Reply to: Global gridded population datasets systematically underrepresent rural population

Láng-Ritter et al. (2025) use population data from 35 countries where rural areas were flooded by dam and reservoir construction. They compare historical population counts prior to flooding with estimates from global gridded population datasets and report that gridded population data underestimate people in these areas by more than 50%. Based on this, they claim that gridded population data systematically undercount rural populations around the world and suggest that rural census data are also flawed. We appreciate the authors’ interest in gridded population data accuracy but believe they misunderstood how these datasets are constructed and try to validate them by their performance in representing historical population changes around very local and rare flooding events, that are not representative for the quality of global, gridded population data measuring contemporary rural population distributions. Global, gridded population data rely on models that redistribute census-based population counts using geospatial data on water bodies and built-up areas, and other factors related to population presence and density. These models assume water bodies to be static over time and do not measure buildings that existed in the past but have since been dismantled, demolished, destroyed, or flooded. They are not designed to capture rare events like villages flooded by newly built water reservoirs. This is a known limitation, not a flaw, that has been clearly described by gridded population data producers. Láng-Ritter et al. (2025) measure effects of these limitations and exaggerate the scale of the problem. The total number of people displaced by dam construction in the analyzed data is small - less than 0.05% of the world’s population - yet they use relative error metrics that grossly inflate their results. They also avoid assessing if gridded population data overestimate post-flooding population in areas where the data on water bodies are out of date, which would invalidate their claim of “systematic underrepresentation”. Moreover, they confuse “missing data” with “misplaced data” - in many cases, the population is still represented in the gridded population data, just in slightly different locations. We estimate that population displacement due to reservoir construction causes such slight misplacements in global gridded population datasets affecting less than 2% of the global rural population, significantly lower than the claim by Láng-Ritter et al. (2025) of over 50% missing rural population globally. In this response, members of the POPGRID data collaborative explain why the conclusions by Láng-Ritter et al. (2025) are incorrect and not supported by the data or methods used. Moreover, we call for awareness of the limitations and fitness for use of global gridded population data.