Professional Documents
Culture Documents
1. Introduction
The verification of a seismic hazard computer code is crucial for ensuring the user that the
calculations performed with it are reliable.
The numerical verification process of CRISIS2008 was carried out by adopting as tools the tests
developed in a project sponsored by the Pacific Earthquake Engineering Research Center (PEER)
and documented in “Verification of Probabilistic Seismic Hazard Analysis Computer Programs” by
Thomas et al. (2010). In that project, several seismic hazard codes were tested, namely:
FRISK88M Version 2.0 (developer: G. Toro)
EZ-FRISK (developer: R. McGuire)
XDC52, HAZ50, TREE50 (developer: B. Youngs)
HazFX v3.f, hazgridX v3.f, filtrate.peer.f (developers: A. Frankel and S. Harmsen)
ALEAS (developer: J. Savy Abramowitz)
EQ-Elements (developer: A. Mendez)
GP-Haz (developer: P. Tan)
NEWHAZ (developer: M. Stirling)
mrs 3.1 (developer: R. LaForge)
OpenSHA (developer: N. Field)
HAZDIR (developer: B. Rowshandel)
The verification exercise consisted of two sets of test problems aimed at testing fundamental aspects
of the codes: treatment of fault sources, recurrence models and rates, attenuation relationships and
their associated uncertainties. For the simplest cases analytical solutions were also provided.
In this chapter, set1 of the PEER tests has been used since only for such set Thomas et al. (2010)
illustrate the results.
Moreover, according to the definition of moment magnitude (Hanks and Kanamori, 1979):
2
M w log M 0 [dyne cm] 16.05 (2)
3
from which:
M 0 [dyne cm] 101.5M 16.05 . (3)
The seismic moment rate, i.e. the seismic moment released by the source in one year, can be
obtained by replacing the average slip on the fault, s, with the slip rate.
Since only one magnitude value (m) is possible, the seismic moment rate is the number of
earthquake of magnitude equal to m in one year times the seismic moment related to such
magnitude m:
M 0 As 101.5m 16.05 (4)
where
s = slip rate on the source [cm/yr].
-1 -1 -1
10 10 10
Site 1 Site 2 Site 3
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4
CRISIS -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1 -1 -1
10 10 10
Site 4 Site 5 Site 6
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4 -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1
10
Site 7
-2
10
-3
10
-4
10
0 0.2 0.4 0.6 0.8
PGA [g]
Figure 3 – Results for test case 1 (the numerical values obtained by PEER are not provided).
-1 -1 -1
10 10 10
Site 1 Site 2 Site 3
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
CRISIS
-4
PEER -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1 -1 -1
10 10 10
Site 4 Site 5 Site 6
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4 -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1
10
Site 7
-2
10
-3
10
-4
10
0 0.2 0.4 0.6 0.8
PGA [g]
Figure 4 – Results for test case 2. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), red dashed curves.
L=25 km
4 6
1
a
b
W=12 km
SET 1 CASE 4
Annual probability of exceedance
-1 -1 -1
10 10 10
Site 1 Site 2 Site 3
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
CRISIS
-4
PEER -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1 -1 -1
10 10 10
Site 4 Site 5 Site 6
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4 -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1
10
Site 7
-2
10
-3
10
-4
10
0 0.2 0.4 0.6 0.8
PGA [g]
Figure 6 – Results for test case 4. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), red dashed curves.
2.8 Set 1 case 5
Input parameters:
The source adopted is “fault 1”.
The seismic activity (magnitude distribution) is described by a truncated exponential model with a
b-value=0.9, a slip rate of 2 mm/yr, minimum magnitude Mmin=5 and maximum magnitude
Mmax=6.5
In CRISIS2008 we adopted a Gutenberg-Richter relation. The seismicity rate is in this case the
number of earthquakes with magnitude M≥ 5. The logic behind is the same of case 1. However, in
this context all the magnitudes between 5 and 6.5 are possible. Following Youngs and Coppersmith
(1985), the moment rate can be written as:
M max
M 0 As M
0 (m) f (m)dm (5)
where
exp( (m M min )
f (m) (6)
1 exp( ( M max M min )
where =ln(10)b.
-1 -1 -1
10 10 10
Site 1 Site 2 Site 3
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
CRISIS
-4
PEER -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1 -1 -1
10 10 10
Site 4 Site 5 Site 6
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4 -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1
10
Site 7
-2
10
-3
10
-4
10
0 0.2 0.4 0.6 0.8
PGA [g]
Figure 7 – Results for test case 5. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), red dashed curves.
-1 -1 -1
10 10 10
Site 1 Site 2 Site 3
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
CRISIS
-4
PEER -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1 -1 -1
10 10 10
Site 4 Site 5 Site 6
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4 -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1
10
Site 7
-2
10
-3
10
-4
10
0 0.2 0.4 0.6 0.8
PGA [g]
Figure 8 – Results for test case 6. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), red dashed curves.
f 1 ( m)
N (M min ) N (M ch ) exp( (m M min ) M min m M max
GR
(8),
1 exp( ( M max M min )
GR
M ch M ch
f 2 (m) n ( M ch ) M ch m M ch
2 2
where the term GR N (M min ) N (M ch ) represents the rate of the non-characteristic,
exponentially distributed earthquakes on the fault and n ( M ch ) is the rate density of the flat portion.
The two parameters needed for the description of the seismicity are the annual seismic rate and
the mean recurrence time between characteristic earthquakes (Tmean). Following the original model
of Youngs and Coppersmith (1985), we assume that:
1. events of any magnitude are possible. This leads to MmaxGR = Mch -Mc=5.95, where
Mmax=6.45 (from PEER) and Mc=0.25×2=0.5.Thus, a uniform distribution is adopted
between 5.95 and 6.45.
2. n (M ch ) n (M max 1) .
GR
Replacing eq. (8) in eq. (5) and solving the integral one obtains:
GR b exp( ( M max M min ) M 0 ( M max N ( M ch ) M 0 ( M max )(1 10 1.5M ch )
GR GR
)
M 0 As
1 exp( ( M max M min ) (1.5 b)
GR
c ln(10)M ch
(9).
The input values are the b-value=0.9 and the slip rate of 2 mm/yr. Hence, with hypotheses 1. and 2.,
GR= 0.0048 and Tmean=157 yr.
Results
Figure 9 shows the results.
SET 1 CASE 7
Annual probability of exceedance
-1 -1 -1
10 10 10
Site 1 Site 2 Site 3
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
CRISIS
-4
PEER -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1 -1 -1
10 10 10
Site 4 Site 5 Site 6
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4 -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1
10
Site 7
-2
10
-3
10
-4
10
0 0.2 0.4 0.6 0.8
PGA [g]
Figure 9 – Results for test case 7. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), red dashed curves.
-1 -1 -1
10 10 10
Site 1 Site 2 Site 3
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
CRISIS
-4
PEER -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1 -1 -1
10 10 10
Site 4 Site 5 Site 6
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4 -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1
10
Site 7
-2
10
-3
10
-4
10
0 0.2 0.4 0.6 0.8
PGA [g]
Figure 10 – Results for test case 8a. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), red dashed curves.
SET 1 CASE 8b
Annual probability of exceedance
-1 -1 -1
10 10 10
Site 1 Site 2 Site 3
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
CRISIS
-4
PEER -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1 -1 -1
10 10 10
Site 4 Site 5 Site 6
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4 -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1
10
Site 7
-2
10
-3
10
-4
10
0 0.2 0.4 0.6 0.8
PGA [g]
Figure 11 – Results for test case 8b. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), red dashed curves.
SET 1 CASE 8c
Annual probability of exceedance
-1 -1 -1
10 10 10
Site 1 Site 2 Site 3
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
CRISIS
-4
PEER -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1 -1 -1
10 10 10
Site 4 Site 5 Site 6
-2 -2 -2
10 10 10
-3 -3 -3
10 10 10
-4 -4 -4
10 10 10
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
PGA [g] PGA [g] PGA [g]
-1
10
Site 7
-2
10
-3
10
-4
10
0 0.2 0.4 0.6 0.8
PGA [g]
Figure 12 – Results for test case 8c. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), red dashed curves.
SET 1 CASE 10
Annual probability of exceedance
Site 1 Site 2
-2 -2
10 10
-4 -4
10 10
CRISIS
-6
PEER -6
10 10
0 0.2 0.4 0.6 0 0.2 0.4 0.6
PGA [g] PGA [g]
Site 3 Site 4
-2 -2
10 10
-4 -4
10 10
-6 -6
10 10
0 0.2 0.4 0.6 0 0.2 0.4 0.6
PGA [g] PGA [g]
Figure 13 – Results for test case 10. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), red dashed curves.
Site 1 Site 2
-2 -2
10 10
-4 -4
10 10
CRISIS
-6
PEER -6
10 10
0 0.2 0.4 0.6 0 0.2 0.4 0.6
PGA [g] PGA [g]
Site 3 Site 4
-2 -2
10 10
-4 -4
10 10
-6 -6
10 10
0 0.2 0.4 0.6 0 0.2 0.4 0.6
PGA [g] PGA [g]
Figure 14 – Results for test case 11. Comparison between the results obtained with CRISIS2008, blue curves with circles, and
those provided in Thomas et al. (2010), gray curves.
The new way of computing hazard, based on occurrence probabilities of events and probabilities of
exceedance of intensity values, and not anymore on exceedance rates, is checked through a test in
which poissonian probabilities are treated in non poissonian way.
The geometry of the source is very simple: a point source located at a depth of 15 Km. In spite of its
simplicity, the test is general enough since, internally, CRISIS 2008 performs all the arithmetic
related to exceedance probability calculations with discrete, point sources. For this example, the
receiver is located on the surface of the Earth, 0.2 degrees West and South of the point source.
The seismicity is described as a G-R relation with 0=0.07/year, =2 (treated deterministically),
M0=5 and Mu=8 (treated deterministically). Once these seismicity parameters are known, it is
possible to compute, under the Poisson assumption, the discrete probabilities of having 0, 1,…,n
events in given time frames. These probabilities were externally computed and later given to
CRISIS 2008 as if they were probabilities obtained from a non-Poisson model of unspecified type.
Results are compared with those obtained giving CRISIS 2008 the same seismicity parameters in
the form of a Poissonian source. Figure 15 to 17 show these comparisons, for time frames of 20, 50
and 100 years, respectively. In each case the curves are coincident, which means that the non-
poissonian occurrence probabilities are correctly handled by the code.
0.1
0.01
frame
Region 1 (Poisson)
0.001
Region 2 (non-Poisson)
0.0001
1 10 100 1000 10000
PGA (gal)
Figure 15 – Comparison between the results obtained with the Poissonian and non-Poissonian sources, for the next 20 years.
0.1
0.01
Region 1 (Poisson)
0.001
Region 2 (non-Poisson)
0.0001
1 10 100 1000 10000
PGA (gal)
Figure 16 – Comparison between the results obtained with the Poissonian and non-Poissonian sources, for the next 50 years.
Time frame: next 100 years
1
Exceedance probability in time frame
0.1
0.01
Region 1 (Poisson)
0.001
Region 2 (non-Poisson)
0.0001
1 10 100 1000 10000
PGA (gal)
Figure 7 – Comparison between the results obtained with the Poissonian and non-Poissonian sources, for the next 100 years.
4. Conclusions
The PEER tests have been adopted for checking the performance of CRISIS2008 code. The tests
showed a satisfactory agreement between the results provided by PEER and those obtained with
CRISIS2008, highlighting the capabilities of the code. The limited differences that have been found
can be explained by different assumptions made inside the code.
Finally the verification of the treatment of the non-poissonian occurrence probabilities, not included
in the PEER test, showed the reliability of the results carried out with this new option.