Flood modeling prior to the instrumental era reveals limited magnitude of 1931 Yangtze flood

Abstract The global flood risk urges an improved understanding of flood magnitude and its mechanism, which needs insights from pre-instrumental flood investigations. Due to data scarcity, reconstructing pre-instrumental flood magnitudes relies on statistical downscaling, failing to capture nonlinear and dynamic characteristics. We developed a dynamical approach, NorESM-WRF-SWAT, integrating a global climate, a regional, and a hydrologic model to investigate the 1931 Yangtze River flood (the deadliest in the world) and compared it with the 1998’s. Through validation, our method outperforms the statistical method in simulating precipitations and river discharges. For the first time, we presented detailed insights into the intensity and duration of the 1931 flood, revealing a smaller magnitude but associated with an amplified loss, likely due to social vulnerability and reduced societal resilience compared to the 1998’s. While successful simulation can be interfered with by model variability, our dynamical method shows promise for simulating pre-instrumental flood and building a long-term pre-instrumental-hydrology database.