Modelling streamflow variability in dry and wet zone river basins in Sri Lanka using satellite soil moisture data

Abstract

Streamflow variability is important in basin water resources management to analyze and plan for the present and future hazards and vulnerabilities affecting effective water management. The unique feature of soil to hold the moisture regulates the precipitation falling on its surface generating the variability in streamflow. The lack of extensive data for distributed hydrological models restrains modelers to accurately simulate temporal and spatial variability of streamflow associated with the soil moisture (SM) in basin-scale. The present study is focused on the use of a simple hydrologic model to assess the impact of SM on the generation of streamflow variability in selected dry and wet zone river basins in Sri Lanka and enhance the model accuracy through the use of satellite soil moisture (SSM) data. The wet and dry zone river basins, Kalu Ganga and Kirindi Oya basins, respectively with a diverse streamflow variability were selected for this study. A semi-distributed hydrologic model was developed to model various events using Hydrologic Engineering Centre’s Hydrologic Modelling System (HECHMS) with soil moisture accounting (SMA) as the loss method. The results obtained from the

model are compared with model results forced with soil moisture active passive (SMAP) SM

data to assess the impact of antecedent moisture on watershed hydrology.

Events of varying magnitude in terms of discharge and precipitation from both Maha and Yala seasons were selected considering different return period discharge to calibrate and validate the model performance. Both models developed for Kirindi Oya and Kalu Ganga performed well with an average Nash-Sutcliffe Efficiency (NSE) of above 0.73 for calibration and above 0.75 for validation along with average root mean square error (RMSE) and observed standard deviation ratio (RMSE std dev) below 0.55 for calibration and below 0.48 for validation. The average coefficient of determination (R 2 ) was obtained above 0.80 indicating a strong correlation. Initial use of SSM improved the model performance of the Kalu Ganga basin whereas deteriorated the performance of the Kirindi Oya basin. The performance was further enhanced by optimizing the soil storage and groundwater parameters yielding an average NSE higher than 0.80, an average R 2 of above 0.90 along with an average RMSE std dev below 0.35 in both basins. Further, the average variation in peak discharge and runoff volume was reduced to 6 % and 2 %, respectively for Kirindi Oya and 15 % and 10 %, respectively for Kalu Ganga basins. The overestimated peak discharge and runoff volume were reduced by 28 % and 18%, respectively upon increasing the soil storage parameters whereas the underestimated peak discharge and runoff volume were increased by 37 % and 43%, respectively by decreasing the soil storage parameters. A minor adjustment in soil storage allowed to manipulate and fine-tune the peak discharge and runoff volume in the basin which substantiates that the runoff is directly associated with the basin SM. The findings of this study can be useful in basins with similar hydrological characteristics to understand the role of SM in runoff generation and for sustainable water management in the basin.