Sensitivity to Data Choice for Index‐Based Flood Insurance

Abstract Despite increasing adoption of earth observations data to inform disaster response and recovery, deciding which measurements to use—and how—remains an open question. An increasing number of flood insurance programs have been using observable proxies—or indices—to activate payouts. However, convincing evaluation of important design features, including choice of index data, are lacking. This study investigates five potential flood data sets at national and regional scales in a simulated index‐based insurance program in Bangladesh: gridded precipitation, river‐height and modeled inundation from the national flood agency, and two satellite data sets of surface‐water‐extent (one state‐of‐practice, the other state‐of‐the‐art). We demonstrate that data choice determines the accuracy and timeliness of indexed payouts, as well as the uncertainty associated with their likelihood, which influences program costs. For example, while river‐height and satellite water‐extent indices activated payouts during the two worst floods in the 20‐year study period (2004 and 2007), the precipitation‐index activated for just one of them. Furthermore, our state‐of‐the‐art satellite index activated on average 1 week earlier and with 21% lower uncertainty than the satellite‐index used in practice. We propose that practitioners leverage the divergence‐of‐evidence among multiple data sets to identify regions where there is lower confidence in making accurate and timely payouts, which can help inform additional programming such as back‐up payout mechanisms. Beyond insights for practitioners leveraging insurance to protect Bangladeshi communities threatened by extreme monsoon floods, this work offers techniques to assess the sensitivity of indexed programs to different data and scales in other flood‐prone regions.