2009 Earthquake Loss Estimation

HAZUS Loss Estimation for California Scenario Earthquakes

Rui Chen, David Branum, and Chris J. Wills
California Geological Survey
June 2009

Using HAZUS, loss estimation software developed by the Federal Emergency Management Agency (FEMA), we estimated economic losses, social impact, and structural damage for fifty-six scenario earthquakes developed by the United States Geological Survey (USGS) (referred to as USGS scenarios). These scenarios were developed using ground motion prediction equations (GMPEs) of Boore et al. 1997 for peak ground acceleration (PGA), spectral acceleration at 0.3 second (SA03), and spectral acceleration at 1.0 second (SA10), and of Joyner and Boore 1988 for peak ground velocity (PGV). Source parameters, including magnitude and fault geometry, were based on the 2002 statewide probabilistic seismic hazard assessments for California. Significant research developments have occurred in the past few years in both ground motion prediction and rupture source characterization. To estimate how these latest developments affect HAZUS loss estimates, we developed and analyzed ten additional scenarios. These scenarios were developed based on source parameters and probability results of the Uniform California Earthquake Rupture Forecast Version 2 (UCERF 2) completed in 2008 by the Working Group on California Earthquake Probabilities and are referred to as UCERF 2-based scenarios. Three of the five Next Generation Attenuation (NGA) models were used to calculate PGA, PGV, SA03, and SA10 for the UCERF 2-based scenarios, namely models of Boore and Atkinson 2008, Campbell and Bozorgnia 2008, and Chiou and Youngs 2008. These three NGA models were used in the 2008 Update of the National Seismic Hazard Maps by USGS. The NGA models are founded on a significantly improved common database of ground motion recordings and to a large extent they have replaced earlier GMPEs for shallow crustal earthquakes in western United States. These NGA models resulted in notable changes in predicted ground motions, particularly for long periods (see figure 1, for example). One can expect that NGA models could lead to substantial differences in estimated losses and related effects than those derived from earlier GMPEs. Our analyses show that the estimated losses for the UCERF 2-based scenarios are 28% – 63% lower than those for the comparable USGS scenarios. Additional work is need to determine how the NGA models for earthquake shaking should be used in the HAZUS models for earthquake loss estimation.

Figure 1. Comparison of ground motions for a repeat of the 1906 M 7.9 San Francisco earthquake on the Northern San Andreas Fault (co-seismic rupture of all four segments).

Based on the analyses of the USGS scenarios in Northern California, the three most damaging scenario earthquakes involve co-seismic rupture of different combinations of three or more Northern San Andreas Fault segments, resulting in an estimated economic loss ranging from nearly $70 billion to over $80 billion (table 1). The most damaging scenario earthquake is a repeat of the 1906 San Francisco M7.9 earthquake. It would rupture all four segments of the Northern San Andreas Fault. Estimates made using HAZUS suggest that this earthquake would cause nearly $84 billion of economic loss, mostly building related. The scenario earthquake is estimated to kill 1000 to 4000 people, displace 64 thousand households, and generate 23 million tons of debris. HAZUS calculations show at least moderate damage on nearly 700 locations of highway bridges, 7 locations of airports, 6 schools, and over 343 thousand buildings. Other earthquakes with over $30 billion estimated losses in Northern California include: an M7.26 earthquake rupturing the entire Hayward-Rodgers Creek Fault causing $39 billion in losses and an M7.42 earthquake rupturing the Santa Cruz Mountain and Peninsula segments of the Northern San Andreas fault causing $36 billion in losses. The estimated loss ratio is less than 10% in most areas affected by scenario earthquake ground motions. However, in some cases, loss ratio exceeds 10% in the vicinity of the rupturing fault. For the three most damaging earthquakes on the Northern San Andreas fault segments, loss ratio as high as 50% was estimated for some census tracts in northeast San Mateo County. In Southern California, the most damaging scenario is the M7.1 earthquake on the Puente Hills fault. With building loss ratios up to 30% in some census tracts, total predicted building loss is $79 billion ($82.8 billion in total economic loss). HAZUS calculations suggest that the earthquake would kill 500 to 2000 people and displace 58-thousand households. The calculations show at least moderate damage to 569 highway bridges, 2 airports and over 470-thousand buildings. These numbers are in the lower range of an earlier estimate by Field and others for the same fault (2005, Earthquake Spectra). Other notable high loss scenarios for Southern California are: an M6.9 earthquake on the Newport-Inglewood fault with over $34 billion in total building loss, an M7.1 earthquake on the Palos-Verde Fault with over $20 billion in total building loss and an M7.8 rupture along the Southern San Andreas Fault with over $20 billion in total building loss. When the same scenario was studied in detail in the 2008 ShakeOut Scenario, the total building loss was higher at $35 billion. This is likely due to a more detailed building inventory and ground failure effects (liquefaction, landslides, and surface rupture).

Table 1. Top Five Most Damaging Scenarios Earthquakes in Northern and Southern California and Associated Economic Losses

			Economic Losses ($M)
	Scenario Earthquakes	M	Buildings Related	Transportation System	Utility System
Northern California
N1	Northern San Andreas Fault	7.90	79,834	1,436	2,583
	(SAS+SAP+SAN+SAO)¹
N2	Northern San Andreas Fault	7.76	70,628	1,172	2,026
	(SAS+SAP+SAN)¹
N3	Northern San Andreas Fault	7.83	66,216	1,162	1,856
	(SAP+SAN+SAO)¹
N15	Hayward-Rodgers Creek Fault	7.26	36,883	826	1,695
	(HS+HN+RC)¹
N4	Northern San Andreas Fault	7.42	34,299	721	1,212
	(SAF_SAS+SAP)¹
Southern California
S8	Puente Hills Fault	7.1	79,662	1178	1,966
S17	Newport – Inglewood Fault	6.9	34,319	482	958
S1	Verdugo Fault	6.7	23,751	270	826
S2	San Andreas Fault – Southern	7.8	20,515	503	1,489
S18	Palos Verdes Fault	7.1	20,084	367	796
¹These are section abbreviations. Click here to see full names.

There are significant and consistent differences in the estimated building related losses for the UCERF 2-based scenarios and for USGS scenarios (figure 2, for example). Among the ten UCERF 2-based scenarios, six can be compared directly to the corresponding USGS scenarios (similar earthquake magnitude and fault model). Comparison shows that the estimated building-related loss for a UCERF 2-based scenario is 28% to 63% lower than that for the comparable USGS scenario (Table 2). This difference is mainly due to different GMPEs used in the UCERF 2-based scenarios and in USGS scenarios.

Figure 2. Distribution of percent loss, calculated as the ratio of economic loss due to building damage to building replacement value times 100, for each census tract.

Table 2. Comparison of Estimated Building Losses for UCERF 2-based Scenarios and USGS Scenarios

Scenario Earthquakes	UCERF 2-Based Scenario		Comparable USGS Scenario		Difference
Scenario Earthquakes	M	Building Related Loss ($M)	M	Building-Related Loss ($M)	(%)
Southern San Andreas Fault (CO)^1,2	6.95	684	7.1	5,138	-87
San Jacinto Fault (SBV+SJV+A+C)¹	7.76	10,983	No comparable scenario
San Jacinto Fault (A+C)	7.49	2,654	No comparable scenario
Hayward-Rodgers Creek Fault (RC+HN+HS)¹	7.29	21,966	7.26	36,883	-40
Hayward-Rodgers Creek Fault (RC)¹	6.98	3,963	6.98	6,582	-40
Northern San Andreas Fault (SAO+SAN+SAP+SAS)¹	7.99	29,412	7.9	79,834	-63
Northern San Andreas Fault (SAP+SAS)¹	7.47	17,031	7.42	34,299	-50
Imperial	6.9	236	7	421	-44
Palos Verdes	7.2	14,513	7.1	20,084	-28
Newport-Inglewood³	7.2	42,980	6.9	34.319	25

¹These are section abbreviations. Click here to see full names.
²The lower UCERF 2 magnitude for this scenario may have contributed to greater difference in estimated loss.
³For this scenario, UCERF 2 used larger magnitude and included rupture along the Rose Canyon Fault.

Most of the scenario ShakeMaps used in the CGS 2005 study are identical to the ones used in the 2009 study for USGS scenarios. Therefore, for a given common ShakeMap scenario, comparison of loss estimates from the 2009 study and from the CGS 2005 study reflects differences in the HAZUS default inventory information on built environment and demographics and in HAZUS analytical models. Comparison is made for building-related losses only because the CGS 2005 study only reported building-related losses. In Northern California, 20 of the 33 comparable scenarios show building related losses from the 2009 study that are within 10% of 2005 estimates. For most of the scenarios on the Northern San Andreas Fault, the 2009 estimates are up to 40% higher than the 2005 estimates. In contrast, the 2009 estimates are up to 36% lower than the 2005 estimates for most other USGS scenarios. In Southern California, 75% of the common scenarios result in lower building related loss when estimated from 2009 study compared with the 2005 study. Half of the differences are between 30% and 71% (see more details).

Uncertainties are inherent in HAZUS loss estimates, just as they are inherent in all other hazard, risk, and loss estimate methodologies. Uncertainties arise from incomplete scientific knowledge concerning earthquake occurrence and ground motion characteristics and their effects on buildings and other facilities. They also result from incomplete or inaccurate inventories of the built environment, demographics, and economic parameters. A third source of uncertainties comes from the approximations and simplifications that are necessary for model analyses. It is estimated that these factors can result in total uncertainties in loss estimates produced by the HAZUS-MH Earthquake Model of up to a factor of two or more (FEMA, 2009).

Conclusions

Three of the four most damaging USGS scenarios involve co-seismic rupture of different combinations of the Northern San Andreas Fault segments. The second most damaging scenario is the rupture of the Puente Hills Fault, a blind-thrust fault directly beneath Los Angeles. This scenario has been studied in greater detail by Field et al. (2005, Earthquake Spectra).
Building related losses estimated using NGA GMPEs are 28% – 63% lower than those using ShakeMap ground motions in six comparable scenarios analyzed. Realistic loss estimation depends on accurate prediction of ground motions. Additional scenarios should be analyzed to evaluate further the implication of NGA models on loss estimation.
Comparison of estimated building related losses for the USGS scenarios from the 2009 study to those from the CGS 2005 study shows no consistent trend that could be attributed to changes in HAZUS methodology.

CGS is in the process of estimating annualized losses based on the 2008 update of the seismic hazard maps.

A complete set of selected HAZUS results for all the scenarios analyzed in the 2009 study can be found in Appendix A of the project report (download). For each scenario, the results include:

Maps

· Shaking intensity map (USGS ShakeMap)

· Peak ground acceleration by census tract (HAZUS map)

· Total building loss by census tract (HAZUS map)

· Loss ratio by census tract [ratio of building damage (includes structural damage and non-structural damage) to replacement value (exposed value of buildings)]

· Displaced households by census tract (HAZUS map)

· Debris generated by census tract (HAZUS map)

Tables

· Dam inventory and ground motions at dam locations

· Direct economic loss for buildings by county

· Short term shelter needs by county

· Casualties summary report by county

Other Summary Reports

· Global Summary Report

· Quick assessment reports

These results may also be viewed by clicking the scenario names in table 3 for Northern California, table 4 for Southern California, and table 5 for UCERF 2-based scenarios. Scenario earthquakes are listed by the abbreviations of fault and fault section names. Click here to see the full names.

Table 3. Northern California Scenario Earthquakes and Associated Economic Losses

	Scenario Earthquakes	M	Economic Losses ($M)
	Scenario Earthquakes	M	Buildings Related	Transportation System	Utility System
N1	SAF_SAS+SAP+SAN+SAO	7.90	79,834	1,436	2,583
N2	SAF_SAS+SAP+SAN	7.76	70,628	1,172	2,026
N3	SAF_SAP+SAN+SAO	7.83	66,216	1,162	1,856
N4	SAF_SAS+SAP	7.42	34,299	721	1,212
N5	SAF_SAS	7.03	6,789	118	353
N6	SAF_SAP	7.15	24,788	472	845
N7	SAF_SAN+SAO	7.70	17,814	426	910
N8	SAF_SAN	7.45	11,474	344	798
N9	SAF_SAO	7.29	161	32	100
N10	HRC_HS	6.67	14,469	221	613
N11	HRC_HN	6.49	7,655	180	518
N12	HRC_HS+HN	6.91	22,660	426	983
N13	HRC_RC	6.98	6,582	153	572
N14	HRC_HN+RC	7.11	19,244	508	1,163
N15	HRC_HS+HN+RC	7.26	36,883	826	1,695
N16	CLV_CS	5.78	168	12	23
N17	CLV_CC	6.23	2,666	45	135
N18	CLV_CS+CC	6.36	3,422	52	153
N19	CLV_CN	6.78	10,176	137	433
N20	CLV_CC+CN	6.90	13,534	182	528
N21	CLV_CS+CC+CN	6.93	14,217	196	540
N22	CGV_CON	6.25	2,644	76	288
N23	CGV_GVS	6.24	1,872	78	280
N24	CGV_CON+GVS	6.58	5,326	122	455
N25	CGV_GVN	6.02	528	48	115
N26	CGV_GVS+GVN	6.48	2,267	96	351
N27	CGV_CON+GVS+GVN	6.71	6,631	144	513
N28	SGF_SGS	7.0	963	19	80
N29	SGF_SGN	7.2	12,111	274	567
N30	SGF_SGS+SGN	7.4	15,005	361	742
N31	GNV_GS	6.6	1,752	59	191
N32	GNV_GN	6.7	3,198	79	319
N33	GNV_GS+GN	6.9	5,175	109	439
N34	MTD	6.65	7,296	118	472
N35	GV	?	3,503	111	513

Table 4. Southern California Scenario Earthquakes and Associated Economic Losses

			Economic Losses ($M)
	Scenario Earthquakes	M	Buildings Related	Transportation System	Utility System
S1	Verdugo Fault	6.7	23,751	270	826
S2	San Andreas Fault – Southern M7.8	7.8	20,515	503	1,489
S3	Chino Hills Fault	6.7	11,390	104	554
S4	San Joce -Carro Prieto Fault	7	267	17	213
S5	Hosgri Fault	7.5	1,020	47	232
S6	Elsinore Fault (Julian)	7.1	1,184	51	129
S7	North Channel Slope	7.4	5,814	140	553
S8	Puente Hills Fault	7.1	79,662	1178	1,966
S9	San Joaquin Hills Fault	6.6	16,557	165	482
S10	Elsinore Fault	6.8	2,481	64	190
S11	Raymond Fault	6.5	16,495	158	587
S12	Whittier Fault	6.8	15,965	155	649
S13	Imperial Valley Fault	7	421	37	341
S14	San Andreas Fault – Coachella Valley	7.1	5,138	152	287
S15	San Andreas Fault – Southern M7.4	7.4	9,138	279	897
S16	San Jacinto Fault	6.7	4,366	89	328
S17	Newport – Inglewood Fault	6.9	34,319	482	958
S18	Palos Verdes Fault	7.1	20,084	367	796
S19	Santa Monica Fault	6.6	16,308	99	490
S20	Rose Canyon Fault	6.9	9,771	183	456
S21	San Andreas Fault – 1857	7.8	12,586	367	1,257

Table 5. UCERF 2-Based Scenario Earthquake and Associated Economic Losses

	Scenario Earthquakes	M	Economic Losses ($M)
	Scenario Earthquakes	M	Buildings Related	Transportation System	Utility System
U1	SAF_CO	6.95	684	3	163
U2	SJ_SBV+SJV+A+C	7.76	10,983	224	798
U3	SJ_A+C	7.49	2,654	69	302
U4	HRC_RC+HN+HS	7.29	21,966	404	1,136
U5	HRC_RC	6.98	3,963	93	424
U6	SAF_SAO+SAN+SAP+SAS	7.99	29,412	503	1,857
U7	SAF_SAP+SAS	7.47	17,031	303	738
U8	Imperial	6.9	236	18	274
U9	Palos Verdes	7.2	14,513	324	782
U10	Newport-Inglewood	7.2	42,980	671	1,679