Rainfall forecasting for flood prediction in the Nilwala basin

Abstract

Flooding is the major natural disaster in Sri Lanka and reliable forecasts with longer lead time is a way of reducing the damages. In this study a weather model was coupled with a hydrologic model and a hydraulic model for predicting floods in Nilwala river basin in southern Sri Lanka. WRF 3.0 (Weather Research and Forecasting) weather model was configured and used to predict rainfall over the basin 24 h into future. The model was configured by investigating the impacts of its physics options on precipitation forecasting. The impacts of microphysics schemes, cumulus schemes, land surface schemes, long/shortwave schemes and boundary layer schemes on rainfall predictions were investigated. The predictions were compared with observed point rainfall data for three rainfall events to find reasonably good physics combination. It was seen that model physics combination; Ferrier microphysics scheme, Kain-Fritsch cumulus scheme, Rapid Update Curve land surface scheme, Rapid Radiative Transfer Model longwave radiation scheme, Dudhia shortwave scheme and Yonsei boundary layer scheme yields better precipitation predictions over the basin. Output of the weather model was coupled with hydrologic model HEC-HMS 3.3 (Hydrologic Engineering Center-Hydrologic Modeling System) with Clark’s, Snyder’s and SCS transformation methods. In all model runs Green-Ampt loss model was executed with recession base flow method. Before using the model with the WRF output HEC-HMS model was calibrated for historical events and Snyder’s method performed better than other methods in calibration and verification. Snyder’s method produced Nash-Sutcliff efficiencies greater than 70% and 50% in calibration and verification respectively. WRF predicted rainfall for May-2003 was introduced to HEC-HMS and the generated river discharges of sub basin were ingested to the HEC-RAS 4.0 (Hydrologic Engineering Center-River Analysis System) hydraulic model for water profile computations along the Nilwala main river. Output of HEC-RAS was exported to Arc- GIS 9.2 where it was two dimensionally visualized as a flood map. Model was capable of predicting the areas as inundated regions but with underestimation of inundation depth.