risk Category III and IV buildings and
or

19  other structures are shown in Figures 32.5-1 and 32.5-2, respectively. Maps for eight effective
nf

20  plan area sizes, ranging from a very small or point target of 1 ft2 (0.9 m2) to a very large target of
21  4 million ft2 (371,612 m2), are provided for each Risk Category. Interpolation is permitted for
rI

22  effective plan area sizes between mapped sizes. Linear interpolation of tornado speed should be
Fo

23  calculated using the logarithm of the effective plan area, rather than the plan area itself.
24 
25  Alternatively, location-specific tornado speeds may be determined using the ASCE Tornado
26  Design Geodatabase, available through the ASCE 7 Hazard Tool
27  (https://asce7hazardtool.online/), or approved equivalent. The ASCE Hazard Tool website
28  provides tornado speeds to the nearest mile per hour based on a defined location using
29  either latitude/longitude or an address, Risk Category, and effective plan area.

15
1  Example (in customary units). Determine VT for a Risk Category IV hospital building located
2  in East Dubuque, Illinois (at the far northwest corner of the state), having an effective plan area,
3  Ae = 200,000 ft2.

4  Single-map solution:
5  a. Select which map to use. The smallest effective plan area that is larger than that of
6  the hospital would be 250,000 ft2, so use Figure 32.5-2F.
7  b. Determine VT using the selected map. East Dubuque is located at the extreme
8  northwest corner of the state, between the 90 and 100 mi/h contours. By linear

LY
9  interpolation between the contours, VT = 99 mi/h.

N
10  Alternate solution, interpolating between two maps:

O
11  a. Determine the tornado speed for the smallest Ae that is larger than the hospital,
12  which is 99 mi/h per the single map solution.
13 
n
b. Determine the tornado speed using the largest Ae that is smaller than the hospital,
io
14  which would be 100,000 ft2. From Figure 32.5-2E, the tornado speed is 94 mi/h
at
15  by linear interpolation between the adjacent contours.
16  c. Use linear interpolation on the logarithm of Ae from results of steps a and b to
m

17  determine VT.


or

𝟗𝟗 𝟗𝟒
18  𝑽𝑻 𝟗𝟒 𝐥𝐨𝐠 𝟐𝟎𝟎, 𝟎𝟎𝟎 𝐥𝐨𝐠 𝟏𝟎𝟎, 𝟎𝟎𝟎
nf

𝐥𝐨𝐠 𝟐𝟓𝟎, 𝟎𝟎𝟎 𝐥𝐨𝐠 𝟏𝟎𝟎, 𝟎𝟎𝟎


19  𝟗𝟖 𝐦𝐢/𝐡.
rI

20 
Fo

21  C32.5.2 Where Design for Tornado Loads Is Not Required. For Risk Category III and IV
22  buildings, other structures, and facilities having a tornado speed VT < 60 mi/h (26.8 m/s), the
23  tornado loads will be of similar or lesser magnitude compared to the wind loads specified in
24  Chapter 26. Therefore, calculation of tornado loads is not necessary. In addition, when VT is less
25  than a certain percentage of the basic wind speed, V, at the same geographic location, the
26  determination of tornado loads is not necessary. These threshold speeds depend on V and
27  exposure, both determined per Chapter 26. These criteria in this section provide approximate
28  limits on VT, above which tornado loads begin to control some aspect of the wind load design.

16
1  These threshold speeds were established through analysis of worst-case combinations of
2  geographic location, exposure, effective plan area, mean roof height, enclosure classification,
3  building shape, and other parameters.

4  C32.5.3 Direction of Tornadic Wind. While tornado paths do have a strong directional
5  preference, the tornadic wind at any point on the building or other structure can come from any
6  direction.

7  C32.5.4 Effective Plan Area. In addition to Risk Category (with its associated return period)

LY
8  and geographic location, tornado speed is also a function of the size and shape of the footprint of
9  the building, other structure, or facility. This is due to the comparatively small width of a tornado

N
10  relative to the plan size of the structure; the width of tornado can be less than the size of the

O
11  structure in some cases, particularly for large structures. At any given return period, tornado
12  strike probabilities and associated maximum tornado speeds impacting the structure increase
13 
n
with increasing plan (i.e., footprint) area. The relevant size of the structure does not always
io
14  correspond to the area of the footprint. For example, the tornado strike probability for a building
at
15  that has a U-shaped plan is essentially the same as the strike probability for a building that has a
16  rectangular plan that encloses the U shape, but those two buildings could have very different
m

17  actual plan areas. Therefore, an effective plan area is used, which accounts for both plan size and
or

18  shape. The effective plan area, Ae, of a building, other structure or facility is equal to the area of
19  the smallest convex polygon enclosing the plan or footprint, meaning that Ae will always be as
nf

20  large or larger than the actual plan area of the building, other structure, or facility. A polygon is
rI

21  convex if all interior angles are less than 180 degrees. For rectangular plan buildings, a rectangle
22  is the smallest convex polygon. For hexagonal and similar multi-sided plan-shaped buildings, the
Fo

23  smallest convex polygon is the area of the respective hexagon or similar shape. For circular or
24  oval plan-shaped buildings, the areas of the smallest convex polygons enclosing these shapes are
25  approximately equal to respective areas of the circle or oval. For a building with a more complex
26  plan configuration, Figure C32.5-1a shows an example of the smallest convex polygon (dashed
27  line) enclosing the maximum plan area (i.e., building footprint). Alternatively, Ae can simply and
28  conservatively be calculated as the area of the smallest rectangle that encloses the maximum plan
29  area, as shown in Figure C32.5-1b.

17

2  (a) (b)

3  Figure C32.5-1. Effective plan areas for a building determined using (a) the

LY
4  smallest convex polygon enclosing the building plan, and (b) a rectangle
5  enclosing the same building plan.

N
O
6  For facilities comprised of multiple adjacent buildings and/or other structures, which all need to
7  survive the tornado strike in order for the facility to remain operational, the convex polygon
8  n
should be constructed to enclose the group of buildings/structures. An example of this situation
io
9  would be a hospital building and its nearby central utility plant (see Figure C32.5-2). This will
at
10  yield a larger effective plan area, that would then be used to determine a single, larger tornado
11  speed, VT, for all of the critical buildings and other structures within the bounds of the polygon.
m

12  An emergency operations center having an adjacent tower for emergency communications and
or

13  parking structure for emergency response vehicles needed following a disaster would be another
14  example.
nf
rI

15 
Fo

16 

17  (a) (b)

18
1  Figure C32.5-2. Effective plan area for a hospital and its central utility plant
2  determined using (a) the smallest convex polygon enclosing the facility, and (b) a
3  rectangle enclosing the facility.

5  C32.5.4.1 Essential Facilities. For facilities comprising multiple adjacent buildings and/or other
6  structures, which need to survive the tornado strike in order for the facility to remain operational,
7  the designer should consider the convex polygon that encloses the Essential Facility and all of

LY
8  the buildings and other structures required to maintain the functionality of the Essential Facility,
9  to capture the facility as a single site. This will yield a larger effective plan area used to

N
10  determine a single, larger tornado speed, VT, for all of the relevant buildings and other structures

O
11  within the bounds of the polygon. An example of this situation would be a hospital building and
12  its nearby central utility plant (see Figure C32.5-2). A second example would be an emergency
13 
n
operations center having an adjacent tower for emergency communications and parking structure
io
14  for emergency response vehicles needed following a disaster.
at

15  C32.5.4.2 Other than Essential Facilities. Where a building (or other structure) consists of
m

16  multiple, structurally independent sections (e.g., buildings with expansion or seismic joints), a
reduction in the effective plan area is permitted. The effective plan area for the entire building
or

17 
18  can be reduced to equal the effective plan area of the largest structurally independent section of
nf

19  the building. The tornado speed determined using this reduced Ae would be applied to the design
of the entire building.
rI

20 
Fo

21  C32.5.4.3 Ground-Mounted Photovoltaic Panel Systems. Most ground-mounted photovoltaic


22  (PV) panel systems are designed, permitted, and inspected as Risk Category I. This includes
23  large-scale systems (also referred to as “utility scale”) PV facilities. Where the nonbuilding
24  structures supporting PV panels are designed as Risk Category III or IV structures, tornado
25  provisions apply. As PV facilities are an intermittent power source, generating power only when
26  the sun is shining, they are not suitable for emergency backup power for Risk Category IV
27  buildings unless paired with an energy storage system.

19
1  Large-scale PV facilities can cover hundreds of acres of land, yet they are composed of hundreds
2  or thousands of small, structurally independent “tables” of PV panels, each with its own
3  independent foundation system. The PV panels on these independent nonbuilding structures are
4  linked by electrical conductors to central inverters that convert DC power to AC power. Large-
5  scale PV facilities can have dozens to hundreds of independent central inverters. If an electrical
6  fault is detected, only the inverter associated with that fault is shut down, and the remainder of
7  the facility remains operational. The entire PV facility will shut down only if the electrical
8  substation is shut down, or if the system otherwise detects a loss of the AC signal from the grid.

LY
9  Substations and grids are outside the scope of ASCE 7.

10  While there is little data on tornado strikes on large-scale PV facilities, in two known cases the

N
11  tornado strike caused only isolated, localized damage. These facilities typically remain

O
12  operational with localized damage. For ground-mounted PV installations, the effective plan area,
13  Ae, should be the effective plan area of the largest structurally independent nonbuilding structure
14  supporting PV panels. n
io
at
15  C32.6 TORNADO DIRECTIONALITY FACTOR
m

16  The tornado directionality factor, KdT, was computed using a procedure similar to that reported
by Vickery and Liu (2018) that was adapted for tornado loads (Vickery et al. 2021). For both the
or

17 
18  main wind force resisting system (MWFRS) and component and cladding (C&C) load cases, the
nf

19  direct wind-induced pressures were computed assuming that the pressure coefficients derived
from boundary layer wind tunnel tests are also valid for tornadic winds (e.g., Roueche et al.
rI

20 
21  2020; Kopp and Wu 2020). The effects of the vertical component of the tornadic winds on the
Fo

22  pressure coefficients were considered separately, as described in Section C32.14. Unlike straight-
23  line winds (including those from hurricanes, extratropical storms, and thunderstorms), the wind
24  speeds acting on a building during a tornado can vary significantly over the building at any given
25  instant in time, particularly for large buildings. These variations in tornado speed as a function of
26  building size are captured in the modeling process used to determine the values of KdT.

27  The analyses for KdT were performed using three buildings of different sizes, ranging from 1,800
28  ft2 (170 m2) to 250,000 ft2 (23,000 m2), including both low-slope and gabled roofs. A total of
29  5,000 simulated tornadoes were used to compute the tornado loads over a range of tornado

20
1  speeds and sizes. C&C loads are based on uplift loads computed for two different deck element
2  sizes, considering elements located in roof Zones 1′, 1, 2, and 3 for the low-slope roof buildings
3  per Figure 30.3-2A and Zones 1, 2, and 3 for the pitched-roof building per Figure 30.3-2B. The
4  MWFRS loads are based on uplift loads computed for the whole roof (for the small building) and
5  roof trusses or large roof panels (for the bigger buildings).

6  In the computation of KdT for C&C loads, KdT is taken as the ratio of the maximum wind induced
7  pressure at a point produced by the tornado, divided by the maximum GCp (over all wind
8  directions) multiplied by the maximum value of the velocity pressure at the same point produced

LY
9  by the passage of the tornado. The wind speed associated with the maximum velocity pressure at
10  the point at which KdT is being computed is always equal to or less than the maximum tornadic

N
11  wind experienced somewhere on the building. The tornado directionality factor for the MWFRS

O
12  is computed in a similar manner, except the normalizing wind load is the maximum MWFRS Cp
13  coupled with the maximum dynamic pressures associated with the maximum wind speed
14  n
experienced somewhere on the structural element (e.g., roof truss, frame section, etc.). Since the
io
15  entire structural element does not necessarily experience the maximum wind speed (unlike
at
16  components, which are small), the effective value of KdT generally decreases as the structure or
17  structural-element tributary area increases, although it is a weak trend. The effective value of KdT
m

18  for C&C behaves in an opposite manner, slightly increasing as the size of the structure increases,
or

19  as a greater portion of the building is associated with Zone 1′ GCp values, which are less
20  directionally sensitive.
nf

Results of the simulations for C&C loads showed that for all buildings and all roof zones except
rI

21 
22  1′, the mean KdT values typically ranged between about 0.65 to 0.75, with a few outliers above
Fo

23  and below. In the case of Zone 1′, KdT ranged from slightly over 0.8 for the intermediate-size
24  building to 0.97 for the large building. In addition to the building size effect, the higher KdT for
25  Zone 1′ is also a result of the fact that there is little variation of GCp with wind direction for
26  locations well removed from the edge of the building, away from any separation zones. The
27  effect of tornado size may also play a role. While no studies were performed examining KdT for
28  components on surfaces other than roofs, the KdT value of 0.75 for Zones 1, 2, and 3 was also
29  recommended for all other zones, including walls (Vickery et al. 2021). This is believed to be
30  conservative, since, as shown in Vickery and Liu (2018), Kd values for C&C on walls are lower

21
1  than those on roofs. For the MWFRS, the value of KdT = 0.8 was adopted based on the results of
2  the analysis (Vickery et al. 2021). For arched roofs, circular domes, and all other structures, KdT
3  is determined using Table 26.6-1, given the lack of tornado-specific data or simulations.

4  Buildings and other structures intended to remain operational in the event of extreme
5  environmental loading from a tornado include not only Essential Facilities, as defined in Section
6  1.2.1, but also the buildings and structures required to maintain the functionality of those
7  Essential Facilities (which comprise a subset of Risk Category IV buildings and other structures,
8  per Table 1.5-1). For both, it is critical to avoid breaches in the building envelope. Loss of any

LY
9  wall cladding, roof decking, or other element of the envelope can permit intrusion of wind, wind-
10  borne debris, wind-driven rain, and/or falling rain into the building. Even minor damage to these

N
11  and other elements, such as roof coverings, can allow significant amounts of rainwater to enter

O
12  the building. This often results in serious damage to building interiors, contents, and mechanical
13  and electrical systems, rendering the facility or parts of it nonoperational. This means that all, or
14  n
a portion of, the building envelope should be considered a Designated Nonstructural System, as
io
15  defined in Section 1.2.1. The directionality factor accounts for “the reduced probability of
at
16  maximum winds coming from any given direction, and the maximum pressure coefficient for a
17  specific load effect occurring for any given wind direction,” per Section C26.6. It is likely that
m

18  for a design tornado event that one or more portions of the building envelope are loaded with the
or

19  maximum pressure, so if the “load effect” is “pressure anywhere on the building envelope,” then
20  there would be no reduction in probability. When all or most of the C&C are part of this
nf

21  Designated Nonstructural System, a KdT value of 1.0 is more appropriate, as some portion of the
rI

22  envelope is nearly certain to have its maximum pressure coefficient at the direction of the
maximum winds.
Fo

23 

24  C32.7 TORNADO EXPOSURE

25  Effects of exposure on tornado characteristics are difficult to resolve near the surface with
26  observational techniques such as radar. Therefore, research has turned to wind tunnel and
27  numerical (e.g., computational fluid dynamics) studies. Most available research suggests that
28  tornado characteristics are modified by exposure, and these modifications are greatest near the

22
1  ground. However, great variability exists on exactly what those modifications are (e.g., Refan
2  and Hangan 2018) due to the both the complexity of tornadoes and the challenge of creating the
3  wealth of realistic terrain environments in the experiments. Based on this variability, the standard
4  does not to define tornado exposure. This results in use of a single tornado velocity pressure
5  profile per Section 32.10.1.

6  Several wind tunnel and numerical studies (e.g., Zhang and Sarkar 2012, Wang et al. 2017)
7  suggest a reduction in tangential wind speed, an increase in radial wind speed, and a reduction of
8  the core radius size for rougher exposures. However, the magnitude of both mean and maximum

LY
9  horizontal wind speeds is unaffected in some simulations (e.g., Nolan et al. 2017). Other studies
10  suggest an increase in wind speed above the surface due to rougher exposure (Davies-Jones et al.

N
11  2001). Intensity and path changes have also been noted in some numerical simulations, but the

O
12  direction of the changes depends on the setup (Lewellen 2014). The peak horizontal wind speed
13  shifts upward (i.e., higher above ground) in simulations with rougher exposure (e.g., Nolan et al.
14  n
2017), which is inconsistent with other simulations (e.g., Matsui and Tamura 2009).
io
C32.8 TOPOGRAPHIC FACTOR
at
15 
m

16  Similar to exposure, the effects of topography on tornado wind speeds are difficult to resolve
17  near the surface with observational techniques such as radar. Observational data, where
or

18  available, defy generalization when considering topography (Houser et al. 2020). Changes in
nf

19  tornado intensity, size, and path were noted in both numerical (Lewellen 2012) and wind tunnel
20  (Razavi and Sarkar 2018) experiments. The sign of the changes in these studies varied, but the
rI

21  most significant changes occurred very close to the ground. A few references suggest an increase
Fo

22  in wind speeds in valleys due to channeling, from both observational studies (Karstens et al.
23  2013) and numerical studies (Satrio et al. 2020). However, similar to studies of tornado exposure
24  (Section C32.7), the challenge of creating the wealth of realistic topographic cases in the
25  simulations makes generalization through numerical simulation difficult. Simulations using
26  various topographic setups show great variability in topographic effects on horizontal tornado
27  winds (Satrio et al. 2020). Based on this variability, the standard does not define a topographic
28  factor for tornadoes at this time.

29  C32.9 GROUND ELEVATION FACTOR

23
1  The effect of ground elevation on air density is not dependent on the type of windstorm;
2  therefore, the Ke provisions of Section 26.9 are appropriate, and there is no need to modify these
3  provisions.

4  C32.10 TORNADO VELOCITY PRESSURE

5  C32.10.1 Tornado Velocity Pressure Exposure Coefficient. For determination of a wind


6  loading profile for tornadoes (represented by KzTor), data from a number of sources were
7  reviewed and analyzed, including field (Refan et al. 2017, Kosiba and Wurman 2013), wind

LY
8  tunnel (Refan and Hangan 2018, Tang et al. 2018), and numerical studies (Dahl and Nolan 2017,
9  Reinhart et al. 2018). Tornadoes from field, wind tunnel and numerical studies used in

N
10  preliminary analysis were generally “strong” tornadoes (wind speeds associated with an EF2
tornado or higher). The profiles included tangential wind speed only (perpendicular to pressure

O
11 
12  gradient) and horizontal (i.e., combination of tangential and radial) wind speed. Since these
13 
n
studies were interested in wind speed, initial analysis focused on the behavior of the wind speed
io
14  profile with height (i.e., √KzTor) and not wind loading (i.e., KzTor). Although significant variability
in the tornado wind speed profiles was noted at all heights, all studies showed a peak wind speed
at
15 
16  occurring relatively close to surface near the horizontal (radial) location of strongest winds (i.e.,
m

17  radius of maximum winds) from the tornado “center.” Tornado wind speed profiles at the radius
or

18  of maximum winds routinely (but not always) displayed a nose-like profile, where the wind
19  speed is lower both above and below the nose. This behavior is different from the atmospheric
nf

20  boundary layer (ABL) flows, in which the wind speed increases monotonically with height until
rI

21  a maximum wind speed is reached at the gradient height.


22 
Fo

23  After preliminary analysis of the tornado wind speed profiles it was determined that only radar
24  profiles would be used to create the ASCE 7 profile, as this was the largest and most consistent
25  data set. These 36 profiles consisted of the maximum radar-estimated wind speed over the time
26  of the tornado. The sampling of the maximum wind speed was over a short time period and
27  therefore did not employ any averaging schemes. No modifications for exposure were made.
28  Radar measurements were typically collected in open, flat terrain. Tornadoes from all studies
29  represent different swirl ratios (i.e., vortex structure). However, the averaging techniques
30  employed, the difficulty of assessing the parameter in the field (Kosiba and Wurman 2013), and

24
1  varying definitions depending on experimental technique precluded any detailed profile analysis
2  conditional on swirl ratio values.

4  The next step was to determine the approximate height where the nose occurs and then to taper
5  the profile away from this height. To better reflect behavior of tornado wind speed profiles close
6  to the ground, profiles that had a minimum observation height below 200 ft (61 m) were kept for
7  analysis. Given this condition, the wind speeds at any height for all studies were normalized by
8  the wind speed at the closest observation height to 164 ft (50 m). It should be noted that the

LY
9  normalized profile behavior is sensitive to the minimum height of data collection. For purposes
10  of statistical analysis, the radar measurements were also binned in height ranges of 66 ft (20 m)

N
11  for elevations below 328 ft (100 m), and 164 ft (50 m) for higher elevations. Where less than 10
12  observations were found in a bin, the bin was extended in height until the condition was met, to

O
13  minimize bin-to-bin variation and the effect of outliers. The results of the normalization and
14 
n
binning processes yield tornado speed profiles ( 𝐾 ) for 36 tornadoes as shown by gray lines
io
15  in Figure C32.10-1. The median values (solid black line) of 𝐾 , as well as median plus and
at
16  minus one standard deviation (dashed lines), are plotted at the mean height of each bin.
17 
m

18  When considering the wind loading profile, the committee chose to approximately follow the
or

19  median radar profile from Figure C32.10-1, resulting in the KzTor profile shown in Figure
20  C32.10-2. Below 200 ft (61 m), KzTor is uniform, with a value of 1.0, effectively “anchored” by
nf

21  the Kz = 1 value for Exposure C at 33 ft (10 m), where the tornado and Exposure C profiles cross.
rI

22  From the 200 ft (61 m) to 328 ft (100 m) levels, a linear function was used to represent the
23  decreasing tapered behavior which is apparent in the median profiles. Above 328 ft (100 m),
Fo

24  KzTor was kept at 0.9, as the median profiles tended to stabilize around that value as height
25  increased, and the volume of available data also decreased above this height.
26 

25
LY
N
O

2  n
Figure C32.10-1. Vertical profiles of tornado winds plotted as the square root of velocity
io
3  pressure exposure coefficient (√KzTor) for tornadoes, showing all the radar profiles (gray), the
at
4  median profile (black solid line), and plus and minus one standard deviation from the median
5  (black dashed line).
m


or
nf
rI
Fo

26
LY
N
O

2  n
Figure C32.10-2. Vertical profiles of tornado velocity pressure (KzTor) versus that of Exposure B
io
3  and Exposure C for nontornadic winds (Kz) in Chapter 26 for the lowest 500 ft (152.4 m).
at


m

5  C32.10.2 Tornado Velocity Pressure. The mass density of air in the tornado velocity pressure
or

6  relation, Equation (32.10-1), which is a component of the constant at the beginning of that
nf

7  equation, is based on standard atmospheric conditions. The wind velocity pressure relation,
8  Equation (26.1-1), uses this same approach (Section C26.10.2). The mass density of air is a
rI

9  function of elevation (Section C26.9), atmospheric pressure, temperature, and humidity.


Fo

10  Although atmospheric pressure drops in a tornado, this effect is not currently incorporated in the
11  formulation of tornado velocity pressure, similar to how the low pressure in hurricanes is not
12  incorporated in the wind velocity pressure equation when used in hurricane-prone regions.

13  C32.11 TORNADO GUST EFFECTS

14  C32.11.1 Tornado Gust-Effect Factor. The gust effect factor for rigid structures takes into
15  account the lack of correlation of the gusts in a turbulent wind field, which results in gust effect
16  factors lower than 1.0. In the case of tornadic winds, while the gustiness due to turbulence may

27
1  be less than in atmospheric boundary layer (ABL) winds, the tornado winds vary in both
2  direction and speed over the building or structure, resulting in lower peak loads compared to
3  ABL winds, whose mean wind speed and direction are constant over the building or structure.
4  The duration of a tornado is sufficiently short such that the gust factor provisions of Section
5  26.11.5 for flexible or dynamically sensitive buildings and other structures do not apply.
6  Consequently, the rigid building/structure gust factor was selected for use with the tornado
7  MWFRS loads.

8  C32.12 TORNADO ENCLOSURE CLASSIFICATION

LY
9  C32.12.1 General. Sealed structures are particularly susceptible to loads induced by atmospheric

N
10  pressure change associated with the passage of a tornado. A new enclosure classification of

O
11  “sealed” was introduced to enable capturing of these effects, which are applicable to other
12  structures, including tanks and vessels that have controlled ventilation such that tornado-induced
13 
n
atmospheric pressure changes will not be transmitted to the inside of the structure. Additional
io
14  information on atmospheric pressure change and its effects is provided in Section C32.13.
at

15  The wind-borne debris protection requirements of this section cover glazed openings, not other
m

16  elements of the building envelope, because wind-borne debris impacts through unprotected
glazing are likely to shatter the glazing and create large openings, whereas other building
or

17 
18  envelope elements are likely to only have penetrations the size of the wind-borne debris, which
nf

19  would not change the enclosure classification. See Section C32.17 for more information.
rI

20  C32.12.2 Openings. Wind-borne debris hazards are greater for tornadoes than for hurricanes.
Fo

21  The updrafts in a tornado can loft debris higher in the air, creating the opportunity for more,
22  larger, and faster traveling debris compared to hurricanes, where debris is mainly transported
23  horizontally. Comparisons of observed window breakage rates between several hurricanes and
24  from the 2011 Joplin, Missouri, tornado demonstrated greater breakage rates for tornadoes over
25  hurricanes in areas that experienced approximately the same maximum wind speeds. This
26  finding is consistent with observations from many field investigations that failure of glazing is
27  common in tornadoes (e.g., Roueche and Prevatt 2013). Since unprotected windows are prone to
28  failure during tornadoes, glazed openings that do not meet the protection requirements of Section

28
1  32.12.3.1 are considered openings for the purposes of determining enclosure classification, as
2  provided in Section 32.12.2. This section requires that for buildings and other structures that
3  would otherwise be classified as enclosed, the criteria for partial enclosure classification are
4  reevaluated for the case that unprotected glazing on each assumed windward wall fails and is
5  considered to be open.

6  C32.12.3 Protection of Glazed Openings. Failure of unprotected glazing is common in


7  tornadoes. The resultant intrusion of wind, wind-borne debris, and wind-driven rain can result in
8  hazards to life safety and significant interior damage, including collapse of interior ceilings and

LY
9  walls (see also Section C32.6). For buildings and other structures intended to remain operational
10  in the event of a tornado (including Essential Facilities) and where a building owner’s tornado

N
11  performance objective is to reduce occupancy disruption and interior damage during a tornado, it

O
12  is important to specify glazing assemblies that have been tested for wind-borne debris impact, in
13  addition to meeting static pressure requirements.
n
io
14  C32.12.3.1 Protection Requirements for Glazed Openings. Assemblies that resist penetration
at
15  by the ASTM E1996 test missile D are likely to resist penetration of most wind-borne debris
16  generated by design tornado events with speeds per Section 32.5. Greater protection is provided
m

17  by assemblies that resist penetration by test missile E. For significantly enhanced protection to
or

18  resist breaching of exterior glazing by more intense tornadoes, glazing assemblies can be
19  specified that have been tested in accordance with AAMA 512 (2011), using tornado test
nf

20  missiles given in ICC 500 (2020) or FEMA P-361 2021a).


rI

21  Impact resistant glazing and permanently anchored impact-protective systems do not require
Fo

22  human intervention in order to achieve wind-borne debris protection. Impact-protective systems
23  that require power for deployment (e.g., motorized shutters) or human intervention for
24  deployment should only be specified when the building owner has procedures that will ensure
25  deployment prior to a tornado. Given the limited warning time for tornadoes (typically ranging
26  from no warning to tens of minutes), many operable impact protections systems used in
27  hurricane-prone regions would not be applicable for tornado protection given the time it takes to
28  deploy them or the need to deploy them from outside the building. Use of operable systems
29  should be limited to facilities that are staffed 24 hours a day, seven days a week, with people

29
1  trained in how to deploy the systems and that maintain situational awareness (such as a fire
2  station or emergency operations center). As remote or automatic deployment of systems becomes
3  available with “smart building” designs, these could be used in place of the 24 h staffing
4  requirement.

5  C32.13 TORNADO INTERNAL PRESSURE COEFFICIENTS

6  For nontornadic winds, internal pressures are caused solely by external wind-induced pressures
7  affecting the interior through openings in the building envelope. In tornadoes, there is a second

LY
8  mechanism that also affects internal pressures, called atmospheric pressure change (APC). The
9  atmospheric (or static) pressure at the center of the tornado is much lower than the ambient

N
10  values, similar to a hurricane but on a much smaller spatial scale. As the core of the tornado

O
11  moves near or over the building, the atmospheric pressure outside of the building drops rapidly.
12  If the air permeability of the building is low, such that this drop in static pressure cannot be
13 
n
swiftly transmitted into the building, this creates a differential static pressure between the
io
14  exterior and interior of the building, which effectively behaves as a positive internal pressure.
at

15  To determine the tornado internal pressure coefficient, GCpiT, tornado wind-induced internal
m

16  pressures, APC-induced internal pressures, and their combined effects were computed for sealed,
17  enclosed (assuming uniform distribution of leakage), and partially enclosed (dominant opening)
or

18  buildings, as described by Vickery et al. (2021). A total of 5,000 simulated tornados were used to
nf

19  generate wind loading data for models of each of the three enclosure classifications. For the
20  enclosed and partially enclosed cases, internal pressures were computed with and without
rI

21  consideration of APC. Each simulated tornado produced a time series of tornado-induced
Fo

22  pressures on the roof and wall panels of the model building. Using these time series of wind
23  pressures, the internal pressure and the APC associated with the time at which the tornado-
24  induced net pressure on an element reached its maximum (or minimum) value was retained and
25  used in subsequent analyses. In the computation of the internal pressures due to tornadic winds
26  only, it was assumed that there was no net flow into or out of the building, and that the balance
27  of the net flow occurs instantaneously. This “conventional” internal pressure has a value near the
28  weighted average of the external pressures. The effective (or net) internal pressure is a
29  combination of 1) the conventional internal pressure due to tornadic wind-induced pressure and

30
1  external pressure change, and 2) the differential between the external static pressure due to the
2  APC and the internal static pressure. The effective internal pressure can also vary significantly
3  over the exterior of the building, particularly if the tornado producing the APC is small compared
4  to the building. Partially open and open buildings and other structures are not subject to APC-
5  induced pressures.

6  The tornado internal pressure coefficient, GCpiT, given in Table 32.13-1 accounts for the
7  effective internal pressure induced by the tornado, i.e., the combined external and internal
8  pressures due to the APC as well as internal pressures due to building porosity and the direct

LY
9  action of the tornadic wind-induced exterior pressures. For enclosed buildings, GCpiT = +0.55
10  and –0.18, where the increased positive internal pressure compared to the GCpi = +0.18 for wind

N
11  loads (per Table 26.13-1) is driven by the contribution of APC. Although analyses were also

O
12  conducted for a (theoretical) completely sealed building, in actuality, no buildings are completely
13  sealed. Therefore, the sealed case GCpiT = +1.0 in Table 32.13-1 is only required for sealed other
14  n
structures, such as tanks and vessels having controlled ventilation such that tornado-induced
io
15  atmospheric pressure changes will not be transmitted to the inside of the structure. It is possible
at
16  that for buildings (or portions thereof) having tightly controlled ventilation, such as certain
17  facilities housing hazardous substances that pose a threat to the public if released, use of a
m

18  positive GCpiT between the values for the enclosed and sealed conditions may be appropriate.
or

19  C32.14 TORNADO EXTERNAL PRESSURE COEFFICIENTS


nf

20  There are strong updrafts that occur near the radius of maximum winds or corner flow region in
rI

21  tornadoes, so the mean winds are no longer always horizontal. This vertical component of the
Fo

22  wind changes the aerodynamics of wind flow around the building and, consequently, the
23  pressures on the building, particularly over the roof where it can result in increased uplift. The
24  tornado pressure coefficient adjustment factor for vertical winds, KvT, is used to modify roof
25  uplift pressure coefficients that were previously developed for boundary layer winds to account
26  for these effects (see Section C32.6 for more information on the applicability of pressure
27  coefficients developed for boundary layer winds to tornadic winds). The parameter KvT takes into
28  account the effect of the vertical component of the wind speed within the core of a tornado.
29  Equations for MWFRS and C&C tornado loads in Sections 32.15 through 32.17 that are

31
1  applicable to uplift on roofs and some rooftop appurtenances include KvT as a modifier on the
2  external pressure coefficient term.

3  The development of KvT was documented by Vickery et al. (2021) and summarized here. The
4  vertical component of the tornadic wind changes the vertical angle of attack of the wind with
5  respect to the surface of the roof. The pressures resulting from the 5,000 tornado simulations
6  were calculated with and without the effect of vertical winds. During each simulated tornado,
7  horizontal and vertical wind speeds, atmospheric pressure, and wind-induced pressures are
8  computed at every time step as the modeled tornado passes by a modeled building. The

LY
9  difference between the two sets of load data is the change in the pressure due to the effect of
10  vertical winds. Each of the 5,000 tornado simulations yields a single value of KvT for each roof

N
11  deck element, at the element centroid. These KvT values are computed for each roof deck element

O
12  as the maximum value of the tornado-induced external wind pressure including the effects of
13  vertical winds, divided by the maximum pressure without the effects of vertical winds.
n
io
14  For sloped-roof buildings, the effects of the change in the angle of attack are handled by altering
15  the effective slope of the roof. The rate of change of GCp with roof slope for sloped-roof
at

16  buildings was developed using GCp data for roof slopes of 4:12 and 7:12. These data use
m

17  directional values of GCp, which are available for eight different zones (all of which can be
18  mapped to one of the three main roof C&C zones used in Figure 30.3-2B). GCp data are modeled
or

19  using the same methodology used in Hazus (Vickery et al. 2006). For low-slope roofs,
nf

20  experiments were performed at the boundary layer wind tunnel at the University of Western
Ontario. The leeward side of the model was tilted down into the floor of the tunnel, such that the
rI

21 
22  horizontal wind flow over the windward roof of the building had a negative roof slope (Vickery
Fo

23  et al. 2020). These experiments were repeated for a range of negative roof slopes in order to
24  develop a relationship between the roof slope and GCp and then determine the KvT values. For
25  both the low-slope and steep-slope cases, KvT values for C&C were determined for Roof Zones 1,
26  2, and 3 (using zone definitions from Figures 30.3-2A through 30.3-2D). For the MWFRS, KvT
27  was computed using the same buildings and tornado simulations described in Section C32.6.

28  For all tornado simulations used to develop these initial KvT values, at least some portion of the
29  building was within the core of the tornado. However, in reality, not all buildings will be
30  impacted by the core of a tornado when affected by a design tornado event. The likelihood of

32
1  being affected by the core increases with increasing design tornado speed. For a small building
2  with a design tornado speed of 100 mi/h (44.7 m/s), the building is not in the core in 33% of the
3  tornadoes, so KvT would be 1.0 for these cases. For the same small building having a design
4  tornado speed of 180 mi/h (80.5 m/s) (which occurs at around 100,000-year return periods, as
5  shown in Appendix G), the building is not in the core in just 4% of the tornadoes. The final KvT
6  values shown in Table 32.14-1 for both C&C and MWFRS were adjusted downward to account
7  for the fraction of the design event tornadoes where the building is within the core, considering
8  the mapped tornado speeds for Risk Category III and IV return periods. These adjustments were

LY
9  typically on the order of 5% decreases. When designing to the higher tornado speeds associated
10  with longer return periods as provided in Appendix G, see Vickery et al. (2021) for

N
11  recommendations on KvT values.

O
12  The effects of tornado updrafts on C&C pressure coefficients are modest at the edges and corners
13  of low-slope roofs (C&C Zones 2 and 3 in Figure 30.2-3A), but increase for these zones for
14  n
steeper roof slopes (see Table 32.14-1). Roof slope does not have much effect on Zone 1
io
15  pressure coefficients. For the MWFRS, KvT = 1.1 for roof uplift. For downward-acting loads on
at
16  roofs and all surfaces other than roofs, the values of KvT are 1.0 for both MWFRS and C&C. The
17  KvT values for building roofs were adapted for use on the roofs of bins, silos, and tanks. The
m

18  C&C zone shapes and locations for these structures are not the same as for buildings, but Section
or

19  32.17.5 provides a mapping to the coefficients for building roof zones provided in Table 32.14-1.
20  One phenomenon not currently accounted for is the spatial gradient of the static pressure and its
nf

21  potential net effects on drag and base-shear loads, a good topic for future research.
rI

22  C32.15 TORNADO LOADS ON BUILDINGS: MAIN WIND FORCE RESISTING


Fo

23  SYSTEM

24  C32.15.1 Enclosed, Partially Enclosed, and Partially Open Buildings. The equation for
25  determination of tornado loads on the MWFRS is similar to that used in Chapter 27, with tornado
26  versions of the parameters substituted for the original terms. The external pressure coefficients,
27  Cp, are multiplied by KvT to account for the effects of the vertical component of tornadic winds
28  (see Section C32.14). The velocity pressure, qi, for use with the tornado internal pressure
29  coefficient 𝐺𝐶 is not multiplied by KdT, since APC contributes the most to 𝐺𝐶 and since

33
1  APC is not a function of the direction of the tornadic winds. Internal pressure evaluation of the
2  roof and walls in partially enclosed buildings uses qi = qzop, where the height zop is defined as the
3  level of the lowest opening in the building that could affect the positive internal pressure. This
4  height is used in contrast to qz at the level of the highest opening in Equation (27.3-1) for wind
5  loads, since the tornado velocity pressure decreases with height, compared to the wind velocity
6  pressure, which increases with height.

7  C32.15.5 Design Load Cases. In cases where a tornado is smaller in plan area than the building
8  or other structure it strikes, the load cases from Figure 27.3-8 are likely to be conservative, since

LY
9  the maximum tornado speed does not extend across the full plan area of the building. The
10  asymmetry of the tornado wind field with respect to the building would tend to increase torsional

N
11  loads compared to other types of windstorms, hence the removal of the exception in Section

O
12  27.3.5 that eliminates the torsional load cases in Figure 27.3-8.

13 
n
C32.16 TORNADO LOADS ON BUILDING APPURTENANCES AND OTHER
io
14  STRUCTURES: MAIN WIND FORCE RESISTING SYSTEM
at

15  C32.16.1 General Requirements. The tornado speed for the building appurtenance should equal
m

16  that used for design of the building. Building appurtenances such as rooftop equipment and
rooftop solar (i.e., PV) panels are typically much smaller than the building itself. By using the
or

17 
18  same tornado speed as the building, the potential for failure of these elements and any associated
nf

19  risk of their being dislodged from the building is reduced to a level consistent with the reliability
goals of the standard. Where building appurtenances or other structures support the continued
rI

20 
21  operation of an Essential Facility, the tornado speed should equal that used in design of the
Fo

22  Essential Facility. See also Section C32.5.4.1.

23  The effective plan area of freestanding walls and signs is based on the data used in the
24  development of the tornado hazard maps (Section C32.5). A sensitivity analysis was performed
25  to determine the aspect ratio at which a building’s aspect ratio would begin to influence the
26  tornado speed at a given return period. This aspect ratio was found to be approximately 20:1.
27  Beyond this ratio, the structure’s azimuth begins to influence the strike probability, due to the

34
1  directional nature of tornadoes (predominantly travelling from southwest to northeast). For this
2  reason, long linear structures such as transmission lines are outside the scope of this document.

3  C32.16.2 Solid Freestanding Walls and Solid Signs. For taller walls and signs, the bullnose
4  profile of the velocity pressure exposure coefficient for tornadoes can lead to higher velocity
5  pressures at midheight of the sign as compared to the top of the sign or wall.

6  C32.16.3 Other Structures

LY
7  C32.16.3.1 Trussed Towers. Damage investigations have shown that wind-borne debris can
8  cling to trussed communications towers (FEMA 2012). Clinging debris such as metal roof panels

N
9  and chain-link fence with privacy slats can increase Af, the projected surface subject to wind
10  loading, thus increasing the wind load on the structure, with tower collapse as a possible result.

O
11  The values for clinging debris in Section 32.16.4.1 were taken from ASCE 7-16, Commentary
12 
n
Section C26.14.6, which were based on tornado damage investigations reported by FEMA
io
13  (2012). To minimize collapse potential, towers and other lattice-type open structures should be
14  designed for the additional tornado load caused by clinging debris.
at
m

15  C32.16.3.3 Roofs of Isolated Circular Bins, Silos, and Tanks. Unlike buildings, the tank
16  envelope is not often punctured to alleviate the pressure differential due to atmospheric pressure
or

17  change. This can result in amplified roof uplift pressures on bins, silos, and tanks. Compounding
nf

18  the effect on the roof is the vertical component of the tornadic winds (Section C32.14).
rI

19  C32.16.3.4 Rooftop Solar Panels for Buildings of All Heights with Flat Roofs or Gable or
20  Hip Roofs with Slopes Less than 7 degrees. The provisions in Chapter 29 for wind loads on
Fo

21  rooftop PV panels not parallel to the roof are based on wind tunnel test data of straight-line
22  winds (Section C29.4.3). Additional testing would be required to evaluate the nominal net
23  pressure coefficient GCrn as it relates to tornado loads. However, committee judgement suggests
24  that the vertical component of the tornado winds would increase the net uplift pressure on the
25  rooftop PV panel. The KvT factor could be applied to Equation (32.16-6) to account for this
26  increase.

35
1  C32.16.3.5 Rooftop Solar Panels Parallel to the Roof Surface on Buildings of All Heights
2  and Roof Slopes. The provisions in Chapter 29 for rooftop PV panels parallel to the roof surface
3  are calculated based on the external pressure coefficient for C&C for the roof zone below the
4  panels. The additional vertical component of the tornadic winds amplifies this net uplift pressure
5  (Section C32.14).

6  C32.17 TORNADO LOADS: COMPONENTS AND CLADDING

7  The procedures for determining tornado design pressures on C&C are based on the methods of

LY
8  Chapter 30, with several modifications. Tornado velocity pressures replace the velocity pressures
9  specified in Chapter 30. For enclosed, partially enclosed, and partially open buildings, the

N
10  velocity pressure used for determination of internal pressures is not multiplied by the tornado

O
11  directionality factor, since APC contributions to internal pressure are not directional as described
12  in Section C32.15.1. Internal pressure coefficients, GCpi, for these buildings are replaced with
13  𝐺𝐶 n
, and external pressure coefficients for equations that include roof uplift are multiplied by
io
14  KvT to account for the increased uplift due to vertical updrafts in the core of the tornado.
at

15  Compared to MWFRS, less is known about the C&C loads in tornadoes. C&C loads are more
m

16  sensitive to other aspects of the wind field, especially turbulence, which is not well understood in
17  tornadoes but a critical aspect of wind loading and bluff-body aerodynamics.
or

18  The development of the KvT factors by Vickery et al. (2021) consisted of research on enclosed
nf

19  and partially enclosed buildings (Section C32.14). Therefore, KvT is excluded from the equations
rI

20  for open buildings and attached canopies until additional data become available.
Fo

21  Wind-borne Debris: The only requirement for protection from wind-borne debris pertains to
22  exterior glazing, per Section 32.12.3. However, during a design tornado event, wind-borne debris
23  can damage other portions of the building envelope. Some debris may penetrate the roof and/or
24  exterior walls. Although such penetrations typically do not result in a change of enclosure
25  classification for enclosed buildings to partially enclosed conditions, the penetration can create a
26  pathway for rain to enter the building. For design guidance to minimize debris and/or rain
27  penetration through roof or wall assemblies, see FEMA (2011, 2012).

36
1  C32.17.5.1 Isolated Circular Bins, Silos, and Tanks. The KvT values for design of C&C for
2  roofs of bins, silos, and tanks are based on committee judgement. Zones 3 and 4 in Figure 30.12-
3  2 are edge zones similar to Zone 2 for a building roof, while Zones 1 and 2 in Figure 30.12-2 are
4  analogous to the interior Zone 1 region of a building roof.

5  C32.18 TORNADO LOADS: WIND TUNNEL PROCEDURE

6  The use of the Wind Tunnel Procedure is limited to determination of pressure and force
7  coefficients for use with MWFRS tornado load provisions for buildings and other structures in

LY
8  Sections 32.15 and 32.16, respectively, and with the C&C load provisions in Section 32.17.
9  Sections C32.6 and C32.14 include information on the applicability of pressure coefficients

N
10  developed for boundary-layer winds to tornadic winds.

O
11  Testing of isolated models in Exposure Category C is specified to be consistent with typical test
12  conditions used for development of pressure and force coefficients in Chapters 27, 29, and 30.
n
Traditional wind tunnel tests would commonly include placing the test building in a proximity
io
13 
14  model to capture interference and shielding effects from nearby buildings and structures.
at

15  However, such effects would not be the same between boundary layer winds and tornadic winds,
m

16  due to the rapid wind direction changes in tornadoes; hence the requirement for testing isolated
17  models.
or

18  While wind tunnels have been used for many years to estimate ABL wind loads on buildings and
nf

19  structures, procedures are currently being developed for determination of aerodynamic loads
rI

20  experienced by buildings and structures in tornadoes using simulators. Several tornado
21  simulators have been designed and constructed over the past years, including the WindEEE
Fo

22  Dome at the University of Western Ontario (Refan and Hangan, 2018), VorTECH at Texas Tech
23  University (Tang et al. 2018), and the Iowa State University simulator (Haan et al. 2008).
24  Increasingly available field data from tornadoes, along with proposed scaling methods (Refan et
25  al. 2014), have contributed to reproducing the velocity field of tornado vortices in simulators.
26  Despite this progress, many challenges remain. Little is known about turbulence levels and its
27  effects on wind loads in tornadoes. Although these laboratory simulations have proven to be
28  successful in replicating the velocity field and atmospheric pressure change in tornadoes, there is
29  little consensus over test conditions, test methodology, and data analysis (e.g., definition of the

37
1  pressure coefficient for tornadic winds and how to compare these coefficients with the ones
2  previously defined for ABL winds; see discussion by Kopp and Wu 2020). Therefore, at this
3  time, the use of tornado simulators in isolation is not considered for determination of pressure
4  and force coefficients for buildings and other structures.

5  REFERENCES

6  AAMA (American Architectural Manufacturers Association). 2011. Voluntary specifications for


7  tornado hazard mitigating fenestration products. AAMA 512-11. Schaumburg, IL:

LY
8  AAMA.
9  ASCE. 2019. Prestandard for performance-based wind design. Reston, VA: ASCE.

N
10  CDC (Centers for Disease Control and Prevention). 2012. “Tornado-related fatalities: Five states,

O
11  southeastern United States, April 25-28, 2011.” Morbidity and Monthly Weekly Report 61
12  (28), July 20. https://www.cdc.gov/mmwr/preview/mmwrhtml/mm6128a3.htm.
13 
n
Dahl, N. A., Nolan, D. S., Bryan, G. H., and Rotunno, R. (2017). Using high-resolution
io
14  simulations to quantify underestimates of tornado intensity from in situ observations.
at
15  Monthly Weather Review, 145(5), 1963-1982.
16  Davies-Jones, R., R. J. Trapp, and H. B. Bluestein. 2001. “Tornadoes and tornadic storms.”
m

17  In Severe convective storms, edited by C.A. Doswell III. 167–221. Boston, MA:
or

18  American Meteorological Society.


19  FEMA (Federal Emergency Management Agency). 2011. “Critical facilities located in tornado-
nf

20  prone regions: Recommendations for architects and engineers.” Tornado Recovery
rI

21  Advisory 6 (RA6). Washington, DC: FEMA.


22  https://www.fema.gov/sites/default/files/2020-07/critical-facilities-tornado-
Fo

23  regions_recovery-advisory.pdf.
24  FEMA. 2012. “Mitigation assessment team report: Spring 2011 tornadoes: April 25–28 and May
25  22. Building performance observations, recommendations and technical guidance.” P-
26  908. Washington, DC: FEMA. https://www.fema.gov/sites/default/files/2020-08/fema_p-
27  908_tornado-mat-combined.pdf .
28  FEMA. 2021b. Taking shelter from the storm: Building or installing a safe room for your home.
29  P-320, 5th ed. Washington, DC: FEMA. https://www.fema.gov/emergency-
30  managers/risk-management/safe-rooms.

38
1  FEMA. 2021a. Safe rooms for tornadoes and hurricanes: Guidance for community and
2  residential safe rooms. P-361, 4th ed. Washington, DC: FEMA.
3  https://www.fema.gov/emergency-managers/risk-management/safe-rooms.
4  Haan, F., P. Sarkar, and W. Gallus. (2008). Design, construction and performance of a large
5  tornado simulator for wind engineering applications. Engineering Structures. 30. 1146-
6  1159. 10.1016/j.engstruct.2007.07.010.
7  Houser, J. B., N. McGinnis, K. M. Butler, H. B. Bluestein, J. C. Snyder, and M. M. French.
8  2020. “Statistical and empirical relationships between tornado intensity and both

LY
9  topography and land cover using rapid-scan radar observations and GIS.” Monthly
10  Weather Review, 1–72.

N
11  ICC (International Code Council). 2014. 2015 International building code. Washington, DC:
12  ICC.

O
13  ICC. 2017. 2018 International existing building code. Washington, DC: ICC.
14 
n
ICC. 2020. ICC/NSSA Standard for the design and construction of storm shelters. ICC 500-2020.
io
15  Washington, DC: ICC and National Storm Shelter Association.
16  Insurance Information Institute. 2020. “Spotlight on: Catastrophes: Insurance issues.” April 28.
at

17  https://www.iii.org/article/spotlight-on-catastrophes-insurance-issues.
m

18  Karstens, C. D., W. A. Gallus Jr, B. D. Lee, and C. A. Finley. 2013. “Analysis of tornado-
induced tree fall using aerial photography from the Joplin, Missouri, and Tuscaloosa–
or

19 
20  Birmingham, Alabama, tornadoes of 2011.” Journal of Applied Meteorology and
nf

21  Climatology 52 (5): 1049–68.


Kopp, G. A., and C.-H. Wu. 2020. “A framework to compare wind loads on low-rise buildings in
rI

22 
23  tornadoes and atmospheric boundary layers.” Journal of Wind Engineering and Industrial
Fo

24  Aerodynamics 204. doi:10.1016/j.jweia.2020.104269.


25  Kosiba, K. A., and J. Wurman. 2013. “The three-dimensional structure and evolution of a
26  tornado boundary layer.” Weather and Forecasting 28 (6): 1552–61.
27  Levitan, M. L., B. Ellingwood, P. J. Vickery, Y. Li, T. McAllister, J. R. Harris, A. Salman, A.
28  Braik. 2021. “Tornado load criteria for ASCE Standard 7-22: A probabilistic approach.”
29  Publication pending. National Institute of Standards and Technology.

39
1  Lewellen, D. C. 2012. “Effects of topography on tornado dynamics: a simulation study.”
2  In Proc., 26th Conference on Severe Local Storms (5-8 November 2012), Nashville, TN.
3  American Meteorological Society.
4  Lewellen, D. C. 2014. “Local roughness effects on tornado dynamics.” In Proc., 27th
5  Conference on Severe Local Storms.(3-7 November 2014), Morgantown, WV. American
6  Meteorological Society.
7  Matsui, M., and Y. Tamura, Y. 2009. “Influence of swirl ratio and incident flow conditions on
8  generation of tornado-like vortex.” In Proc., EACWE5.

LY
9  McAllister, T. P., N. Wang, and B. R. Ellingwood. 2018. “Risk-informed mean recurrence
10  intervals for updated wind maps in ASCE 7-16.” Journal of Structural Engineering 144

N
11  (5), 06018001. https://doi.org/10.1061/(ASCE)ST.1943-541X.0002011.
12  NIST (National Institute of Standards and Technology). 2014. Final report: NIST technical

O
13  investigation of the May 22, 2011, tornado in Joplin Missouri. NCSTAR 3, March.
14 
n
http://www.nist.gov/manuscript-publication-search.cfm?pub_id=915628.
io
15  Nolan, D. S., N. A. Dahl, G. H. Bryan, and R. Rotunno. 2017. “Tornado vortex structure,
16  intensity, and surface wind gusts in large-eddy simulations with fully developed
at

17  turbulence.” Journal of the Atmospheric Sciences 74 (5): 1573–97.


m

18  https://journals.ametsoc.org/doi/10.1175/JAS-D-16-0258.1.
Pintar, A., E. Simiu, F. Lombardo, and M. Levitan. 2015. Maps of non-hurricane non-tornadic
or

19 
20  wind speeds with specified mean recurrence intervals for the contiguous United States
nf

21  using a two-dimensional Poisson process extreme value model and local regression.
Special Publication 500-301. National Institute of Standards and Technology.
rI

22 
23  http://dx.doi.org/10.6028/NIST.SP.500-301.
Fo

24  Refan, M., and H. Hangan. 2018. “Near surface experimental exploration of tornado
25  vortices.” Journal of Wind Engineering and Industrial Aerodynamics 175: 120–35.
26  Refan, M., H. Hangan, and J. Wurman. 2014. “Reproducing tornadoes in laboratory using proper
27  scaling.” Journal of Wind Engineering and Industrial Aerodynamics 135: 136–48.
28  Refan, M., H. Hangan, J. Wurman, and K. Kosiba. 2017. “Doppler radar-derived wind field of
29  five tornado events with application to engineering simulations.” Engineering
30  Structures 148: 509–21.

40
1  Razavi, A., and P. P. Sarkar. 2018. “Laboratory study of topographic effects on the near-surface
2  tornado flow field.” Boundary-Layer Meteorology 168 (2): 189–212.
3  Reinhart, A. E., D. J. Bodine, and F. T. Lombardo. 2018. “The impact of terrain on supercells
4  using idealized numerical simulations.” Proc., 29th Conference on Severe Local Storms.
5  American Meteorological Society.
6  Roueche, D. B., and D. O. Prevatt. 2013. “Residential damage patterns following the 2011
7  Tuscaloosa, AL and Joplin, MO tornadoes.” Journal of Disaster Research 8 (6) 1061–
8  67.

LY
9  Roueche, D. B., D. O. Prevatt, and F. Haan. 2020. “Tornado-induced and straight-line wind loads
10  on a low-rise building with consideration of internal pressure.” Frontiers in Built

N
11  Environment 6. https://doi.org/10.3389/fbuil.2020.00018.
12  Satrio, M. A., D. J. Bodine, A. E. Reinhart, T. Maruyama, and F. T. Lombardo. 2020.

O
13  “Understanding how complex terrain impacts tornado dynamics using a suite of high-
14 
n
resolution numerical simulations.” Journal of the Atmospheric Sciences: 1–68.
io
15  Tang, Z., C. Feng, L. Wu, D. Zuo, and D. L. James. 2018. “Characteristics of tornado-like
16  vortices simulated in a large-scale ward-type simulator.” Boundary-Layer
at

17  Meteorology 166 (2): 327–50.


m

18  Texas Tech University. 2006. “A recommendation for an Enhanced Fujita Scale.” Revision 2,
October. Wind Science and Engineering Center, Lubbock, TX.
or

19 
20  http://www.depts.ttu.edu/nwi/Pubs/FScale/EFScale.pdf.
nf

21  Twisdale, L. A., S. Banik, L. Mudd, S. Quayyum, F. Liu, M. Faletra, M. Hardy, P. Vickery, M.
Levitan, and L. Phan. 2021. “Tornado risk maps for building design: research and
rI

22 
23  development of tornado hazard risk assessment methodology.” Publication pending.
Fo

24  National Institute of Standards and Technology.


25  Vickery, P. J., S. Banik, M. L. Levitan, and L. A. Twisdale. 2021. “Development of new K-
26  factors and internal pressure coefficients for use in the tornado provisions of ASCE 7-
27  22.” Publication pending. National Institute of Standards and Technology.
28  Vickery, P. J., S. Banik, and F. Liu. 2020. Tornado induced loads on lowrise flat roofed
29  buildings with comparisons to wind loads derived in atmospheric boundary layer wind
30  tunnels. Final report to the National Institute of Standards and Technology on grant
31  number 70NANB17H253. Applied Research Associates, Inc.

41
1  Vickery, P. J., J. Lin, P. Skerlj, L.A. Twisdale, and K. Huang. 2006. “HAZUS-MH hurricane
2  model methodology. I: Hurricane hazard, terrain, and wind load modeling.” Natural
3  Hazards Review, 7 (2).
4  Vickery, P. J., and F. Liu. 2018. Development of wind directionality factors for component and
5  cladding and main wind force resisting loads in hurricane and non-hurricane prone
6  regions of the United States. Raleigh, NC: Applied Research Associates.
7  Vickery, P. J., D. Wadhera, L. A. Twisdale Jr, and F. M. Lavelle. 2009. “U.S. hurricane wind
8  speed risk and uncertainty.” J. Struct. Eng. 135 (3): 301–20.

LY
9  Wang, J., S. Cao, W. Pang, and J. Cao. 2017. “Experimental study on effects of ground
10  roughness on flow characteristics of tornado-like vortices.” Boundary-Layer

N
11  Meteorology 162 (2): 319–39.
12  Zhang, W., and P. P. Sarkar. 2012. “Near-ground tornado-like vortex structure resolved by

O
13  particle image velocimetry (PIV).” Experiments in Fluids 52(2): 479–93.
14 
n
io
at
m
or
nf
rI
Fo

42

You might also like