Not all flood maps are created equal
For assessing flood risk in California, not all flood maps are created equal. We compared FEMA's flood maps to independent modeling, including climate-informed flood risk. This comparison illustrates very different pictures of flood risk, depending on the map used and how those maps were created, with implications for how California manages infrastructure and prepares for future flooding.
Since the 1960s, the Federal Emergency Management Agency (FEMA) has been charged with mapping and managing flood risk in the US. FEMA's Special Flood Hazard Area (SFHA) defines the 100-year floodplain (or 1% annual chance of flood) and is the regulatory basis for floodplain management nationwide, including whether or not flood insurance is required for many US homeowners.
FEMA mapping has known limitations (Schubert et al., 2024[1]). Most mapping of U.S. rivers is outdated and incomplete. In California, less than 20% of the state's stream miles have been mapped by the NFIP (Schaefer and Pinter, 2019[2]). Furthermore, FEMA's SFHA does not reflect climate-driven or other "non-stationary" changes.
In contrast, independent models of flooding worldwide have been created and marketed commercially, for example, to insurance companies. FEMA itself uses an ensemble of these models to price National Flood Insurance premiums in the US (Miller et al., 2021[3]). We have partnered with Fathom - one of these independent flood modelers - and used their US Flood Map (2020 conditions) and their 2050 climate projections. Fathom's flood models are climate-aware, including both past shifts ("non-stationarity") and projecting future climate-change impacts on flooding. In contrast, FEMA flood maps assume no change over time in the occurrence or probability of flooding.
Mapping State Property Flood Risks
Our team's ongoing work in assessing flood risk to California state properties presented an opportunity to explore interesting differences between flood models. In this work, we used data from the Department of General Services (DGS) to analyze this flood risk. This inventory includes over 24,000 state structures totaling 250 million square feet, primarily California State University buildings and state prisons. The DGS database excludes state-rented space, infrastructure like roads, and some categories of structures, including many University of California facilities. In addition, we used Microsoft's building footprint data to analyze all California structures' flood risk, regardless of ownership.
We compared flood risk maps from FEMA, Fathom 2020, and Fathom 2050, identifying which properties fall within "100-year" flood-risk areas according to each source. Finally, we used the flood depths from the Fathom models to estimate flood depths for state properties for 100-year floods.
Results of the Comparison
Zooming into local areas highlights surprising differences between where FEMA maps show 100-year flood risk versus where an independent and climate-aware model shows the same probability of inundation. The area of California State University, Long Beach (Figure 1) illustrates these differences. FEMA's 100-year floodplain mostly follows the main channel of the San Gabriel River and nearby drainage canals. In contrast, Fathom's 2020 model shows a much larger area of urban land at risk, including several buildings on the campus. A second example shows Fortuna in Humboldt County, where the FEMA and Fathom 2020 floodplain maps overlap significantly, but Fathom's extends beyond the FEMA floodplain (Figure 2).
A primary reason for the differences in mapped floodplains is that the models were created very differently. FEMA flood maps are constructed from local hydrologic and hydraulic ("H&H") modeling. These are detailed local studies that focus on flow in larger stream and river channels. In contrast, Fathom applies similar methods (i.e., physical equations which translate rainfall to runoff to inundation) but with less-precise national datasets to achieve spatially consistent hazard estimates. The math and the hydrology of both types of models are presumably valid (Bates et al., 2020[4]), but they result in very different maps, as shown in Figure 1.
Using the 24,691 California-owned structures as a more-or-less representative statewide sample, nearly the same numbers of properties are at-risk in FEMA's and Fathom's current maps (Figure 3). Of the ~2,800 state-owned structures mapped in the floodplain in either the Fathom map or in FEMA, only 347 structures were in both sets of maps. The large majority of state-owned structures mapped at risk of flooding were in either the FEMA or Fathom maps, but not both (~1400 each).
FEMA's modeling identified many structures in rural areas along larger streams or rivers. In contrast, the Fathom's maps identified broad areas of shallow inundation away from river channels. This type of inundation is sometimes called "pluvial flooding" and includes inundation due to intense downpours, inadequate drainage, and/or flooding of alluvial fan piedmont areas that are common in southern California. Some of the most damaging floods in the US have resulted from such pluvial flooding, including in Louisiana in 2016 (Pinter et al., 2016[5]), in Houston during Hurricane Harvey, and in the Chicago area multiple times including 2023. FEMA mapping is known to overlook most pluvial flood risk. Historically, at least 40% of FEMA National Flood Insurance Program claims are for properties outside the SFHA[6].
Adding Future Climate-Change Impacts
Another potential explanation for the differences between FEMA flood maps and the Fathom maps is that climate change is explicitly considered in the independent modeling. Both of these models are based on the decades-long record of historical stream flows, in the US mostly measured by the US Geological Survey. FEMA 100-year flood levels are derived from those historical flows, using the assumption that past conditions are representative of present and future flood probabilities. In reality, both climate change and a wide array of human alterations of the landscape and river channels alter these probabilities over time. These "non-stationary" changes impact calculations of both present and future flood risk derived from historical hydrology. Part of the difference between the FEMA and the Fathom 2020 flood maps may be the recognition of non-stationary hydrology in the latter.
In addition, we assessed the impact on flood probabilities of future climate change in California, as estimated in the Fathom 2050 model results. Fathom 2050 uses the Shared Socioeconomic Pathways' model of 2050 projected climate change based on current greenhouse gas emissions (SSP2-4.58). Surprisingly, we found only a small increase in flood risk when comparing Fathom's 2020 and 2050 projections for state-owned properties. While there were some differences - 30 properties dropped out of 2020 100-year inundation areas, and 52 new properties appeared in the 2050 model - the overall cumulative changes were minimal. The main takeaway is that both Fathom models incorporate "climate-aware" and non-stationary hydrology (compared to the FEMA floodplain, which does not), with limited additional climate-driven flood risk projected in the 2050 model.
Comparing State Properties' Flood Risk Against All California Structures
We compared state-owned facilities' flood risk to the risk to all structures (private and public) using Microsoft's satellite-derived building outlines database, which has extensive coverage of California.
Looking at state properties alone, we identified a near equal count of properties at-risk for flooding in either FEMA's or Fathom's mapping. Looking at all structures statewide (Microsoft building outlines), we see a significantly larger percentage of properties in the floodplain in Fathom's modeling compared to FEMA (Figure 4). State-owned properties are disproportionately more likely to be in core urban areas (which are often near rivers), or serve river-focused activities (park facilities, fish hatcheries). This likely explains why FEMA's floodplain captures more state-owned properties than other buildings. Other studies have noted systematic or possibly systemic differences between FEMA flood assessment and independent assessments (Schubert et al., 20241).
Fathom Modeled Flood Depths
In addition to just the extent of flooding, the Fathom models provide depth estimates for 100-year flood events at any location. For California state-owned structures, the Fathom 2020 and Fathom 2050 modeled flood depths (Figure 5) are predominantly shallow, consistent with the pluvial flooding most broadly identified in these models. This is important information for designing mitigation strategies, as shallow flooding can be mitigated by a broader range of approaches, usually at much lower cost, than deep inundation.
In Summary…
A side-by-side comparison of FEMA flood maps for California and independent modeling of the state's 100-year flood extent shows surprising and stark differences. Using a statewide database of state-owned facilities, only about 11% of all structures at-risk for flooding in either FEMA or Fathom's floodplain are in both flood maps. According to our observations, FEMA's local H&H-based mapping seems better at local scales near rivers and larger streams, but Fathom better captures the risk of broad, distributed pluvial inundation and the impacts of non-stationary hydrology such as driven by climate change.
By combining the two types of flood assessments, California receives a better, broader, and more robust assessment of its exposure to present and future flood risk. For planners and flood managers wanting to just comply with flood regulations, use the FEMA maps; but for those wanting to assess and plan for a wider breadth of flood hazards, including the impacts of climate change, combine the two mapping methodologies. FEMA has already begun using independent flood models to price its NFIP insurance policies. Together, FEMA and independent flood maps like Fathom better equip California to plan for risk both today and in a changing climate future.