Application of Machine Learning in Rainfall Disaggregation and Flood Inundation Mapping: A Case Study
摘要
In recent times, urban floods have become a recurring phenomenon, extensively impacting low-lying areas in densely populated Indian cities. Flash flooding is mainly driven by rapid suburbanization and climatic change. Built-up areas, reduction or blockage of existing drains, and poor management of drainage systems can lead to floods in urban areas. The current work focuses on finding the flood inundated regions in Zone XV of Hyderabad city. Sub-hourly rainfall data is temporal disaggregation at 15-min intervals using the Gaussian process regression (GPR) approach. Intensity duration frequency (IDF) curves were then developed from sub-hourly disaggregated rainfall data and applied in the stormwater management model (SWMM) to estimate the runoff generated in the existing storm drainage system. The runoff generated from each subbasin for the study area in the SWMM model was further integrated into the 2D–HEC RAS model to delineate flood inundated areas of the study area.