Continuous hydrological modelling using soil moisture accounting for water resources assessment in Kelani river basin, Sri Lanka

Abstract

The assessment of water resources in a river basin for fulfilling various needs in the present and future requires a proper estimation of water availability. This is possible through hydrological modelling. The Kelani river basin in Sri Lanka experiences water stress under the current water uses, development, and urbanization effects. It requires a continuous hydrological model for the assessment of its water resources, focusing on impending climate change impacts. Continuous hydrological models, unlike event-based models, simulate longer periods that include both dry and wet conditions. Soil moisture accounting (SMA) model in the Hydrologic Engineering Centre-Hydrologic Modelling System (HEC-HMS) is chosen to simulate the streamflow. However, the SMA loss model requires precise and updated soil and land use data for parameter estimation, which is not available for the study area. In addition, the lumped nature of the model comparing to distributed models is also in question. This research discusses the development, parametrization and calibration methodologies for the 14 parameters of the HEC-HMS model with the SMA algorithm by considering a catchment divided into several sub-catchments. This division is based on the maximum drainage area method to improve the model accuracy in a scarce soil data situation. The SMA loss model requires 14 parameters to be set. Among these, the impervious percentage is calculated from a land use map; the groundwater 1 and 2 storage as well as the groundwater 1 and 2 coefficients are calculated through the streamflow recession analysis. The maximum infiltration, soil storage, tension storage, and soil percolation rate are calculated from the similar studies; and the groundwater 1 and 2 percolation with four initial parameters are calculated only through a calibration procedure. The model is calibrated using daily data from 2007 to 2012 and validated from 2012 to 2017. The mean ratio of absolute error (MRAE) is used as a primary objective function. The coefficient of determination (R2), percent volume error (PVE), and Nash-Sutcliffe efficiency (NSE) are also used to compare and evaluate the model performance. The results indicate that the performance of the rainfall-runoff model significantly improves when the basin is subdivided into three to eight sub-catchments and the optimum result is found with the five sub-catchments. For the calibration period, the performance of the model is adequate with a R2 of 0.83, a NSE of 0.82, a PVE of 5.3%, and a MRAE = of 0.38. Similarly, adequate results are also retrieved for the validation period, with a R2 of 0.81, a NSE of 0.80, a PVE of 13.1%, and a MRAE of 0.36. The results of the statistical analysis indicate that the simulated and observed flows are reasonably well correlated. The parameter analysis shows that the soil percolation and tension zone storage rates are the most sensitive and second storage of ground water (GW2) is the least sensitive parameters. Furthermore, for the Kelani river basin up to the Hanwella catchment, the simple surface, simple canopy, ModClark, recession and Muskingum methods are found to be the most suitable methods alongside the SMA model. The model performance can potentially be improved through further calibration using hourly climatic input data instead of daily data and with using multiple gauging stations instead of single gauge station. In the future, the validated HEC-HMS model can be employed with seasonal climate forecasts under long-range land use and climate projections. Besides, radarbased precipitation data can be used to represent the climatic variability on a grid-based scale.