The Effect of Incorporation of Embankment Information for Flood Simulation of the Gin River, Sri Lanka

抄録

As flooding is inevitable and becoming increasingly frequent, efficient flood management strategies should be developed to manage floods, especially in developing countries. Rainfall-Runoff-Inundation (RRI) model, which is based on a diffusive wave model, was applied to Gin River Basin, Sri Lanka using daily rainfall data. The RRI model was calibrated and validated for three past flood events (2003, 2016, and 2017) based on observed discharge data and inundation maps developed from ground survey data and satellite images. The Nash–Sutcliffe efficiency (NSE) values for river discharge obtained at the downstream gauging station were greater than 0.7 during both the calibration and validation experiments. Simulated inundation data showed good agreement with the limited observational records. The Critical Success Index (CSI) value for inundated extent in large flood event (May 2017) within downstream was greater than 0.3. Incorporation of embankment information significantly improved the accuracy of the simulation of inundation extent during large flood events (May 2017). The CSI value without embankment information for large flood event (May 2017) within downstream decreased to around 0.1. On the other hand, the embankment information was less useful for smaller flood events caused by less extreme rainfall. Inclusion of embankment information for large flood events enhanced the model performance, thus ensuring the availability of accurate inundation information for efficient flood risk planning and management in the basin.