Coastal population and settlement trends in the U.S. 1990–2020: Where and why data choice matters for exposure assessment in coastal resilience studies

Understanding the population density and underlying demographic trends in low-lying coastal areas is crucial for assessing population exposure to future sea-level rise and supporting climate adaptation efforts. Many global population datasets are used for these purposes, but they rely on coarse input data and assumptions that may limit their applicability in densely settled coastal areas. This study compares widely used global population grids (GPW, GPW-UN, GHS-POP, LandScan, WorldPop) with detailed U.S. Census block data from 1990 to 2020 to assess how consistently they estimate population exposure in the Low Elevation Coastal Zone (LECZ, 0–10 m elevation). Using spatial refinement methods based on land-use data (NLCD) and the Historical Settlement Data Compilation for the U.S. (HISDAC-US), we demonstrate how national and global datasets differ in their estimates of coastal populations. National-scale estimates for 2020 vary by up to 6.4 million people (19% relative difference), with the unrefined overlay method systematically overestimating compared to spatially refined approaches. In general, population growth in the LECZ consistently outpaced that in non-coastal areas across all datasets and time periods, with LECZ growth rates 2–7 %age points higher than those outside the LECZ from 1990 to 2020, depending on the allocation method. At the county scale, maximum discrepancies exceed 57,000 people, with some counties showing differences of several hundred thousand people across datasets. The results highlight how data choice can affect exposure estimates at the local scale, and underscore the importance of high-resolution data for effective planning in coastal regions as well as the usefulness of comparative, transparent analysis across different data sources to inform the research and practitioner community about the appropriate use of global population gridded datasets in areas where no local, detailed data is available.