Sea-level rise is a key measure of how climate change will shape our future. Coastal communities rely on accurate coastal hazard assessments to manage escalating flood risks, protect infrastructure, and mitigate economic losses—but new findings show that most of the research is inherently flawed.
The study, published Wednesday in Nature, found that a staggering number of sea-level rise impact assessments are based on drastic underestimations of current sea levels. Of the 385 peer-reviewed scientific papers the authors analyzed, more than 90% relied on assumed sea levels based on gravitational models called “geoids” rather than direct sea-level and land-elevation measurements. Consequently, these assessments have underestimated the timing and severity of sea-level rise impacts.
After accounting for this widespread methodological issue, the authors determined that 3.28 feet (1 meter) of global sea level rise could inundate 37% more land area than previously thought, impacting 77 million to 132 million people across the globe.
“What our study uncovers is, in a way, a methodological blind spot that is positioned between—I would say—traditionally disconnected scientific disciplines,” co-author Philip Minderhoud, an associate professor of coastal geoscience at Wageningen University in the Netherlands, said at a Tuesday press briefing. He conducted the study alongside Katharina Seeger, a geographer at the University of Padova in Italy and a guest researcher at Wageningen.
Modeling versus measuring
To find this blind spot, it helps to understand how various methods of sea-level assessment differ from each other.
Geoids are not based on direct, real-time sea-level measurements. These models are mathematical representations of Earth that approximate average sea level based on two factors: gravity and the planet’s rotation. They essentially represent the surface of the ocean in the absence of tides, winds, and currents.
But actual sea levels are constantly influenced by these forces, not to mention other factors such as temperature and salinity. Only direct measurements—primarily via tide gauges and satellite observations—can capture this complexity. While geoids can provide a theoretical baseline, using them as a starting point for sea-level rise projections and coastal hazard assessments inevitably leads to inaccuracies, as this new study shows.
“I thought that was terrible that people have done that and were unaware of what the authors point out,” David Holland, a New York University professor of mathematics and atmosphere-ocean science who was not involved in the study, told Gizmodo. “I think the authors have made an excellent point and a major contribution.”
A different future
Most of the studies that the authors analyzed understated coastal sea levels by 9.4 to 10.6 inches (24 to 27 centimeters), depending on the geoid used. In areas where the models are least accurate—often in parts of the Global South—actual sea levels could be 18 to 24.9 feet (5.5 to 7.6 meters) higher than assumed. The greatest discrepancies were in Southeast Asia and the Pacific region, where many island nations are already experiencing catastrophic sea-level rise.
But Bob Kopp, a Rutgers University climate scientist and professor who was not involved in the study, told Gizmodo in an email that while Seeger and Minderhoud make an important technical point, it’s easy to overstate its broader significance.
For example, the study challenges the Intergovernmental Panel on Climate Change’s 2022 estimate that almost 11% of the global population—about 896 million people—live in low-elevation coastal zones, arguing that the true proportion is between 12.3% and 13.7%, or roughly 970 million to 1.07 billion people.
“While certainly important to a summary global metric such as this, such statements are only important to indicate that coastal risk matters for a lot of people,” Kopp said, adding that uncertainties tied to human behavior—such as climate migration and coastal adaptation measures—are likely to have a greater influence on future coastal risk than technical differences in sea-level baselines.
Still, fixing this methodological problem will improve the accuracy of local coastal hazard assessments and international reports like the IPCC, Holland said. Minderhoud and Seeger said ensuring such reports represent the realities of sea-level rise in the world’s most vulnerable areas is critical, as this spurs global climate action and helps communities garner international support for adaptation efforts and, in the worst cases, migration.
Bridging the gap
The authors therefore call for a paradigm shift in coastal hazard research. They urge researchers to reevaluate their assessments to ensure that they are based on properly integrated sea-level and land-elevation data, and not solely on geoid models.
Proper data integration is key here. While 9% of the 385 studies tried to combine actual sea-level and land elevation data with geoid models, most handled it incorrectly, leading to conversion errors and misaligned measurements. Only one of the studies Seeger and Minderhoud evaluated nailed its analysis, with full data documentation and proper integration.
The authors hope their study helps the scientific community improve its methodology by providing open-source data and ready-to-use tools that facilitate the accurate alignment of land and sea-level measurements. They even converted several state-of-the-art digital elevation models to coastal sea-level height, providing a jumping-off point for researchers to rework past assessments and produce better ones going forward.
As for policymakers, Minderhoud and Seeger encourage them to check the information that underlies their decision-making and ensure that it is based on locally validated data.
By charting a path toward more accurate coastal hazard research, the authors aim to help scientists and policymakers better protect communities from the accelerating impacts of sea-level rise. Their work could reshape planning and adaptation efforts, helping ensure that vulnerable regions receive the resources they need before it’s too late.
