Abstract
Accurate population data is crucial for assessing exposure in disaster risk assessments. In recent years, there has been a significant increase in the development of spatially gridded population datasets. Despite these datasets often using similar input data to derive population figures, notable differences arise when comparing them with direct ground-level observations. This study evaluates the precision and accuracy of flood exposure assessments using both known and generated gridded population datasets in Sweden. Specifically focusing on WorldPop and GHSPop, we compare these datasets against official national statistics at a 100 m grid cell resolution to assess their reliability in flood exposure analyses. Our objectives include quantifying the reliability of these datasets and examining the impact of data aggregation on estimated flood exposure across different administrative levels. The analysis reveals significant discrepancies in flood exposure estimates, underscoring the challenges associated with relying on generated gridded population data for precise flood risk assessments. Our findings emphasize the importance of careful dataset selection and highlight the potential for overestimation in flood risk analysis. This emphasises the critical need for validations against ground population data to ensure accurate flood risk management strategies.