FLOOD FREQUENCY ANALYSIS IN MONSOON-CLIMATE CATCHMENT USING SIMULATED RUNOFF SERIES

In developing countries, many catchments are either partially gauged or completely ungauged, making accurate runoff prediction and flood forecasting a major challenge. The Nilwala catchment, a monsoon-climate catchment in Sri Lanka, is highly flood-prone, yet hydrological monitoring is limited to two upstream runoff gauging stations, leaving 70% of the downstream area, including flood-vulnerable Akuressa city ungauged. Integrated frameworks combining continuous rainfall-runoff  modeling with both Annual Maximum (AM) and Peak Over-Threshold (POT) flood frequency approaches remain rare for ungauged, humid tropical catchments. This research addresses this gap by developing a novel, integrated hydrological modeling and flood frequency analysis framework for the ungauged, monsoon flood-prone downstream reaches of the Nilwala catchment. A SWAT model was setup for upper Nilwala catchment, calibrated (2006-2010) and  validated (2011-2015) at Urawa and Pitabeddara gauges, with performance assessed using  statistical criteria, hydrographs, and flow duration curves. At Pitabeddara, the model  achieved a Nash-Sutcliffe efficiency (NSE) of 0.72 and 0.67, percent bias (PBIAS) of -2.7% and 6.8%, and root mean square error standard deviation ratio (RSR) of 0.53 and 0.58 for calibration and validation, respectively; at Urawa, NSE was 0.65 and 0.51, PBIAS was  1.00% and -4.00%, and RSR was 0.60 and 0.70, indicating good to very good performance  overall. A Random Forest analysis was applied to assess the influence of spatial variability on simulated runoff coefficients, identifying forest cover as the most influential spatial  factor. Calibrated parameters were then regionalized from the gauged Pitabeddara sub catchment to the hydrologically similar, ungauged Akuressa sub-catchment. This transfer preserved strong representation of peak flows (