Optimization of linear stream temperature model parameters in the soil and water assessment tool for the continental United States
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
2018 Elsevier B.V. The Soil and Water Assessment Tool (SWAT) is a watershed model that has been commonly used to simulate hydrological and biogeochemical processes in natural as well as human-modified landscapes. Currently, the SWAT model simulates stream temperature using a sub-model involving linear regression of water temperature on observed air temperature. The use of constant universal regression coefficients in this sub-model, however, significantly limit the model's ability to estimate temperatures in stream networks throughout the continental United States, especially when water temperatures vary due to a combined effect of climate, topography, land use, and hydrologic processes. We present a methodology to improve SWAT's default stream temperature sub-model by optimizing geographic-specific values of coefficients in the regression model. Specifically, the coefficients were optimized to better represent observed stream temperature data at 408 stations across the continental United States. Different models were also constructed for each of the four seasons at each of the 408 locations. The spatially and seasonally varying optimized model parameters led to consistent model improvements when compared with observed data for all 408 locations and for all seasons. The median seasonal NSE improved from 0.58 to 0.77 in the Fall, 3.40 to 0.30 in the Winter, 0.26 to 0.67 in the Spring, and 1.01 to 0.36 in the Summer. The simplicity of the proposed approach provides SWAT model practitioners an easy and flexible solution to seasonally and geographically adjust the stream temperature sub-model, and to better represent local conditions.
