Towards reanalysis of human-flood interactions: case study in the United States

It is of paramount importance for improved water resource management to understand long-term changes in human-flood interactions. Although spatiotemporally homogenous data of long-term changes in human and flood are necessary to deepen our understanding of complex human-flood interactions, such data are unavailable in regional and global scales. Here, we present the prototype of a “reanalysis” dataset of human-flood interactions as an analogy of atmospheric and oceanic reanalysis datasets in Earth science. Following practices in earth science, we develop a socio-hydrological data assimilation system which integrate observations in both human and hydrology domains into a socio-hydrological flood risk model. Then, we generate long-term socio-hydrological data in counties in the United States. By assimilating levee height and population data into a flood risk model, we successfully constrain the trajectory of human-flood interactions. Unobservable variables, like social collective memory, create persistent uncertainties. The output of the data assimilation system (i.e. reanalysis data) clearly indicates that many communities have been getting vulnerable to flood as the flood protection level rises since the social collective preparedness to flood has been declined. Overall, we propose the data assimilation approach and the concept of reanalysis of human-flood interactions as promising ways to realize effective flood management strategies.