Climate-driven 21st-century sea-level rise is exposing coastal populations, infrastructure, and ecosystems around the world to more frequent marine inundation ( 1 -4 ). At many coastal locations, downward vertical land motion (i.e., land subsidence) sometimes exceeding 5 mm/y is amplifying local relative sea-level rise (RSLR), defined as the change in sea level at a specific location relative to the land, and increasing flooding frequency ( 5 -10 ). However, along much of the coast of Washington, Oregon, and northern California, gradual coastal uplift caused by crustal deformation during the interseismic phase of the current Cascadia subduction zone (CSZ) earthquake cycle locally mitigates the effects of climate-driven sea-level rise ( 11 -14 ). Coastal uplift rates of 1 to 3 mm/y exceed the current rate of climate-driven sea-level rise at locations such as Astoria, OR, Port Orford, OR, and Crescent City, CA, with tide gauges recording RSL fall. At other locations, such as Yaquina Bay, OR, and Coos Bay, OR, where uplift rates are lower, tide gauges show 0.3 to 1.2 mm/y of RSLR, well below the global sea-level rise rate of 4.5 ± 1 mm/y ( 14 -16 ). An exception is Humboldt Bay in Northern California, where complex regional tectonics are causing gradual subsidence, resulting in the highest recorded Pacific-coast RSLR rate of 4.7 mm/y ( 17 ).

The tectonic tempering of climate-driven sea-level rise along the Washington, Oregon, and northern California coasts is projected to be short-lived; by ~2030, rates of climate-driven sea-level rise are expected to outpace gradual uplift. By 2050, central (50th percentile) sea-level projections for a high emissions scenario [SSP3-7.0 ( 18 ) show 0.1 to 0.3 m of RSLR. By 2100, sea levels are projected to rise 0.4 to 0.9 m. The acceleration of RSLR will require Washington, Oregon, and northern California residents and planners to contend with compromised roadways and bridges, more frequently and/or permanently inundated lifelines and critical infrastructure, increased high-tide flooding and vulnerability to storm-surges, increased coastal erosion and barrier dune breaching, and eroding or inland-migrating coastal marshes ( 12 , 19 -21 ).

Yet, gradual climate-driven sea-level rise is not the only inundation threat facing CSZ coastlines. Coastal subsidence from the next great (>M8) CSZ earthquake may produce >1 m of sudden RSLR much sooner than 2100 as evidenced in Cascadia's intertidal wetland stratigraphy ( 22 , 23 ). Stratigraphic evidence of earthquake-driven subsidence from the most recent great earthquake along the CSZ, which occurred on 26 January 1700 CE, indicates sudden (over minutes) 0.5 to 2 m RSLR, resulting in submergence of low-lying intertidal wetlands and floodplains that may persist for decades to centuries after an earthquake ( 23 -29 ). Radiocarbon dating of plant fragments preserved within pre-earthquake peat or overlying mud suggests >11 great earthquakes along Cascadia's coasts in the last 6 to 7 ka, recurring every ~200 to 800 y ( 30 ).

Earthquake-driven coastal subsidence following recent historical earthquakes has had severe consequences for communities, leading to permanent land loss, infrastructure damage, and forced relocation ( 31 , 32 ). The 1960 Chile earthquake caused up to 2.5 m of coastal subsidence, permanently submerging coastal pine forests and farms and converting them to intertidal marshes ( 33 ), and flooding coastal towns and forcing residents to abandon homes and rebuild inland ( 34 ). In 1964, the Alaska earthquake lowered coastal areas by over 2 m, rendering roads, docks, and waterfront areas uninhabitable, in some cases necessitating relocation of communities to higher ground or raising waterfront facilities and airstrips above high tide ( 35 ). The 2004 Sumatra-Andaman earthquake caused land subsidence of up to a meter that led to chronic tidal flooding in waterfront areas used for aquaculture, resulting in oversalinization ( 36 ), and causing coastal erosion and land loss ( 36 ). Similarly, the 2011 Tōhoku earthquake in Japan caused up to 1 m of subsidence, disrupting ports ( 37 ), causing shoreline erosion ( 38 ), and permanently altering the morphology of river mouths ( 39 ).

At the CSZ, the National Seismic Hazard Model calculates a time-independent 15% probability of a M ≥ 8 rupture sometime in the next 50 y ( 40 ). Such an earthquake could suddenly lower coastal areas by 0.5 to 2 m, drastically altering shorelines and causing profound, lasting impacts to coastal populations, infrastructure, and ecosystems. Unlike gradual climate-driven RSLR, this earthquake-driven RSLR will happen within minutes, leaving no time for adaptation or mitigation. Moreover, climate-driven sea-level rise will make coastal areas even more vulnerable to the effects of future earthquake-driven subsidence as it progresses paired with the increased probability (29%) of a M ≥ 8 earthquake occurring by 2100 ( 40 ).

Here, we use earthquake rupture and deformation modeling in combination with geospatial analysis to quantify the projected expansion of coastal floodplains at 24 CSZ estuaries and surrounding communities if earthquake-driven subsidence occurs today (2023), or in 2100, when projected climate-driven RSLR will amplify flooding. We assess the impacts of expanded floodplains on land use, residents, structures, and roads, illustrating the importance of considering the compound hazards of earthquake-and climate-driven RSLR in coastal planning on the Pacific coast of the United States and other tectonically active coastlines.

Effects of Earthquake-Driven Subsidence Today. Using 2023 as a baseline, we use geospatial analysis to quantify the expansion of the 1% floodplain area following earthquake-driven subsidence and its impact on land use, residents, structures, and roads (41-50) at 24 CSZ estuaries and surrounding communities (Fig. 1A; see the Data Availability section and SI Appendix for geospatial dataset information). The 1% (100-y) floodplain includes land that is covered in water during a flood that has a 1% chance of being equaled or exceeded each year. We define the perimeter of the 1% floodplain as the 1% annual exceedance probability water level as measured at a series of National Oceanic and Atmospheric Administration (NOAA) tide gauges along the Washington, Oregon, and northern California coasts (51) (Materials and Methods). Our 1% floodplain perimeters are broadly aligned with the Federal Emergency Management Administration (FEMA) high-risk flood zones within which residents and businesses are required to have flood insurance (52). To depict the floodplains, we overlayed the local 1% annual exceedance probability waterlevel boundary, which ranges from 1.08 to 1.23 m above mean higher high water (MHHW) (Fig. 1C and Table 1), on 10-m (1/3 arc-second) resolution digital elevation model (DEM) tiles (53) (Materials and Methods). We then adjusted the elevation of the 1% floodplain boundary upward by the modeled low (50th percentile), medium (10th percentile), and high (maximum recorded) earthquake-driven subsidence projections for each estuary defined in the FakeQuake Catalog, a forward modeling tool for earthquake ruptures used to simulate coseismic subsidence along the CSZ (54, 55) (Fig. 1 B and C and Table 1). These ruptures range in magnitude from 7.7 to 9.2 and were chosen due to their ability to match the coastal subsidence records correlated to the 1700 CE earthquake. The catalog includes fault slip heterogeneity and variable rupture areas, including both fullmargin and smaller partial-margin ruptures. At the CSZ estuaries analyzed, the modeled low subsidence ranges from 0.23 to 0.67 m, the medium subsidence ranges from 0.46 to 1.34 m, and the high subsidence ranges from 0.93 to 2.67 m. For each subsidence scenario, we use a constant value of subsidence throughout the estuaries and limit our analysis to ~30 km inland from the coast due to the uncertainty in how coseismic subsidence will decay inland (56). Most sites analyzed lie within 10 km of the coastline, except for those in Washington, which extend out to our 30 km inland analysis limit. We note that we report the change in the floodplain area, rather than the total floodplain area before and after subsidence since some parts of the current 1% floodplain are already covered by water.

Our analysis shows that if a CSZ earthquake occurred today, earthquake-driven subsidence would increase the area of the 1% floodplain at the 24 estuaries by 90 km 2 (low subsidence), 160 km 2 (medium subsidence), or 300 km 2 (high subsidence; Table 2 , Figs. 2 and 3 and SI Appendix, Figs. S1-S53 ). The land-use categories with the largest increase in land area within the 1% floodplain are parks and open space (340 km 2 to 410 km 2 ) and farm use (100 km 2 to 160 km 2 ) under the high subsidence scenario. Other notable impacts to land use under the high subsidence scenario include increased exposure to flooding of residential and rural residential (60 km 2 to 100 km 2 ) and commercial (100 km 2 to 120 km 2 ) land.

Along with impacts to land use, earthquake-driven subsidence will cause significant impacts to coastal residents, structures, and roads ( Table 2 , Fig. 2 and 3 and SI Appendix, Figs. S1-53 ). Within the 2023 1% floodplain at the 24 estuaries, there are 8,120 residents, 13,370 structures, and 700 km of roadway exposed to flooding. Following high-end earthquake-driven subsidence today, an additional 14,350 residents (177% increase), 22,500 structures (168% increase), and 1,250 km of roadway (179% increase) are estimated in the 1% floodplain, more than doubling flood exposure.

Postseismic land-level change occurring in the months to years after the next great CSZ earthquake could either temper or exacerbate coseismic subsidence. Luo et al. (2022) ( 58 ) modeled the coseismic and postseismic deformation of the 1700 CE CSZ earthquake along coast-perpendicular profiles in southern Washington and northern Oregon and found that postseismic deformation from viscoelastic relaxation is negligible after 1 y, but afterslip-the slip that may occur between the Episodic Tremor and Slip zone and the seismogenic zone-could produce decimeters of uplift along the coast, depending on the downdip width of the afterslip. Also at the CSZ, high-resolution dating of post-1700 CE sediments shows the reestablishment of intertidal wetlands following coseismic subsidence takes centuries, suggesting a sustained submergence of the coast ( 29 ).

At other subduction zones, geodetic studies following the 2004 Mw 9.2 Sumatra-Andaman earthquake and the 2011 Mw 9.0 Tōhoku earthquake show that in some locations, coseismic subsidence has been exacerbated by continued postseismic subsidence ( 59 -61 ), while other studies found that coastal locations recovered between 10%-50% of their subsidence through postseismic uplift within years ( 61 , 62 ). At the CSZ, the magnitude and direction of postseismic deformation following a future great earthquake is uncertain. For the purposes of our study, postseismic land-level change may, for example, cause projected subsidence to increase from the "medium" to "high" scenario, or decrease from the medium to "low" scenario, depending on postseismic land-level change.

Climate-Driven Sea-Level Rise. The probability of a CSZ earthquake increases with time, and with time, climate-driven sealevel rise is projected to expand CSZ floodplains, compounding the impacts of earthquake-driven subsidence when it does occur. To explore this amplification effect, we use a central estimate (50th percentile) from the Intergovernmental Panel on Climate Change (IPCC) AR6 SSP3-7.0 localized RSLR projections (63) (18) (Fig. 1D) to depict the climate-driven expansion of the 1% floodplain at the 24 CSZ estuaries for the year 2100 (Figs. 2 and 3). SSP3-7.0 assumes emissions and temperatures rise steadily and CO 2 emissions roughly double from current levels by 2100 (63).

The central estimates of RSLR for 2100 along the Washington, Oregon, and northern California coasts range from 0.4 to 0.9 m ( Table 1 ). Our geospatial analysis shows that by 2100, climatedriven sea-level rise is projected to increase the land area within the 1% floodplain by 100 km 2 . This expansion of the 1% floodplain would produce similar land-use impacts to the low earthquake-driven subsidence scenario described in the previous section ( Table 2 ).

In addition to leaving CSZ shorelines more vulnerable to high-tide flooding and storm impacts ( 20 ), the expansion of the 1% floodplain due to climate-driven RSLR will amplify the effects of earthquake-driven subsidence. If a CSZ earthquake occurs in 2100, compared to the 2023 1% floodplain, combined climate-driven RSLR and earthquake-driven subsidence would increase the land area within the 1% floodplain by 170 km 2 (low subsidence), 240 km 2 (medium subsidence), or 370 km 2 (high subsidence; Table 2 and Figs. 2 and 3 ). The land-use categories with the largest increase in land area within the 1% floodplain continue to be parks and open space (340 km 2 to 430 km 2 ), farm use (100 km 2 to 180 km 2 ), residential and rural residential (60 km 2 to 120 km 2 ), and commercial (100 km 2 to 130 km 2 ) under the combined climate-driven SLR and high-subsidence scenario compared to the 2023 1% floodplain.

The combined effects of climate-driven RSLR and earthquakedriven subsidence amplify the impact to coastal residents, structures,

Necanicum River Nehalem River Tillamook Bay Netarts Bay Sand Lake Nestucca Bay Salmon River Siletz Bay Yaquina Bay Alsea Bay Siuslaw River Umpqua River Coos Bay Coquille River Sixes River Elk River Rogue River Pistol River Chetco River Winchuck River OREGON CALIFORNIA WASHINGTON Salem Eugene Medford Ca sca dia Subduct ion Zone 0 45 90 km N Columbia River Willapa Bay Humboldt Bay Projections for all sites are available in Table 1.

pnas.org and primary roads ( Table 2 and Figs. 2 and 3 ). Compared to the 2023 1% floodplain, high-end earthquake-driven subsidence amplified by climate-driven RSLR in 2100 more than triples flood exposure. This most extreme scenario would expose an additional 17,710 residents (218% increase), 29,060 structures (217% increase), and 1,620 km of roadway (231% increase) to flooding.

The Cascadia Rising Scenario conducted in 2016 and 2022 outlined the potential impacts of shaking, tsunami inundation, landslides, and liquefaction from a ~M9 CSZ earthquake in Oregon and Washington, projecting >30,000 casualties, 2,000 destroyed bridges, >170,000 damaged or destroyed coastal structures, and heavy damage to >75% of coastal roadways, >60% of coastal fire stations, >75% of coastal schools, and >80% of seaports, for a resulting economic impact of >$81 billion ( 64 -66 ). However, the potential effects of earthquake-driven subsidence, which may persist over decades to centuries, and the additional flooding exposure it will cause has not been previously quantified and could substantially increase the timeline to recovery.

Our results demonstrate the significant and lasting effects that sudden earthquake-driven subsidence would have on low-lying coastal communities along the CSZ and, therefore, the need for considering subsidence in future hazard assessments. We also highlight the role that 21st-century climate-driven RLSR will have in amplifying the impacts of a future earthquake. If a great CSZ earthquake occurred today, between 90 km 2 (low subsidence) and 300 km 2 (high subsidence) of low-lying coastal land area would be lowered into the 1% floodplain by earthquake-driven subsidence. The greatest impacts to people and infrastructure (i.e., structures and roads) are in the more densely populated areas of southern Washington, northern Oregon, and northern California. Farmlands developed for cattle grazing and farming through diking and draining in the early 20th century ( 42 , 67 ) are one of the most heavily impacted land-use categories along the CSZ. More frequent marine inundation of farmlands will result in salination of agricultural soils and higher salt levels in groundwater, resulting in substantial economic losses ( 68 , 69 ).

In Oregon, our 2023 high-earthquake-driven subsidence scenario depicts a similar amount of flooding as detailed in the Oregon Sea-Level Rise Inventory for Oregon's estuaries ( 20 ) in 2100, which shows that such an expansion of the 1% floodplain would impact 5 airports, 18 critical facilities (e.g., public schools, hospitals, fire stations, police stations, city halls, etc.), 8 wastewater treatment plants, 1 electric substation, and 57 potential contaminant sources (animal feeding operations, gas stations, solid waste facilities, chemical storage, liquid waste storage). If the next earthquake occurs in 2100 (after climate-driven RSLR has already begun to impact the coast) and RSLR rates exceed postseismic and/or interseismic uplift rates, low-lying areas along the CSZ may never recover. Today, and more so in 2100, the immediate effect of earthquake-driven subsidence will be a delay in response and recovery to the earthquake due to compromised assets; long-term effects could render many coastal communities uninhabitable ( 70). Although we do not quantify damage to seaports, previous reports suggest that earthquake-driven subsidence will also compromise jetties, inlets, and navigation channels, affecting port operations and disaster response ( 65 ). Additionally, liquefaction and lateral spreading could locally amplify subsidence in river valleys, waterfronts, and artificially filled coastal locations where critical assets along the CSZ coastline are often located ( 66 , 71 ).

Beyond the direct impacts on infrastructure, sudden earthquake-driven subsidence can appreciably impact natural systems-particularly coastal estuaries, intertidal wetlands, and protective dunes and beaches. Wetland loss is a concern: Intertidal wetlands typically migrate inland in response to rising sea levels, but this inland movement can be constrained by topography and human development. This is especially true along the Oregon pnas.org coast, where Brophy et al. (2018) ( 21 ) demonstrated that a sea-level rise of ~2.7 m could lower ~50% of existing Oregon intertidal wetlands to mudflat elevations, a result comparable to that in this study's high-subsidence scenarios in 2100. Thorne et al. (2018) ( 72 ), who also considered intertidal wetland accretion rates, found that under ~1.4 m of RSLR, Oregon would lose all of its high and middle intertidal wetland environments. The loss of intertidal wetlands directly impacts ecosystem services such as water filtration, habitat for fisheries and shorebirds, and carbon storage capacity ( 72 ). Intertidal wetlands function as natural carbon sinks, and their erosion or conversion to tidal flats reduces their ability to sequester carbon ( 73 ). The erosion and drowning of coastal wetlands caused by earthquake-driven subsidence will also diminish their role as natural buffers against storm surges. Intertidal wetlands can dissipate wave energy, keeping storm surges from penetrating inland and preventing sediment erosion and property damage ( 74 , 75 ). Earthquake-driven subsidence also puts ocean-exposed sandy coastlines at risk. For example, during the 2015 to 2016 El Niño year, a modest RSLR of 7 to 17 cm along the Pacific coast of the United States resulted in substantial coastal erosion, with shoreline retreat 70% greater than during normal winter conditions ( 76 ). Sudden earthquake-driven subsidence can also increase the tidal range within an estuary, exacerbating issues such as high-tide flooding and the impacts of storm surges coinciding with high tides. A study in the Columbia River estuary showed that projected earthquake-driven subsidence could result in up to a 10% increase in the local tidal range ( 22 ).

Finally, sudden earthquake-driven subsidence and climate-driven sea-level rise also need to be considered in tsunami inundation maps. The current tsunami inundation maps for Washington, Oregon, and California take into account the subsidence that will occur during the next CSZ earthquake and how this will increase tsunami inundation ( 77 -79 ). However, tsunami hazard maps do not consider climate-driven sea-level rise and the amplification effect it will have on future tsunamis. ( 80 ) showed that under future climate-driven sea-level rise scenarios, tsunamis created by more common, smaller magnitude earthquakes can have the same coastal wave heights as rare, great-earthquake generated tsunamis. This lesser-considered effect of climate-driven RLSR, especially combined with earthquake-driven subsidence and tides, may imply increased flooding risk in future hazard assessments ( 80 , 81 ).

Our findings stress the importance of incorporating the effects of earthquake-driven subsidence into future flood hazard assessments at the CSZ, as well as considering how climate-driven RSLR will amplify the impacts of a future earthquake and tsunami. Preparing for these compound hazards can minimize long-term damage, ensure resilient communities, and protect critical coastal ecosystems from permanent degradation. Given the global prevalence of subduction zones, these insights hold relevance beyond the CSZ, informing hazard assessments and mitigation strategies for tectonically active regions worldwide.

Geospatial Analysis. To assess the impacts of potential earthquake-driven and climate-driven sea-level rise, we created a series of "bathtub" style 1% floodplain contour polygons on 10-m (1/3 arc-second) resolution DEM tiles from the U.S. Geological Survey (USGS) National Map 3DEP data collection (53). The contour polygon elevations were determined from combinations of potential earthquakedriven subsidence and sea-level rise values. Site-specific 1% exceedance water level elevations are from the NOAA Tides and Currents database for sites at Astoria, OR, Charleston, OR, and South Beach, OR (51). For each site, we apply the closest 1% exceedance water level value to define the perimeter of the 1% floodplain. The starting elevation of the 1% floodplain at each site is reported in Table 1 relative to MHHW. The 1% floodplain contours presented here closely correspond to the FEMA "still water" elevations (82), and likely represent the lower end of potential impacts, as additional impacts from river flow, snow melt cycles, precipitation, and wave action are not included.

To quantify earthquake-driven subsidence and sea-level rise impacts, contour polygons were intersected with a variety of data including state and countylevel land use zoning, road, structure footprint, and population data (41-50) (Dataset S1). To remove inconsistencies with land use data coding between states, a unified land use code was created for use in this study. Since a wide variety of subcategories existed within certain land zones like commercial and industrial, these subcategories were combined into a single category for the entire study area, eliminating regional coding discrepancies. We note that our starting 1% floodplain areas include open water and estuary land. Because of this, in the main text, we emphasize the change in land area in the 1% floodplain rather than the total area. For our land-use impacts analysis, we removed "shorelands" and "estuary" in order to focus more on on-land impacts. Despite open water sometimes being classified as "parks and open space", we kept it in the dataset because the "parks and open space" category is also often found on land.

Sea-Level Rise Projections. The IPCC AR6 sea-level change projections used in this work are medium-confidence projections for the SSP3-7.0 emissions scenario (83). The medium-confidence projections use methods and assumptions about the individual processes that contribute to sea-level change that are assessed to have medium confidence or stronger by the IPCC and therefore do not include contributions that could lead to more extreme sea-level rise, but which have lower confidence levels (such as Marine Ice Cliff Instability). The sea-level rise projections are provided on both a 1×1 degree grid, and at 1,030 tide gauge locations from around the world. The sea-level rise projections are provided in decadal time steps starting in 2020 and extending to the year 2150; here, we focus on the 50th percentile of projections for the year 2100 (18, 63)-and therefore do not account for the possibility of more extreme, tail-area sea-level rise totals.

For this work, we use the publicly available NASA Sea Level Projection tool (83) to isolate the projected sea-level rise at points that are most relevant for our work. This allows us to select the best sea-level rise value for each location on a case-by-case basis, whether that value comes from the nearest 1×1 degree ocean grid cell in the gridded sea-level rise dataset or a tide gauge location along the Pacific coast. Because sea-level rise values from the gridded dataset will have interpolations that capture vertical land motion to varying degrees of success, this manual approach to selecting sea-level rise projection values allows us to ensure that we are using the best sea-level projection for each site, based on how well vertical land motion is captured within both the gridded data and the tide gauge dataset.

Modeled and Observed Earthquake Subsidence Estimates. Subsidence estimates are calculated at each site based on about 1,600 kinematic, stochastic slip rupture models (54) of varying magnitudes between 7.7 and 9.2. These models come from a larger catalog of 37,500 hypothetical ruptures (55). Each of these ruptures is unique from one another, with rupture area, amount of slip, and location of dominant slip patches varying between ruptures. These ruptures were initially chosen based on their abilities to match the coastal subsidence records correlated to the 1700 CE event (23,57). Subsidence estimates for the 1700 CE event are distributed along the entire length of the CSZ, likely representing a full-margin rupture. To account for the possibility of shorter or segmented rupture scenarios, each kinematic rupture model must reproduce the observed 1700 CE subsidence for sites located within 50 km of the modeled rupture area. This allows for a wider range of subsidence estimates to be modeled.

For each kinematic rupture model, coseismic subsidence is calculated at each site using the analytical solution for angular dislocations for triangular subfaults in an elastic half space (84). Based on 1,600 model results, three coseismic subsidence values are determined for each site: a small, medium, and high value. We base the high subsidence value on the largest subsidence modeled at each location to function as the "worst-case scenario." Stated prior, modeled subsidence values are validated with respect to coastal subsidence estimates previously determined for the 1700 CE event, although not all geologic sites with subsidence estimates are colocated with the 24 sites modeled in this study. As a result, modeled sites closest to the geologic sites with estimated subsidence values more closely resemble the upper bounds of the 1700 CE geologic subsidence estimates. Locations that are farther from sites with observed subsidence estimates are less constrained by the 1700 CE data, and as a result the models produced higher "worst-case scenario" subsidence estimates there (e.g., Sixes River, Elk River, Rogue River, Pistol River, Chetco River, Winchuck River, Oregon). Due to modeled subsidence estimates being unrealistically high at these locations, we used the closest, better constrained subsidence estimate (e.g., Coquille River) for these sites (Table 1). At the 24 sites, mean subsidence values are -0.4 m, -0.9 m, and -1.7 m for the low, medium, and high modeled subsidence values, respectively. The modeled subsidence values follow the low (50th percentile), medium (10th percentile), and high (maximum recorded) earthquake-driven subsidence values of the 1,600 ruptures for each site location.

Data, Materials, and Software Availability. All data integral to the stated conclusions are presented within the results, Materials and Methods, or SI Appendix. All shapefiles generated in this study are available at