Probabilistic Flood Inundation Prediction Using Copula-Corrected Precipitation Forecast

Abstract

Flood forecasting is an effective disaster management and risk reduction tool, especially as climate change and urbanisation increase the frequency and intensity of flood events worldwide. We propose a flood inundation forecasting system incorporating a copula-corrected forecast rainfall mechanism to rectify the spatio-temporal inconsistency between observed (from the India Meteorological Department – IMD) and forecast rainfall (Global Ensemble Forecast System – GEFS) patterns. The Dynamic Budyko hydrological model and a conceptual flood inundation model were coupled successively to this corrective mechanism and executed continuously to map the inundation extent for a 1 in 100-year flood event across Kerala, India. For the peak flood day of 16 August 2018, the forecast inundation captured 61.7% of the observed flooded locations at a one-day lead time, decreasing to 48.3% at a seven-day lead time. We evaluated the conceptual inundation modelling framework across Kerala as a proof-of-concept for emergency flood mitigation applications, with runtimes of 2–3 h/lead day.