Effectiveness of hydrologic models for streamflow prediction in the Nam Song River Basin

Abstract

The Nam Song River is a tributary of the Nam Lik River. Itis one of the most attractive natural tourism places of Vientiane in Lao PDR. The Nam Song River basin has long been affected by natural disasters especially floods. The severity of major floods continues to increase in recent years. This is probably a result of changes in the pattern and amount of rainfall which is highly variable in monsoon regions. While realizing the importance of reliable predictions of streamflow and flood for improved water resources and disaster management in the Nam Song basin, this issue has remained a challenge due to scarce meteorological and hydrological gauges and unequally distributed of the gauges over the basin. To predict streamflow in data scarce basins, a rainfall–runoff modeling technique has been used as an alternative to observed streamflow data. The overall aim of this study is to assess the performance of rainfall-runoff models for streamflow predictions under the data scarcity. Three rainfall-runoff models, i.e. HEC-HMS, IFAS and SWAT, with different complexities were tested. The hydrological data were obtained from four rain gauges and two streamflow gauges. The period of the study was from 1996 to 2013. The training data set was from 1996 to 2004 for parameter calibration and sensitivity analysis. A testing data set was from 2005 to 2013 for validation of the model parameters. The performance of the models was evaluated at two temporal resolutions including daily and monthly scales. The correlation coefficient (r) and Nash-Sutcliffe efficiency (NSE) were used as performance indices. While all the rainfall-runoff models tested in this study performed equally well for predicting daily and monthly streamflow time series, they had different capabilities in prediction high flows that might lead to flooding. These results demonstrated that IFAS outperformed HEC-HMS and SWAT when predicting high flows. Therefore, IFAS was considered to be the most suitable rainfall-runoff model for the case of data scarcity in the Nam Song basin. This study could be improved further by searching for parameter sets that lead to increased accuracy of predicting other components of hydrographs apart from peak flows.