A spatially and temporally disaggregated twenty first century global flood record for flood impact analysis

Abstract

Given that floods are one of the most widespread and costly disasters, there is significant attention on understanding the hydrologic and socioeconomic drivers of these events. However, current research efforts are limited by a lack of detailed, comprehensive, and global-scale data on historical flood events and impacts. In this study, we combine flood damage records with climate reanalysis data and satellite-based flood detection to create a spatially and temporally disaggregated 21st century flood dataset. We start with 4073 inland floods from 2000-2024 contained in the EM-DAT international hazard database. We then disaggregate each event into sub-national administrative levels (e.g., states or provinces) and, for longer duration events, by calendar month to enable analysis at finer spatial and temporal scales. For each event, we derive flooded pixel maps from MODIS satellite imagery. By combining these high-resolution flood maps with gridded population density data, we calculate the number of people exposed to flooded areas in each state or province over time. Lastly, we use climate reanalysis data to extract historical precipitation data at the administrative region and month level in order to characterize the meteorological conditions of each flood. The result is a spatially and temporally consistent dataset of global flood characteristics and impacts to enable more granular impact analysis. In our exploratory analyses, we use the dataset to investigate how historical precipitation contributes to observed flood impacts—both in terms of people affected and economic damages—across different regions and through time. We find a statistically significant relationship between monthly extreme precipitation and flood impacts using a global-scale fixed-effects panel regression model. In future work, this model could serve as the basis for attributing flood impacts to underlying changes in the precipitation distribution.