Why aren't earthquake hazard maps better. Seth Stein1, M Why aren't earthquake hazard maps better? Seth Stein1, M. Liu2, Edward M. Brooks1, Bruce D. Spencer3 1 Department of Earth & Planetary Sciences and Institute for Policy Research, Northwestern University 2 Department of Geological Sciences, University of Missouri 3 Department of Statistics and Institute for Policy Research, Northwestern University

Society is playing a high-stakes game of chance against nature We want to - assess the hazard - how often major earthquakes happen and how bad they will be mitigate or reduce the risk - the resulting losses. - Our limited knowledge about what the earth will do, and limited resources, limit how well we can do either Society can still take sensible measures to reduce damage in future events

Ideally, how much mitigation is enough? Societally optimal level minimizes total cost = sum of mitigation cost + expected loss Expected loss = ∑ (loss in ith expected earthquake x its assumed probability) Expected loss depends on event & mitigation level Compared to optimum Less mitigation decreases construction costs but increases expected loss and thus total cost More mitigation gives less expected loss but higher total cost Stein & Stein, 2012

Limitation1 (science): Inaccurate hazard and loss (damage) estimates produce nonoptimal mitigation Large uncertainties remain despite scientific & engineering advances Stein & Stein, 2013

Still, undermitigation is better than no mitigation Limitation 2 (economics): Even without uncertainty, mitigation usually isn’t optimal because of limited resources & other needs Still, undermitigation is better than no mitigation Communities should do what they can

Hazard assessment failed 2010 map predicts probability of strong shaking in next 30 years But: 2011 M 9.1 Tohoku, 1995 Kobe M 7.3 & others in areas mapped as low hazard In contrast: map assumed high hazard in Tokai “gap” Geller 2011

Hazard model divided trench into segments These were assumed to break individually in future earthquakes Expected Earthquake Sources 50 to 150 km segments M7.5 to 8.2 (Headquarters for Earthquake Research Promotion)

Giant earthquake broke many segments 2011 Tohoku Earthquake 450 km long fault, M 9.1 Expected Earthquake Sources 50 to 150 km segments M7.5 to 8.2 (Headquarters for Earthquake Research Promotion) (Aftershock map from USGS) J. Mori

NY Times 3/21/11

Hazard maps are hard to get right: successfully predicting future shaking depends on accuracy of four assumptions over 500-2500 years Where will large earthquakes occur? When will they occur? How large will they be? How strong will their shaking be? Uncertainty & poor performance can result because these are often hard to assess

Africa-Eurasia convergence rate varies smoothly (5 mm/yr) Slow plate boundary Africa-Eurasia convergence rate varies smoothly (5 mm/yr) NUVEL-1 Argus, Gordon, DeMets & Stein, 1989 Predicted hazard from historic seismicity is highly variable GSHAP 1999 Swafford & Stein, 2007 11

Africa-Eurasia convergence rate varies smoothly (5 mm/yr) Slow plate boundary Africa-Eurasia convergence rate varies smoothly (5 mm/yr) NUVEL-1 Argus, Gordon, DeMets & Stein, 1989 Predicted hazard from historic seismicity is highly variable Likely overestimated near recent earthquakes, underestimated elsewhere 2003 M 6.3 2004 M 6.4 GSHAP 1999 Swafford & Stein, 2007 12

EARTHQUAKE RECURRENCE IS HIGHLY VARIABLE Sieh et al., 1989 Extend earthquake history with paleoseismology M>7 mean 132 yr s 105 yr Estimated probability in 30 yrs 7-51% 13

constant with time (time-independent) or We don’t know whether to assume that probability of a major earthquake is constant with time (time-independent) or small after a large earthquake and then increases (time-dependent ). Time dependent predicts lower until ~2/3 mean recurrence Results depend on both model choice & assumed mean recurrence Hebden & Stein, 2008

Have We Seen the Largest Earthquakes in Eastern North America? Largest known east of Appalachians is 1886 Charleston SC, M ~7, recurrence time ~ 500 yrs Could there be bigger? M 7.2 -7.4 occur offshore Canadian east coast What do past event locations tell us?

Simulations address issue of short record record needed to see real hazard Swafford & Stein, 2007 1933 1929

Given 300 years of data, what Mmax would we observe if Mmax were really 7.0, 7.2, 7.4, 7.6? Simulated seismicity – 10,000 catalogs Generally miss largest events and so underestimate Mmax Most likely to observe Mmax ~7 with recurrence time ~ sample length Events larger than observed to date in ENA are possible Locations of large events to date need not indicate higher hazard

Sensitivity analysis - predicted hazard depends on - Assumed magnitude of largest future events Assumed ground prediction motion model Neither are well known since large earthquakes rare 180% 275% Newman et al., 2001

Seismic hazard uncertainty typically factor of 3-4 At best partially reducible because some key parameters poorly known, unknown, or unknowable Uncertainties not communicated to users Stein et al, 2012

Compare 510-year shaking record to Japanese National Hazard (JNH) maps Miyazawa and Mori, 2009 The short time period since hazard maps began to be made is a challenge for assessing how well they work. Hindcasts offer long records, but are not true tests, as they compare maps to data that were available when the map was made. Still, they give useful insight. Brooks et al., 2015

Uniform & random maps Geller (2011) argued that “all of Japan is at risk from earthquakes, and the present state of seismological science does not allow us to reliably differentiate the risk level in particular geographic areas,” so a map showing uniform hazard would be preferable. Test: By the one measure uniform and randomized maps do better, but by another the detailed JNH maps perform better. JNH worse JNH better Brooks et al., 2015

Intermediate level of detail may be better for hazard maps. Maps may be overparameterized (overfit) A high order polynomial fits past data better than linear or quadratic models, but this more detailed model predicts the future worse than the simpler models. Intermediate level of detail may be better for hazard maps.

NEPAL ILLUSTRATES Because plausible alternative parameter choices yield quite different maps, maps have significant uncertainties. Shaking variations from past earthquakes are useful for characterizing site effects in hazard maps. Locations of past earthquakes may not be useful for assessing locations of future ones.

At oceanic subduction zones, GPS data show variation in coupling that seem to indicate locations of future earthquakes and likely reflect locations of past ones Loveless & Meade, 2011 Avouac et al., 2015 Nepal GPS data show no significant variation in coupling between areas of recent large earthquakes, or the 2015 earthquake Moreover, earthquakes in past few hundred years have released less plate motion than is accumulating. Hence with present knowledge, the entire zone can be regarded as equally hazardous and perhaps vulnerable to much larger earthquakes than those currently known, with long recurrence times.

Communities should do what they can Although major scientific questions remain, they don’t need to be solved to make sensible mitigation policies, given available resources & other needs Communities should do what they can