Earthquake scenario development in conjunction with the 2023 USGS National Seismic Hazard Model

The National Seismic Hazard Model (NSHM) developed by the US Geological Survey (USGS) has served many different users throughout the years ranging from industry to academia and policymakers. This probabilistic model considers a wide range of seismic sources, resulting ground motions, and uncertainties associated with each. As such, the USGS continually updates this model with the latest science to provide the best possible estimates of seismic hazard for the country. The 2023 update to the NSHM (Petersen et al., 2023) supersedes the 2018 NSHM (Petersen et al., 2020) and features updates to the earthquake catalog, earthquake rupture forecast, ground-motion models (GMMs), and site/basin amplifications, among other features. To communicate the impact of the updates, we present deterministic earthquake scenarios for select urban regions across the United States.

Earthquake scenarios serve a range of local and regional needs, from developing proactive mitigation strategies to enhancing post-disaster emergency response. Past scenario efforts, such as ShakeOut (Jones et al., 2008) and HayWired (Hudnut et al., 2018), examined the impacts to large earthquakes in California. ShakeOut specifically helped provide information to stakeholders on the risk and potential impacts caused by a large earthquake centered in southern California. The study helped highlight the consequences to local communities by featuring in-depth research from hundreds of scientists and engineers representing government, academia, and industry partners to focus on assessing broad community-wide impacts to serve local needs. Similarly, the M9 Project (Frankel et al., 2018; Wirth et al., 2021) produced synthetic ground-motions records from a simulated moment magnitude (M_W) 9 earthquake occurring on the Cascadia subduction zone (CSZ), which provided insight into potential impacts from a large interface earthquake in the Pacific Northwest. Other scenario efforts that have been tied to the NSHM include the Building Seismic Safety Council 2014 (BSSC2014) catalog (Thompson et al., 2016) and numerous scenario legacy catalogs for regions across the United States (U.S. Geological Survey (USGS), 2017). The BSSC2014 set was developed as a deterministic subset of fault rupture sources from the 2014 NSHM (Petersen et al., 2015) and considered the source models and GMMs used in that version of the NSHM. The legacy catalogs include scenario sets for Alaska, Cascadia, Hawaii, Montana, northern California, Nevada, Oklahoma, southern California, Utah, and a global set. These legacy catalogs were developed at varying points of time and for different uses.

The scenarios developed in the present investigation have a slightly different scope than the previous scenario studies in the region. This work focuses on the development of a handful of scenarios located at urban centers across the United States that have seen either substantial changes in the 2023 NSHM Update or have implications in terms of risk from changing hazard. As such, we produce scenarios for Hawaii, Utah, Alaska, and Virginia. Focus is given to impact beyond the seismic hazard, including economic loss, estimated fatalities, and the probability of landslides and liquefaction. In addition to presenting scenarios and their impacts, we discuss the lessons learned throughout the development process including updates to the 2023 NSHM itself. Differences in impacts are explored for NSHM modeling decisions for the Wasatch fault zone (WFZ) beneath Salt Lake City (SLC) and for the gridded seismicity in the northern Hawaiian Islands (defined as all of the islands except the Island of Hawai‘i).

The earthquake scenarios developed as a part of this study are composed of three main components: (1) the new source and ground-motion characterization in the 2023 NSHM, (2) existing scenario efforts or catalogs, and (3) input from working groups and consultation with local experts and stakeholders. These three pieces shaped the scenarios developed for the four regions, and additional details on the specifics for each location are discussed in the following sections of this article.

Locations for scenario development focused where the NSHM received substantial updates from previous cycles and where there are major urban regions with a high seismic risk. In the 2023 NSHM, both Hawaii and Alaska benefited from major improvements with the development of new statewide seismic hazard models, motivating their selection for this study. The SLC region received consideration for scenario development due to moderate updates in the 2023 NSHM and the high seismic risk posed by the nearby WFZ. Virginia was selected as a site to showcase the updates to the seismic hazard modeling of the central and eastern United States (CEUS). For all the scenarios developed, the underlying science used, including the source models and GMMs, are consistent with the 2023 NSHM. The updated fault and seismic zone models (Bender et al., 2021; Hatem et al., 2022; Thompson Jobe et al., 2022a, 2022b) were used to develop scenarios in Utah, Alaska, and Virginia. In Hawaii, the new model includes an updated seismicity catalog, and seismicity originating from décollement faults and the Kīlauea volcano’s caldera collapse (Llenos and Michael, 2022; Petersen et al., 2022). The GMMs used in the work by Petersen et al. (2023) were used within the USGS ShakeMap (Worden et al., 2020) with site soil amplification using the Heath et al.’s (2020) Vs 30 data however without using any non default basin factors to estimate median ground shaking for all scenarios as well.

When considering the development of the scenarios themselves, one tool we used was the consideration of past scenario efforts conducted in the subject region to help guide our list. For example, in Hawaii, we reviewed and updated the Hawaii Legacy Catalog (USGS, 2017) and then developed scenarios of our own, independent of the scenarios in that catalog. In SLC, we consulted the Earthquake Engineering Research Institute (EERI) M_W 7.0 Wasatch Fault Scenario (Pankow et al., 2015) and the work conducted by the Working Group for Utah Earthquake Probabilities (WGUEP; Wong et al., 2016). In Alaska, we consulted the Alaska Legacy Catalog (USGS, 2017) and Fozkos et al. (2023). Finally, we also considered past historical earthquakes in scenario development for all the regions. Earthquakes such as the 2002 M_W7.9 Denali and the 2011 M_W5.8 Mineral events were major influences for the scenarios developed in Alaska and Virginia.

Finally, we incorporated direct inputs from the scenario development working groups to identify specific source and their seismicity potential in Hawaii, Utah, Alaska, and Virginia. The working groups were composed of government officials, academic researchers, and members of industry who were experts in the study region being discussed. The variety of participants ensured the scenarios could be used to fulfill multiple needs. In each case, we first presented the background information on preliminary 2023 NHSM updates available at the time of deliberations, discussed existing scenario efforts in the respective study region, and then presented preliminary set of scenarios to jumpstart the discussion. We then held a discussion period to gather feedback and recommendations from the group members.

As mentioned in earlier sections, Hawaii was selected as a location for scenario development due to the major updates from the previous version by Klein et al. (2001). The 2023 NSHM update incorporates the new 2021 model for the State of Hawaii (Petersen et al., 2022). Key features of Petersen et al. (2022) include updates to the earthquake catalog, updated GMMs, and a new caldera collapse model.

The updated hazard model separates the Hawaiian Islands into two regions: (1) southern (the Island of Hawai‘i), and (2) northern (all other islands). The northern islands have no mapped faults in the 2023 NSHM, resulting in a seismic hazard completely governed by gridded seismicity. The southern region, consisting of the Island of Hawai‘i, features previously delineated seismic sources (e.g. décollements, volcanic caldera). In addition, the seismic catalog for both regions is separated into deep earthquakes (hypocentral depths > 20 km) and shallow earthquakes (hypocentral depths ≤ 20 km), for which there are separate GMMs in each case (Petersen et al., 2022). Honolulu is located on the island of O‘ahu, in the northern region, and therefore, its seismic hazard is dominated by gridded seismicity (58% shallow and 41% deep). In contrast, the hazard for Hilo on Island of Hawai‘i, in the southern region, has contributions from deep gridded sources (84%) and the décollement on the southern flank of the island (16%). The seismic hazard disaggregations are shown for Honolulu and Hilo in Figure 1a and b, respectively.

As shown in Figure 1, Honolulu has a substantial hazard stemming from M_W 5 to M_W 6.5 earthquakes at ∼10 and 50 km away, with the largest contributor (3.5% of total hazard) from an M_W 5.5 earthquake occurring ∼14 km from the downtown Honolulu. Meanwhile, Hilo has substantial contributions from M_W 6 to 8 earthquakes occurring about 50 km away, with the largest contributor (13.8% of total hazard) being an M_W 7.3 event at ∼48 km from downtown Hilo.

Using this model, input from existing scenario catalogs and working group feedback, we developed a suite of eight new earthquake scenarios. These scenarios are shown in Figure 2 and listed in Table 1. The three scenarios developed in the northern islands were run with two hypocentral depths of 10 and 30 km. The scenarios run for 10-km depth featured different GMMs than the 30-km depth scenarios, consistent with the methods presented in the work by Petersen et al. (2022). Differences between the impacts between the two depths are further explored later in this article.

The M_W 7.5 Lāna‘i scenario was developed considering the findings of Butler (2020), who explored the 1871 Lāna‘i earthquake that likely occurred in the Molokai fracture zone (MFZ) south of O‘ahu. Similarly, the M_W 7.5 Maui scenario was developed considering a possible rupture on a nearby section of the MFZ. The M_W 6.5 O‘ahu scenario is similar to an M_W 6.0 scenario initially developed for the Hawaii Legacy Catalog (USGS, 2017), but it was moved closer to Honolulu and increased to M_W 6.5 to be consistent with the maximum magnitude modeled in the work by Petersen et al. (2022). Finally, the South and West Flank scenarios were developed to be consistent with the rupture of décollement faults on the southern and western flanks of the Island of Hawai‘i, respectively. These scenarios, along with an update to the USGS’s Hawaii Legacy Catalog (USGS, 2017) to be consistent with GMMs used in the work by Petersen et al. (2022), were presented to the Hawaii Earthquake and Tsunami Advisory Council (HETAC) for consideration for use by state officials.

Impacts from each scenario were also estimated using ShakeMap (Worden et al., 2020) to model the rupture and estimate ground shaking, Prompt Assessment of Global Earthquakes for Response (PAGER; Earle et al., 2009; Jaiswal et al., 2009, Wald et al., 2010) to estimate economic losses and fatalities, and Ground Failure (Allstadt et al., 2022; Nowicki Jessee et al., 2018; Zhu et al., 2017) to estimate the likelihood of landslides and liquefaction. The most likely range of economic losses and fatalities from PAGER is shown for each scenario in Table 1 above. Figure 3 shows impacts estimated for the M_W 7.5 Lāna‘i scenario (10-km hypocentral depth) for both ground shaking intensity (Figure 3a) and landslides (Figure 3b). Ground shaking intensities for this event were estimated to be Modified Mercalli Intensity (MMI) VIII in Kahului, MMI VI in Honolulu, and MMI IV in Hilo. Economic losses were estimated to be most likely between US$1 and US$10 billion with the highest probability of 10 or fewer fatalities. USGS Ground Failure estimated the likelihood of liquefaction to be very low but estimated a high probability of landslides on the islands of Maui, Moloka‘i, and Lāna‘i (Figure 3b).

The SLC region was selected as an important location for scenario development due to the high seismic hazard posed by the WFZ combined with the relatively vulnerable building stock (e.g. presence of large fraction of unreinforced masonry (URM) dwellings). In addition, the incorporation of sedimentary basin effects for the Utah in the 2023 NSHM also served as motivation for our analysis.

The seismic hazard in Utah is dominated by the WFZ, a west-dipping normal fault with a history of producing M_W 7 + surface rupturing earthquakes. The West Valley fault zone (WVFZ) also lies to the west of downtown SLC and poses an additional hazard. Recently, the 2020 M_W 5.7 Magna earthquake occurred in the vicinity of the WVFZ, although research indicates this event occurred on the SLC segment of the WFZ (Pang et al., 2020). The location and importance of the 2020 M_W 5.7 Magna event and how it changes our understanding of the WFZ is explored in more detail later in this article. The hazard disaggregation for a site located in downtown SLC is shown in Figure 4. Disaggregation results show both M_W 6.0 and 6.9 earthquakes from WVFZ and WFZ at ∼3–5 km away from the site, each contributing more than 20% to the total hazard. Other major urban areas, including Brigham City, Ogden, and Provo, also have a high seismic hazard from the WFZ.

As discussed in previous sections, the scenario development for Utah was informed by existing scenario efforts (Pankow et al., 2015), the WGUEP (Wong et al., 2016), and a working group convened by the USGS team for the effort presented in this study. Resulting scenarios include ruptures on the SLC segment and the WFZ, both with and without a synchronous West Valley fault (WVF) rupture, and other WFZ ruptures on the Brigham City, Weber, and Provo segments. These scenarios are shown and described in Figure 5 and Table 2. The scenario rupture lengths and magnitudes are largely a product of the WGUEP results, which provide WFZ characteristic magnitudes for each segment. The scenarios developed in the study feature full ruptures of the SLC, Brigham City, Weber, and Provo segments. In addition, complex ruptures of multiple segments or partial segments are considered. Finally, the inclusion of a synchronous WVF rupture along with an SLC segment rupture was also considered based on the paleoseismic evidence that the two faults had ruptured twice in the Holocene (Wong et al., 2016).

As was done for the Hawaii scenarios in the previous section, impacts from each scenario were again estimated using ShakeMap, PAGER, and Ground Failure. The most likely ranges of economic losses and fatalities from PAGER are shown for each scenario in Table 2. Furthermore, the impacts from the M_W 7.0 South SLC and North Provo scenario are shown in Figures 6 to 8. Figure 6a shows the surface rupture location and provides information on the magnitude (M_W 7.0), depth (0–15 km), and dip angle (50° west). The resulting estimated shaking intensities are shown in Figure 6b, with expected MMI VIII in SLC and Provo. PAGER estimates of economic losses are likely to be above US$100 billion with an estimated 10–100 fatalities (Table 2).

Almost 2 million people are expected to experience shaking of MMI VII and stronger, with 1.6 million experiencing MMI VIII and MMI IX. Ground Failure estimates a high probability of both landslides and liquefaction as illustrated in Figure 7. Similarly, we also present the M_W 7.1 Wasatch SLC scenario loss estimates obtained using the Federal Emergency Management Agency’s (FEMA) Hazus software, which provides an economic loss estimate of US$75 billion to the built environment (Figure 8).

Alaska was selected as a location for scenario development due to the major overhaul to the state seismic hazard model conducted for the 2023 NHSM. The updated 2023 model supersedes the prior 2007 Alaska hazard model by Wesson et al. (2007) and features updates to the earthquake catalog, gridded seismicity sources, crustal model, subduction zone interface and inslab source models, magnitude scaling relations, and GMMs.

The seismic hazard for a site in downtown Anchorage has contributions from interface, inslab, and crustal sources. In the 2023 NSHM, the subduction zone geometry is divided into 12 sections with multiple down-dip widths, updating the 2007 model that had seven sections with a single down-dip width. The Prince William Sound (PWS) section of the new subduction geometry contributes substantially to the hazard in Anchorage. In addition, the 2023 NSHM features updated faults and fault zones (85 + fault sections), including prominent faults, such as the Denali fault and Castle Mountain fault (CMF), which are close to Anchorage (Haeussler et al., 2023). The interface and inslab hazard contributions dominate the 2475-year PGA for a B/C site condition as shown in Figure 9. The largest contributor (23%) to the total hazard is from an M_W 8.5 interface earthquake occurring at 46 km from the downtown Anchorage.

Similar to Hawaii and SLC, we convened a working group to provide insight into what would be useful for the community. Using the input (suggestions of location, magnitude, and rupture plane characteristics) of this working group and past scenario efforts (Fozkos et al., 2023; USGS, 2017), we developed 10 new scenarios described in Figure 10 and Table 3.

Particular focus was given to scenarios in the Anchorage region, as Anchorage is the most populated city in Alaska and the port of Anchorage manages a substantial fraction of goods that come into the state. Crustal scenarios on the Castle Mountain, North Cook Inlet, and Broad Pass Thrust faults (Figure 10b) were selected to give insight into how moderate-to-large crustal earthquakes occurring close to downtown Anchorage could impact the city. Similarly, large inslab and interface earthquakes centered beneath Anchorage were also developed (Figure 10a). The inslab event was developed to provide insight if a similar (but larger and shallower) event to the 2018 M_W 7.1 earthquake were to occur, while the interface event was developed to investigate the impacts of a large interface rupture of the PWS section southeast of the city. The Denali fault scenario was developed to explore the impacts of a similar M_W 7.9 strike-slip event to the 2002 M_W 7.9 Denali earthquake, except with the western section of the fault rupturing. Finally, the Queen Charlotte fault scenario was developed to provide insight into a larger earthquake close to Juneau and oil and gas operations in the southern part of the state. Notably, the inclusion of a scenario for large interface earthquake similar to the 1964 M_W 9.2 event was considered, but we decided to omit it from working group discussions because many past scenario efforts have focused on this type of event.

The M_W 7.5 Anchorage Inslab scenario earthquake and its impacts are shown in more detail in Figures 11 and 12. The scenario was developed considering the 2018 M_W 7.1 earthquake, past scenario efforts—including the Alaska Legacy Catalog (USGS, 2017) and Fozkos et al. (2023)—and the updates to the 2023 NSHM. The rupture plane used was informed by all these sources and the Slab2 geometry model (incorporated in the 2023 NSHM) was developed by Hayes (2018). The rupture chosen for this study in consultation with the working group was larger and shallower than the 2018 M_W 7.1 earthquake. The resulting impacts feature an estimated MMI of VIII in Anchorage with approximately 406,000 people exposed to that shaking intensity (Figure 11b). PAGER losses are estimated to be likely between US$10 and US$100 billion with the highest probability of 10 or fewer fatalities. Note that the PAGER loss estimates are from direct shaking-induced damage only and does not account for potential impacts from earthquake-induced tsunami. USGS Ground Failure estimates that there is a moderate probability of both landslides and liquefaction, although this would likely be occurring away from densely populated areas (Figure 12).

The CEUS had major updates in the 2023 NSHM with updates to fault sources, the earthquake catalog, GMMs, and site amplification models. In addition, new seismic zones were added in the New Madrid seismic zone to capture sources from observed seismicity where fault structures were previously not well constrained. One example is the Central Virginia seismic zone (CVSZ), which hosted the 2011 M_W 5.8 Mineral earthquake (Pazzaglia et al., 2021; Thompson Jobe et al., 2022a; Tuttle et al., 2021). The earthquake was thought to have occurred on the Quail fault, a blind and poorly constrained reverse fault. The Quail fault was not modeled explicitly in the 2023 NSHM, as its length and location are not well understood. Based on the occurrence of this earthquake and findings by Tuttle et al. (2021), which indicate one to two historical events of M_W 6–6.5 occurred in this region, two seismic zone polygons were included in the 2023 NSHM to represent the CVSZ. These new polygons are composed of one local polygon containing the location of the 2011 M_W 5.8 and another regional polygon containing areas of other historical seismicity. Washington, D.C. and Richmond, Virginia are two of the major urban centers close to the CVSZ. The seismic hazard in Washington, D.C. is dominated by gridded seismicity (95%) for PGA at the 2475-year return period, while the CVSZ contributes 5% to the city’s hazard. Accordingly, seismic disaggregation (Figure 13a) shows the earthquake that contributes to Washington, D.C.’s hazard is an M_W 4.9 at a distance of ∼13 km from the city. Similarly in Richmond, gridded seismicity contributes 87% to the total hazard, while the CVSZ contributes 13%. The earthquake that contributes most to Richmond’s total hazard is an M_W 4.9 at ∼11 km (Figure 13b). It is worth noting that the minimum magnitude considered for PSHA is typically M_W 5.0 for building and nuclear power plant design (Electric Power Research Institute (EPRI), 2005); however, the NSHM considers a lower threshold of M_W 4.7 in the CEUS.

We developed a single earthquake scenario originating from the CVSZ for this region. The scenario was informed by the 2023 NSHM science and discussions with regional experts. We repeat the 2011 Mineral earthquake (location and depth), except an increase in its magnitude to M_W 6.5. An M_W6.5 scenario was selected since it was the upper range of the maximum magnitude distribution (0.2 weight) for the CVSZ in the 2023 NSHM. Tuttle et al. (2021) suggests that one possible scenario of a past earthquake is an M ≥ 6.4 occurring in the Mineral area would explain paleoliquefaction observations on the Pamunkey River (∼65 km away). Per the discussion in the work by Thompson Jobe et al. (2022a), we decided to keep the scenario modeled as a point source because the location, orientation, and size of the Quail fault are highly uncertain. It is also worth noting that the ShakeMap produced for the 2011 M_W 5.8 Mineral earthquake (ShakeMap ID: se609212) modeled the event as a point source, with no finite fault.

The M_W 6.5 scenario results in an “Orange” PAGER alert with estimated losses likely being between US$100 million and US$1 billion (median loss estimates ∼US$768 million) and an estimated 10 or fewer fatalities. Richmond is estimated to experience shaking with MMI VI, while Washington, D.C. is estimated to experience MMI V. In total, approximately 26.7 million people are estimated to experience MMI V, about 2 million are estimated to experience MMI VI, and 159,000 are estimated to experience MMI VII. For comparison, the 2011 M_W 5.8 earthquake resulted in an estimated US$200–US$300 million in losses with no reported fatalities (Horton et al., 2015). Figure 14 shows the location of the CVSZ and 2011 M_W 5.8 event and the estimated ground-motion intensities from the M_W 6.5 scenario. Figure 15 shows the estimated probabilities of landslides and liquefaction.

We obtained several valuable insights about the hazard characterization in Hawaii, SLC, Alaska, and Virginia during our scenario identification exercise. In almost all cases, the discussions held in the working group meetings were critical in formulating new scenarios and revising the candidate scenarios presented to the working groups. In the following subsections, two examples, one in SLC and one in Hawaii, are discussed in detail.

The characterization of the WFZ beneath SLC (i.e. the northern section of the SLC segment) in our scenario development proved to provide a critical learning experience and opportunity to improve the 2023 NSHM. The 2023 NSHM (and previous versions) only modeled fault sections as singular planes and did not capture potential listric geometries (the flattening of the fault’s dip with depth). The 2018 NSHM (Petersen et al., 2020) modeled this section of the Wasatch fault as a 50° west-dipping normal fault. The occurrence of the 2020 M_W 5.7 Magna earthquake sparked a renewed discussion about the listric nature of the fault (Pang et al., 2020). Figure 16 shows the postulated characterization of the northern SLC segment and the location of the 2020 M_W 5.7 Magna earthquake in black. Through discussions in the working group and in the work by Pechmann et al. (2023), it was determined that the 50° west dip used in the 2018 NSHM (Petersen et al., 2020) should be updated to reflect the newly postulated fault characterization. The multiple dip angles represented by red dashed lines (Figure 16) were partially the product of the discussions of Pechmann et al. (2023) and reflect the updated characterization of this segment of the Wasatch fault used in the work by Petersen et al. (2023). As shown, a 35° west dip was selected as the highest weighted planar representation of the hypothesized listric fault to be used in the 2023 NSHM.

We further explore the implications of the modeled fault dip (35° versus 50° or listric geometry) through estimated impacts by comparing the three different ruptures for the M_W 6.6 North SLC scenario (Figure 5 and Table 2). Figure 17 shows the differences in PGA (Figure 17a), with the scenario featuring a 35° dip, estimating 0.10 -0.14 g more than the 50° dip rupture (2023 NSHM versus 2018 NSHM). Similarly, spectral acceleration at 0.3 s (SA(0.3 s)) was estimated to be 0.25 - 0.35 g more (Figure 17b), and MMI was estimated to increase to 0.7 units. As such, when comparing MMI from the 50° rupture to the 35° rupture, estimated MMI in SLC increases from VIII to XI and from V to VI in Provo. The consideration of the fault as a 35° dip results in higher hazard in comparison to a 50° dip.

We also compare estimated ground motions between the modeled 35° planar dip and the listric geometry shown in Figure 16. Generally, the estimated ground-motion intensities for the modeled 35° dip are stronger in downtown SLC (i.e. when the 35° plane is closer to the surface than the listric representation) and weaker farther west of downtown SLC (i.e. when the 35° plane is farther from the surface than the listric representation), as observed in Figures 17 and 18. The 35° dip scenario estimates PGA and SA(0.3) to be as much as 0.16 and 0.37 g more, respectively, in downtown SLC. Conversely, the 35° dip scenario yields PGA and SA(0.3) estimates of 0.19 and 0.42 g less, respectively, in areas west of downtown SLC. When considering downtown SLC has a much higher population and economic exposure when compared to western Utah, the 35° planar dip model could produce more costly impacts when compared to the hypothesized listric geometry shown in the work by Pang et al. (2020). These estimates are especially important for structures that are more sensitive to high-frequency energy, such as URM, that is prevalent across much of SLC.

When considering longer periods, such as SA(1.0) and SA(3.0), we observe similar trends to PGA and SA(0.3). Comparing the 35° dip to the listric representation, SA(1.0) is estimated to be up to 0.22 g higher in downtown SLC and up to 0.25 g less west of the city. Meanwhile, SA(3.0) is estimated to be up to 0.04 g higher in downtown SLC and up to 0.08 g less west of the city. However these estimates do not include potential amplifications from the sedimentary basin in the region, and as such, differences at long periods could be even higher. Moschetti et al. (2021) found that sites in the deepest part of the basin could amplify ground motions substantially, with greater amplifications at longer periods ≥ 1 s. These types of estimates are important for tall buildings, like those in downtown SLC, which are more susceptible to strong long-period ground motions.

Table 4 shows the PAGER-estimated populations exposed to MMI VII, VIII, and IX for the M_W 6.6 North SLC rupture scenarios featuring a 35° west dip, a 50° west dip, and a listric geometry. When comparing the 35° dip to the 50° dip scenario, the 35° dip scenario yields estimates of 72,000 more people exposed to at least MMI VII. The largest difference was an estimated 115,000 more people exposed to MMI IX when the rupture was modeled to have a 35° dip. Notably, an estimated 104,000 fewer people are predicted to experience MMI VIII when the dip angle is modeled at 35° as opposed to 50°. It is likely that most of this population would experience an increase in shaking, to MMI IX, when the rupture plane is changed to 35° from 50° (Table 4). In general, the 35° dip scenario is predicted to result in similar population exposed to MMI VII, MMI VIII, and MMI IX. The largest differences are the 35° dip scenario estimates 37,000 more people exposed to MMI IX than the listric scenario, and 89,000 fewer people exposed to MMI VIII.

Note that the Wong et al.’s (2022) study found that the ground motions estimated close to the Northern SLC segment (i.e. downtown SLC) are not heavily influenced by the fault geometry (listric versus planar). However, their findings are similar to the results of this study about the effects of fault geometry at greater distances away from the WFZ, near Magna, Utah, as shown in Figure 18. Differences between the estimates of ground-motion estimates close to the WFZ likely stem from the time-averaged shear-wave velocity in the top 30 m (Vs30) used in both models. Wong et al. (2022) used a Vs30 of 300 m/s, whereas this study used default Vs30 estimates based on Heath et al.’s (2020) global mosaic Vs30 implemented within the ShakeMap 4.0 (Worden et al., 2020).

As noted, the hazard in the northern Hawaiian Islands, including Honolulu (Figure 1a), is dominated by gridded seismicity because no mapped faults are in this northern region in the 2023 NSHM. This proved to be a difficult factor when developing earthquake scenarios in this area. There are a couple of techniques that have been used in past studies to deal with this issue. In one example, Lin et al. (2022) developed scenarios based on a user-defined consequence, using a grid of possible earthquakes (modeled as point sources) with a range of magnitudes to determine which combinations produced the targeted consequence (e.g. bridge damage in the case of the work by Lin et al., 2022). Because our study did not have any predefined consequences of interest, we went a different route of holding working groups to gather feedback from local experts and stakeholders.

As shown in Table 1, the three scenarios in the northern islands were modeled with hypocentral depths of 10 and 30 km. These two depths were chosen due to the uncertainty of the seismic hazard in the region. As shown in Figure 19, Petersen et al. (2022) divided the earthquake catalog into earthquakes with depths shallower than 20 km and depths greater than 20 km, each with its own GMMs. The center of the distribution of earthquake depths from 0 to 20 km is approximately 10 km and is approximately 30 km for the depths from 20 + km. As such, we selected these depths for possible hypocenters. We also consulted Petersen et al. (2022), the USGS Hawaii Legacy Catalog scenarios (USGS, 2017), input from the working group, and the HETAC when determining the magnitude and location of the potential scenarios.

Here, we look further into the changes in impacts from the different hypocentral depths for the M_W 6.5 O‘ahu scenarios. Figure 20 shows the difference in estimated PGA and SA(0.3) between the 30- and 10-km depth M_W 6.5 O‘ahu scenarios. The PGA for the 10-km deep scenario has a maximum of 0.14 g stronger shaking and is approximately 0.05 g stronger in the city of Honolulu. The 10-km deep scenario results in MMI VIII in Honolulu, with an estimated economic loss of US$1–US$10 billion and 10 or fewer fatalities. This results in an estimated “Red Alert” in the PAGER system. Almost 800,000 people are estimated to experience shaking of MMI VII or MMI VIII. The 30-km deep scenario has an estimated MMI VII in Honolulu with an economic loss of US$100 million to US$1 billion and 10 or fewer fatalities. This results in an estimated “Orange Alert” with approximately 672,000 people experiencing MMI VII or VIII.

Through the scenario development process and the examples discussed in the previous sections, the uncertainty in the hazard model and its propagation to the process of selecting scenario characteristics can have a large downstream influence on the final estimated impacts on a community. Its influence on our scenario selection is apparent in both the development of the scenarios in SLC and Hawaii. As shown, the dip angle of the SLC segment of the Wasatch fault has a large influence on the potential for high shaking intensities throughout the city. A shallower dip results in stronger ground motions that more people could experience as violent shaking in the M_W6.6 scenario. However, the actual location of the Wasatch fault underneath SLC is highly uncertain, and we relied on the current preferred fault model within the 2023 NSHM to inform our scenario development. Similarly, in the northern Hawaiian Islands, the lack of mapped faults results in gridded seismicity as the dominating hazard source for Honolulu. As shown, a 3-km-deep M_W 6.5 event results in “Orange Alert,” while the same 10-km-deep event results in a “Red Alert.”

Scenario exercises such as these can help inform users and highlight important facets of the underlying mechanism that may not have been obvious using probabilistic methods. For example, metrics such as average annual loss (AEL) can be helpful for determining insight into which areas of a region are most vulnerable to seismic risk (Jaiswal et al., 2023), but AEL can be lacking when trying to assess which underlying factors contribute most to these potential impacts. For example, URM is a prevalent and vulnerable building type found in the SLC area. Typically, URM structures have short natural periods, and their damage correlates well to intensity measures such as PGA and SA(0.3s). Accordingly, the modeling of a shallower Wasatch fault under SLC results in higher estimated PGA and SA(0.3) and an increased risk of damage or collapse in SLC. Meanwhile, tall buildings correlate more to long-period intensity measures. A similar exercise can be performed to estimate the potential impacts to these structures through scenario analyses. For example, increased long-period intensities can be expected of subduction motions (Chase et al., 2021), posing higher risks of damage and collapse to these buildings in a city like Anchorage. Notably, as shown in Figure 9, an M_W 8.5 interface earthquake occurring 46 km from downtown Anchorage contributes 23% to the 2475-year hazard. This indicates that large subduction earthquakes can be a relatively high hazard for tall buildings in Anchorage.

Finally, earthquake scenarios such as these can be used for several different applications ranging from emergency response planning to proactive mitigation strategies. Each of these scenarios can provide insight into potential impacts from plausible shaking events in a given built environment. Cities, states, and the federal government can use this information to gain insight into how to properly prepare for when a similar event occurs. For example, estimates of building collapses, debris, injuries, and displaced persons can all be critical in a community’s immediate and long-term response to a disaster. In addition, the same information can also be used to help mitigate the magnitude of these negative outcomes. For example, California enacted a soft-story retrofit ordinance (FEMA, 2023) to strengthen vulnerable buildings across the state in an effort to lessen the potential number of building collapses in an earthquake. Information such as the HayWired scenario effort (Hudnut et al., 2018) helped inform decision-makers on the potential dangers and how to plan for and reduce the impacts. This work can be continued and expanded in future cycles of NSHMs. For example, sensitivity analyses can reveal which scenario inputs are more influential than others (e.g. Price et al., 2010; Remo and Pinter, 2012).

As a part of the release of the 2023 NSHM, earthquake scenarios were developed for Hawaii, SLC, Alaska, and Virginia to help communicate the underlying seismic hazard. For Hawaii, eight new earthquake scenarios were developed, and the existing USGS legacy catalog of scenarios for Hawaii was updated. In SLC and Alaska, 9 and 10 new scenarios were developed, respectively. Finally, a single scenario was developed for Virginia. Particular focus was given to incorporate the expertise of local industry professionals, academic and federal researchers, and other stakeholders into the scenario development process. This input, along with information gathered from existing scenario efforts, combined with the 2023 NSHM science, was used to develop these new scenarios. In addition to scenarios produced, we used these deterministic exercises to investigate the impact of modeling decisions within the 2023 update and to help improve the NSHM further. The uncertainty in modeling decisions in SLC and Hawaii was explored, and we found that a lower dip angle of the Wasatch fault under SLC and a shallower rupture close to Honolulu were more damaging with the potential for higher losses and fatalities. Of note are significant model-related uncertainties that exist in estimating earthquake impacts and thus, not all of those get propagated fully into final loss estimates. For example, the PAGER fatality models are calibrated from the key historical deadly earthquakes (e.g. 1989 Loma Prieta and 1994 Northridge earthquakes in California, 2001 Nisqually earthquake in Washington) and thus perhaps underestimating the number of likely fatalities in SLC scenarios. Their applicability maybe limited in the context of SLC in Utah where a large portion of vulnerable URM buildings do exist and continue to house a large population in Utah. Similarly, the PAGER loss estimates are for shaking-related damage only, and thus, the potential impacts from tsunami are not explicitly modeled into PAGER methodology.

This work can be continued and expanded in future cycles of NSHMs to help communicate the underlying science and updates to the highly complex probabilistic modeling found in the NSHM. In addition, scenario development techniques can also be improved on. Consequence-based scenarios, which target a particular impact (e.g. US$100 million in economic loss or 100 fatalities), can be used to provide communities with tailored estimates to provide the best possible information. Information such as this, and continued collaboration with local stakeholders, can be used to produce even more effective results to help improve communities vulnerable to earthquake risk.

Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government. The authors thank the contributions from Chris Duross, Mike West, James Pechmann, Allison Shumway, Peter Powers, Morgan Moschetti, Simon Kwong, Eric Thompson, and the scenario working group members. Critical feedback received from Ivan Wong, Doug Bausch, Brian Shiro, and the anonymous reviewers from Earthquake Spectra Editorial Board has helped the article significantly.