Parameter Regionalization of Semi-Distributed Hydrologic Model for Peak Flow Predictions in Flood-Prone Monsoon-Climate Catchment in Sri Lanka

Abstract

Listen

Translate article

Most of the flood-prone rivers in developing countries are either partially gauged or completely ungauged and thus accurate runoff prediction in these catchments has become a challenging task. Wide range of parameter transferability techniques have already been assessed using various hydrological model applications to several catchments. However, there exists a knowledge gap urging to explore the transferability of semi-distributed hydrological model parameters within a monsoon-dominated catchment having a gauged upstream and an ungauged downstream to accurately predict peak flows in flood-prone downstream areas. Thus, this study aims to evaluate the spatial transferability of model parameters from upstream to downstream and vice versa, by applying semi-distributed Soil and Water Assessment Tool (SWAT) hydrological model to Nilwala catchment in Sri Lanka (1044 km2) which is frequently susceptible to floods. The SWAT model was calibrated(2006-2010) and validated (2011-2015) with observed stream flow data at Urawa (6.24 N, 80.57 E) and Pitabeddara (6.21 N, 80.48 E) gauging stations in the Nilwala catchment. The calibrated and validated model parameters were then transferred from upstream to downstream (Urawa to Pitabeddara, Transfer scheme 1) and vice versa (Pitabeddara to Urawa, Transfer scheme 2) where both upstream and downstream catchments were having similar physiographic characteristics. As measured using the performance indices, the model performance efficiency was acceptable and the simulated hydrographs well represented the rising and falling limbs of the observed hydrographs for both transfer schemes. Results showed that Transfer scheme 1 performed best with high prediction performance as measured using the efficiency measures: coefficient of determination (R2) of 0.72, Nash–Sutcliffe efficiency (NSE) of 0.63, Kling-Gupta efficiency (KGE) of 0.70 and, root mean standard deviation ratio (RSR) of 0.61 with the best simulation shown in the peak flow regime (< Q10). Results revealed that transfer of the SWAT model parameters from upstream to downstream in flood-prone monsoon region catchments is viable, indicating the possibility of applying calibrated semi-distributed SWAT model to other partially gauged catchments in the monsoon region, particularly to the catchments with ungauged downstream to accurately predict the peak flow (< Q10) in the downstream areas in supporting effective flood management.