Improved Understanding of Water Balance in the Malwathu Oya River Basin Using SWAT and Remote Sensing

Sammanfattning: As the need for climatic data is increasing in times of climate change and water scarcity, remote sensing (RS) and hydrological modelling are ways to battle these problems, especially in data scarce areas. The actual evapotranspiration (ETa) is one of the key parameters when assessing the water balance and a good estimate of this parameter is thus of great importance. In this study a hydrologicalmodel was created with the Soil and Water Assessment Tool (SWAT) over the Malwathu Oya river basin, Sri Lanka, and the SWAT ETa estimates were compared to RS derived ETa from FAO’s open access database WaPOR. A sensitivity analysis and a calibration with observed streamflow data of the SWAT model was conducted with the SUFI-2 algorithm in SWAT-CUP. The calibration was satisfactory and showed the following values for the performance parameters: R2 = 0.72, Nash-Sutcliffe Efficiency, NSE = 0.69, and Percent of Bias, PBIAS = -10.4. The most sensitive parameters were CN2 (runoff curve number for moisture condition II), SOL_AWC (soil available water capacity), and ESCO (soil evaporation compensation factor). The water balance partitioning from the calibrated SWAT model showed a ratio of 0.68 between ETa and precipitation as an annual average between 2012–2020. In the comparison between SWAT ETa and WaPOR ETa the SWAT ETa showed a clear underestimation, particularly during the drier Yala growing season (May – August). However, the SWAT land use classes representing the cultivated rice fields agreed well with WaPOR while the forest and range grasses were underpredicted. To increase the performance of SWAT in estimating ETa the following was recommended: improvement of the simulation of the shallow aquifers, more accurate forest parameters, deactivation of the default dormancy period in SWAT, calibration with ETa instead of streamflow, and a higher resolution soil map together with more soil measurements.