Improving process-based prediction of stream water temperature in SWAT using semi-Lagrangian formulation Journal of Hydrology

Accurately predicting stream water temperature at large scales requires modeling techniques that account for various physical, chemical, and biological processes that influence water temperature. One common approach is to use process-based hydrodynamic models that simulate water movement throughout a river basin. In this study, we assessed whether enhancing the Soil and Water Assessment Tool (SWAT) by incorporating a semi-Lagrangian formulation that explicitly accounts for various thermal processes improves its ability to predict stream water temperature. A basin-scale evaluation of the modified module, conducted in the Gunpowder Falls River and San Antonio River Basins using data from 2012 to 2022, showed that the modified model consistently demonstrated strong predictive capabilities for daily stream water temperature with R2 and KGE values ranging from 0.71 to 0.91, and 0.77 to 0.92, respectively. It also more accurately captured the monthly and seasonal variations in stream temperature compared to the default formulation, which is simply a function of air temperature. We recommend accounting for thermal processes using a semi-Lagrangian approach wherever these variables are available to improve the SWAT’s ability to predict stream temperature. Grant no. NA19NES4320002