The ocean knows where it is. The map disagrees.
At Galveston's Pier 21, NOAA's datum tables show that mean sea level sits 0.158 meters above the zero point that federal flood maps use as their vertical reference. Mean higher high water, the average of the highest daily tides, sits 0.335 meters above it. More than thirteen inches. On a coast as flat as the upper Texas Gulf, where a foot of elevation can separate dry ground from standing water across several blocks, thirteen inches can move a contour line deep into a neighborhood. And the contour line on a flood map is the line that separates families who must buy flood insurance from those who don't. The line that determines buyout eligibility, planning priorities, bond-rating assumptions, and whether a neighborhood shows up in the system designed to protect it.
A 2026 Nature paper by Seeger and Minderhoud found that correcting baseline errors in coastal impact assessments puts 31–37% more land and 48–68% more people below projected sea level, raising global exposure estimates to between 77 and 132 million people.
A paper published this year in Nature by Katharina Seeger and Philip Minderhoud found that more than 99 percent of coastal impact assessments worldwide use sea-level or elevation baselines that undercount who actually lives below projected flood levels. Correcting this baseline error, under a hypothetical one meter of relative sea-level rise, increases the amount of land identified as below sea level by 31 to 37 percent and the number of people by 48 to 68 percent. The corrected global estimate: between 77 and 132 million people exposed.
This weekend, millions of Americans will drive to the coast, set up chairs near the water, and watch fireworks over the Gulf. They will trust, without thinking about it, that the systems measuring where the water will go are measuring from the right starting point. The Seeger and Minderhoud paper says, with considerable evidence, that nearly all of them are not.
The map's zero
NAVD88, the North American Vertical Datum of 1988, is the official elevation reference for the United States. When a flood map says a base flood elevation is twelve feet, it means twelve feet above NAVD88's zero. When a surveyor certifies an elevation certificate for a mortgage closing, the number references NAVD88. When an engineer designs a seawall, the drawings reference NAVD88.
But NAVD88's zero was not set to match the ocean. It was set by a single tidal benchmark at Rimouski, Quebec, then propagated across the continent through a network of leveling surveys. NOAA's own National Geodetic Survey acknowledges the problem:
"Biased by about one-half meter and tilted about one meter coast to coast" relative to current global geoid models.
The agency notes that the datum relies partly on more than a million physical survey marks in the ground, marks that can be "deteriorated, destroyed, or not maintained," making height data "outdated and inaccurate." NGS identifies sea-level rise, subsidence from oil and gas withdrawal, and tectonic activity as situations where outdated marks hinder mitigation.
None of this is secret. NOAA has been working on a modernized National Spatial Reference System for years. But the flood maps haven't changed. The insurance pricing hasn't changed. The elevation certificates that determine whether a family's mortgage requires flood coverage haven't changed. The gap between what's known and what's operationalized sits there, published on government websites, visible to anyone who looks at two numbers and subtracts. The downstream systems, the ones that actually touch families, don't subtract. They take the map's zero as given.
From zero to someone's kitchen table
FEMA's National Flood Hazard Layer for Galveston County shows base flood elevations referenced to NAVD88, with maps effective as of August 2019. If the map's zero is half a foot lower than actual mean sea level, and more than a foot lower than mean higher high water, the contour line defining the Special Flood Hazard Area is drawn too close to the water. Properties that should appear inside the risk zone don't. On terrain as flat as Galveston's, where the Strand Historic District and residential neighborhoods along the bay side sit only a few feet above any reference point you choose, that shift plays out in concrete terms: the difference between a block that appears on every planning map and emergency checklist and a block that doesn't appear on any of them.
From there the consequences compound. The Congressional Research Service reports that the National Flood Insurance Program held more than 4.7 million policies and $1.3 trillion in coverage as of January 2025. Property owners in mapped Special Flood Hazard Areas with a one-percent annual flood chance must purchase flood insurance as a condition of receiving a federally backed mortgage. Outside the line, most people don't buy. The line is the trigger for mandatory purchase.
Whether that trigger still matters as much as it used to is a genuine point of contention. FEMA's Risk Rating 2.0, introduced in April 2022, bases individual premiums on actual flood risk rather than flood zones alone, incorporating distance to water, flood frequency, and building characteristics. In principle, this should erode the map-line's binary power: a property priced on its actual risk profile shouldn't need a zone boundary to tell it what it faces. Some flood-risk professionals argue that RR 2.0 already mitigates the datum problem by pricing risk continuously rather than categorically. But the mandatory-purchase requirement still follows the map, not the risk rating. A homeowner outside the Special Flood Hazard Area faces no federal requirement to carry flood insurance regardless of what RR 2.0 would charge them. The zone line determines who must buy. The risk rating determines what they pay. These are different mechanisms controlled by different parts of FEMA's infrastructure, and the datum error affects the first one directly. Until the mandatory-purchase trigger decouples from the mapped zone boundary, the line on the map remains the most consequential product of the measurement convention.
In nearby Harris County, the Houston Chronicle reported this year that draft FEMA maps could move nearly 200,000 properties into high-risk zones, with many single-family NFIP policies costing around $1,000 annually and some above $2,000. The mechanism is identical across the region: where the line falls determines who pays and who can sell, which neighborhoods hold property value and which erode.
And the mechanism points in one direction. Families currently outside the mandatory-purchase zone are, by definition, the ones least likely to carry flood insurance voluntarily. When the line eventually corrects to include them, they absorb both the insurance cost and the property-value adjustment simultaneously, with no transition period built into any existing federal program. Research on NFIP premium increases and mortgage performance suggests these costs can push marginal borrowers toward delinquency, though the magnitude depends on local housing markets and household financial reserves. The communities least likely to have a surveyor in the family, least likely to know that NOAA publishes datum relationships online, are the ones most likely to be making decisions with incomplete information about where the water actually sits relative to their floor.
A problem two decades in the making
Seeger and Minderhoud didn't discover the vertical-reference problem. They quantified its global scale in a way that makes it harder to treat as a local anomaly. The concern has been building in the peer-reviewed literature for nearly two decades.
In 2009, Dean Gesch published in the Journal of Coastal Research establishing that elevation-data accuracy is a first-order problem for coastal risk assessment. In 2012, Benjamin Strauss and colleagues published in Environmental Research Letters arguing that coastal exposure estimates should be referenced to tidal water levels rather than a generic elevation zero. In 2019, Scott Kulp and Strauss, researchers at Climate Central, a nonprofit climate science and communications organization, published in Nature Communications that the widely used SRTM elevation dataset had a positive vertical bias comparable to projected sea-level rise this century, and that correcting it tripled many global exposure estimates.
The distinction between these findings matters. Kulp and Strauss asked whether the land surface was being measured correctly. Seeger and Minderhoud ask whether the water surface and the land surface are being referenced to the same vertical system. One is an error in the ruler. The other is an error in what the ruler measures from. Both produce maps that tell people they're safe when they may not be, through different mechanisms. Fixing one doesn't fix the other.
A caveat worth sitting with: the paper's supplementary data shows U.S.-specific mean offsets of 0.242 meters (EGM96) and 0.308 meters (EGM2008) at 90-meter resolution. But these are national averages. The paper notes that global geoids represent coastal sea level "relatively well" in data-rich regions like the eastern United States, meaning offsets there may be smaller than in the Global South. The supplementary statistics do not break out the U.S. Gulf Coast separately. The global and national findings cannot be cleanly applied to a specific parish or county without local tide-gauge and geoid calculations. What can be verified locally is simpler: at Galveston Pier 21, the gap between NAVD88 and actual mean sea level is published in NOAA's own data. The Nature paper confirms a global pattern. The two numbers on the same government website confirm it at Pier 21.
The ground moves too
Along the Gulf Coast, the datum gap isn't static.
In coastal Louisiana, Jankowski, Törnqvist, and Fernandes reported in 2017 present-day relative sea-level rise rates of 12.0 ± 8.3 mm per year, with median shallow-subsidence rates of 6.0 mm per year in Mississippi Delta sites. They made a point that matters enormously for the datum question: tide gauges in coastal Louisiana often don't capture the full relative sea-level rise experienced at the land surface, because gauge benchmarks are typically anchored tens of meters below ground while much subsidence occurs in the uppermost five to ten meters. A NAVD88 benchmark can be internally consistent while still missing community-scale land-surface loss.
In the Galveston area, the Harris-Galveston Subsidence District monitors ground movement from groundwater withdrawal. Current GPS rates near Pier 21 are modest, around 0.2 to 3.9 mm per year depending on the station. But NOAA's sea-level trend data for Galveston Pier 21 shows approximately 6.6 mm per year of relative sea-level rise over 121 years, a rate combining ocean rise and land subsidence into a single upward creep of water against infrastructure.
Every millimeter of subsidence widens the gap between the map's zero and the water's zero. A benchmark set in 1988 that has physically moved downward doesn't announce that its elevation is now wrong. The map doesn't update itself. The family living at what the map says is four feet above base flood elevation doesn't get a letter explaining they might actually be at three.
The cost of a correct map
As of publication, no public-facing guidance from FEMA, NOAA, any state coastal agency, or any bond-rating agency has addressed the Seeger and Minderhoud paper's specific findings, based on checks of the Nature paper and supplement, FEMA NFHL data, NOAA CO-OPS and NGS datum pages, and Harris-Galveston Subsidence District materials. Internal discussions may be underway. But the public-facing infrastructure of coastal risk assessment continues to operate from the same reference points it used before the paper was published.
The measurement convention persists as a systemic methodological default. NOAA knows NAVD88 has problems. FEMA's own data model carries fields for vertical datum identification and conversion factors, showing the agency's mapping infrastructure recognizes datum as a variable. The modernized National Spatial Reference System has been in development. No one has answered what happens in the years between knowing the ruler is off and actually redrawing the lines.
Correcting the baseline redistributes risk. Properties that appeared above the flood line move below it. Mandatory insurance requirements activate. Property values adjust. Municipal bond ratings face new inputs. Buyout programs, already underfunded and slow, confront expanded eligibility. Every institution downstream of the contour line absorbs a new obligation.
The political math follows a pattern visible across disaster recovery and environmental policy: a corrected map is more accurate, and accuracy creates costs, and those costs bypass the institutions that draw the map. They land on the families who discover their home is in a zone they didn't know existed. On the local governments whose fiscal projections assumed a stable tax base. On the school districts and fire departments funded by property taxes that may decline. The measurement convention persists in part because correcting it creates problems that no single institution is structured to solve.
What you can ask
This piece cannot tell you whether your home's elevation certificate is referenced to a datum that matches your local sea level. That determination requires your nearest NOAA tide gauge's published datum relationships, your community's FEMA flood map metadata, and, if you're in a subsidence-prone area, current benchmark-stability data. All of these are publicly available. None of them are easy to find or interpret without training.
But there are questions worth bringing to your next city council meeting or planning commission hearing:
- When were your community's flood maps last updated, and what vertical datum do they reference?
- Does your local tide gauge show a gap between NAVD88 and mean sea level, and how large is it?
- Does your area have measurable subsidence, and have the benchmarks your maps rely on been resurveyed since installation?
- Has your local planning authority evaluated what a datum correction would mean for the boundaries of its flood zones?
These are not hypothetical questions. They have answers, in feet and inches, that determine which side of a line your home sits on.
Thirteen inches, at Galveston's Pier 21. The distance between the map's ocean and the real one. Small enough to step over on a Fourth of July walk along the seawall. Wide enough to leave a neighborhood off every map, every checklist, every system that was supposed to see it.
-
Flood maps meet hazardous sites: A 2025 Nature Communications study identified 5,500 U.S. hazardous facilities at risk of a 1-in-100-year flood by 2100, with renters, households without vehicles, and linguistically isolated communities disproportionately likely to live nearby, raising the question of whether those facility-level flood assessments use the same flawed baselines.
-
Harris County's 200,000-property reclassification: The Houston Chronicle reported that draft FEMA maps could push nearly 200,000 Harris County properties into high-risk flood zones, offering a real-time test of what happens to families, property values, and local tax bases when the line moves at scale.
-
Louisiana's land is sinking faster than gauges show: Jankowski, Törnqvist, and Fernandes found that coastal Louisiana relative sea-level rise rates reach 12.0 mm per year, with shallow subsidence that tide-gauge benchmarks anchored deep underground may not capture, compounding the datum gap in the places least able to absorb it.
-
NOAA's datum replacement is coming: The National Geodetic Survey has been developing a modernized National Spatial Reference System to replace NAVD88, but no timeline has been set for when downstream systems like FEMA flood maps, elevation certificates, and mandatory-purchase triggers will transition to the corrected reference.