FLOOD PROJECTION AND ANALYSIS THROUGH STOCHASTIC DOWNSCALING

Abstract

Due to limitations in the General Circulation Models and Regional Climate Models, high resolution rainfall projections are not available. To this end, this study applies a novel stochastic downscaling method that was originally developed for rainfall nowcasting using radar data. In this study, the stochastic model was first modified to adapt the tropical weather data. Singapore’s radar data were utilized to form the basis of the rainfall library of the model. The projected rainfall at high spatial and temporal resolution (1 km and 5 minutes respectively), derived through the stochastic downscaling (using dynamically downscaled outputs), were then used to derive future Intensity-Duration-Frequency (IDF) curves. The IDF curves were then used in flood simulations. Flood maps were generated and the vulnerability of some areas (flood extent and flood depths) were identified. The results are useful for policy makers to propose flood mitigation measures in context of the anticipated increasing rainfall extremes.