What a Baseline Error in Coastal Risk Assessment Means for Every Decision Built on Top of It
Every flood map starts with a zero. A line that represents where the water sits, not during a surge or a storm but on an ordinary day, at rest. The insurance premium, the mortgage requirement, the elevation certificate, the adaptation budget, the evacuation trigger: every one of those numbers is measured from that point. If the zero is wrong, every number stacked on top of it inherits the error.
Ben Strauss has spent years building coastal risk tools that planners, journalists, and governments use to understand flood exposure. As CEO of Climate Central, he runs one of the few organizations whose 2019 study correctly measured where the ocean surface actually meets the land, using tide-gauge data rather than gravitational models. When a team of Dutch researchers published a finding in Nature this March showing that more than 99% of coastal hazard assessments got the baseline wrong, Strauss was not among the surprised. He described the problem in terms any homeowner could follow: to understand how much higher a piece of land is than the water, you need to know both the land elevation and the water elevation. What the vast majority of studies did was assume that zero in the elevation dataset was the level of the water. It wasn't.
The researchers, Katharina Seeger and Philip Minderhoud of Wageningen University in the Netherlands, analyzed 385 peer-reviewed studies on coastal hazard impacts published between 2009 and 2025. Ninety percent relied solely on mathematical models of the Earth's gravitational field, called geoid models, to estimate where the ocean surface sits relative to the land. Another 9% exhibited issues with improperly aligning land elevation and sea level measurements. Less than 1% got both right.
The ocean does not conform to the gravitational models. Winds, currents, tides, water temperature, salinity push the real sea surface away from the idealized surface. The two most commonly used global geoid models, EGM96 and EGM2008, underestimate actual coastal sea level by a global average of 0.24 to 0.27 meters. The standard deviations are large (0.52 to 0.76 meters), reflecting enormous geographic variability. In some regions, the gap reaches several meters.
Future projections of sea-level rise remain unchanged. The floor those projections rise from was lower than anyone realized. Jonathan Bamber, director of the Bristol Glaciology Centre, called the result "genuinely surprising." He offered a comparison that deserves a moment of quiet:
"The average baseline discrepancy exceeds the total amount of global sea level rise measured since the beginning of the twentieth century."
A century of observed rise, the entire quantity that adaptation strategies and infrastructure investments are designed to address, was already hiding inside a modeling assumption about where the water starts. Every assessment built on the wrong zero was undersized before it began.
Where the Error Is Smallest and Where It Is Largest
Geoid models perform best where gravitational data is densest: Northern Europe, Western Europe, the Eastern United States. These are also the places where most coastal researchers are based and most studies are published. Seeger noted that the "methodological blind spot" persisted partly for this reason. The models looked accurate from where the modelers were standing.
The largest errors are in Southeast Asia and the Pacific, where measured sea levels exceed geoid estimates by roughly 0.9 to 1.1 meters. Latin America, East Africa, the Caribbean, the Indo-Pacific. The Mekong Delta. The Ganges-Brahmaputra. Large, low-lying deltas with dense populations, limited measurement infrastructure, and the fewest resources to respond.
On the shoreline of Vanuatu, 17-year-old climate activist Vepaiamele Trief has watched the ocean take ground her grandmother walked. Beaches eroded, coastal trees uprooted, homes pushed to within three feet of high tide. On her grandmother's island of Ambae, the coastal road from the airport to the village has been rerouted inland. Graves have been submerged. The global assessments that were supposed to quantify the threat to places like hers, and to size the adaptation funding available to address it, underestimated the baseline by nearly a meter.
Seeger and Minderhoud recalculated flood exposure using corrected baselines and four global elevation datasets. Under a hypothetical one-meter rise in relative sea level, 37% more land area and up to 132 million more people would fall below sea level than previous assessments estimated. That range reflects different model and elevation dataset combinations, not a traditional confidence interval. It does not account for levees, pumps, or future population shifts. It measures how much the starting point was off.
I want to be direct about what this means for a US-based readership. The Eastern United States faces the smallest version of this baseline error. If you own a home on the Jersey Shore or in coastal Florida, the discrepancy between the geoid model and measured sea level is relatively modest compared to the global average. The finding still matters to you. But the sharpest consequences land in places most American readers will never visit, in communities whose adaptation funding depends on the global-scale studies that got it wrong. Letting the geography of the readership stand in for the geography of consequence would be its own kind of measurement error.
Dr. Natasha Barlow, Technical Director of Coastal Resilience at Haskoning, identified this structural gap directly: many densely populated delta and estuary regions in the Global South face significant risk from sea-level rise while lacking the high-quality elevation data and resources required to respond effectively to increasing flood hazards.
The Distance Between a Dataset and a Decision
Seeger and Minderhoud released corrected coastal elevation datasets alongside their paper. Openly accessible, ready for use. The science is, in principle, correctable. Minderhoud said the work could "save many researchers complicated calculations" and help make future analyses more realistic.
A corrected dataset now sits on a server. The flood maps in county planning offices remain unchanged. The distance between those two things is measured in years.
In the United States, FEMA's flood maps determine mortgage lending requirements and insurance obligations for millions of properties. Under Risk Rating 2.0, fully implemented as of April 2023, elevation is a core input to insurance pricing. FEMA works with federal, state, tribal, and local partners to update maps, but funding constraints mean only a limited number of communities can be studied or restudied each year. A Physical Map Revision can take up to 18 months to complete. Many existing maps have not been updated in decades.
Running parallel to this, the National Geodetic Survey is replacing the US vertical datum itself. The current system, NAVD 88, is biased by about half a meter and tilted by roughly a meter coast to coast relative to modern global geoid models. The replacement, NAPGD2022, has been in development for years. The Federal Geodetic Control Subcommittee may vote on it in mid-2026. If approved, federal civilian agencies have five years to comply under the Geospatial Data Act of 2018. FEMA flood maps, currently referenced to NAVD 88, would need to be re-referenced as communities restudy and update them under the new datum.
In one worked example from NGS, the difference between NAPGD2022 and NAVD 88 orthometric height was -0.26 meters. Control points, project baselines, and flood elevations that professionals have relied on for years will change on paper, even though nothing moved on the ground.
A corrected global dataset is available now. A new US vertical datum awaits a mid-2026 vote, with a five-year federal adoption window. FEMA map revisions proceed community by community, funding-constrained, 18 months each. From corrected science to a revised flood map affecting an insurance premium, the practical timeline is years to decades.
"Simply put, if sea level in reality is higher for your particular island or coastal city than was previously assumed, the impacts from sea-level rise will happen sooner than projected," Minderhoud said. He also acknowledged: "It is still unclear how much the underestimations of coastal impacts in scientific studies have found their way into policy and its implementation."
What does a person do when they know the map is wrong but the map is still the official map? The flood zone printed on the FEMA panel is what the mortgage lender sees. It determines whether insurance is required, what the premium costs, what the property appraises for. A family in a flood-adjacent neighborhood weighing whether to renovate or relocate is working from the same map that was there last year and will likely be there next year. The correction is technically available. Institutionally, it is years away. The map is the map until it changes, and it changes slowly.
What the Dispute Clarifies
In late April 2026, a group of 17 prominent coastal scientists, led by Gonéri Le Cozannet of the French geological survey, posted a preprint critique arguing that Seeger and Minderhoud overstate the problem. (The critique has not been peer-reviewed, unlike the Nature paper it challenges.) Their core claim: at national and local scales, mean and extreme sea levels are routinely tied to vertical datums, including in assessments of flood hazards and adaptation plans. The geoid-based error matters most for global-scale studies, not local planning. Rutgers sea level expert Robert Kopp echoed this: "Most local planners know their coastal issues and plan accordingly."
Seeger and Minderhoud have not published a formal response. But the contours of the disagreement are worth tracing. Both sides agree the correction is needed. They disagree about how far the error travels into actual decisions. Le Cozannet's group says local practitioners often already know their water levels. Seeger and Minderhoud counter that the global studies are precisely what inform IPCC reports, international climate finance, and the adaptation priorities of countries that lack local measurement infrastructure.
Both claims are probably true, and which one matters depends on where you are. In places with tide gauges, local expertise, and well-resourced planning offices, the error may never reach the decision that affects a family. In places without that infrastructure, the global-scale study may be the only number anyone has. Those places are also where the error is largest. Whether the baseline mistake travels all the way to a planning decision depends on local measurement capacity, and local measurement capacity tracks with resources. Minderhoud pointed to Vietnam as a place where local practitioners have accurate elevation awareness. The problem is that the global-scale science used to allocate resources did not reflect what those practitioners already knew.
The Direction of Every Error
This finding does not stand alone. In the past two years, AMOC slowdown projections were revised roughly 60% worse than the multimodel mean. The global land carbon sink was found to be 20% smaller than models calculated. Eighteen of 40 major river deltas are sinking faster than sea levels rise. These are independent measurements, and they point the same direction. Measurement gaps are widest where resources are thinnest.
What Remains
Coastal geologist Patrick Barnard of UC Santa Cruz offered the most practical takeaway: the finding "underscores how important it is for planners to avoid using findings from big picture studies in local adaptation plans without additional verification." Sound advice for a planner with resources, technical capacity, and time. Communities navigating systems already built around assumptions that didn't account for their reality may not have any of those things.
The corrected data is public. The science is clear. The institutions that translate science into protection have yet to catch up, and who bears the cost of the gap between a corrected dataset and a revised map is a question with familiar answers.
The zero was wrong. Everything built on top of it still stands, for now, as the official map.
- NFIP reauthorization, September 2026: The National Flood Insurance Program faces its next expiration on September 30, 2026, after 35 short-term reauthorizations since 2017, raising the question of whether corrected baselines will factor into any restructured program.
- Deltas sinking beneath the baseline: A January 2026 Nature paper found that 18 of 40 major river deltas are subsiding faster than sea-level rise, compounding the baseline error in precisely the regions where it is already largest.
- The Le Cozannet critique, unresolved: The preprint challenge from 17 coastal scientists has not yet been peer-reviewed or formally answered by Seeger and Minderhoud, leaving the central disagreement about how far the error propagates into real decisions still open.
- US vertical datum vote pending: The Federal Geodetic Control Subcommittee is expected to vote on NAPGD2022 adoption in mid-2026, a decision that will reset the reference point for every flood elevation certificate in the country.