Spatial calibration and uncertainty reduction of the SWAT model using multiple remotely sensed data

Authors: LEE, S. - University Of Seoul; KIM, D. - University Of Seoul; McCarty, Gregory; Anderson, Martha; Gao, Feng; LEI, F. - Non ARS Employee; Moglen, Glenn; ZHANG, X. - South Dakota State University; YEN, H. - Texas A&M University; QI, Y. - University Of Maryland; Crow, Wade; YEO, I. - University Of Newcastle; SUN, L. - University Of Maryland

Submitted to: Heliyon
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/8/2024
Publication Date: 5/9/2024
Citation: Lee, S., Kim, D., McCarty, G.W., Anderson, M.C., Gao, F.N., Lei, F., Moglen, G.E., Zhang, X., Yen, H., Qi, Y., Crow, W.T., Yeo, I.Y., Sun, L. 2024. Spatial calibration and uncertainty reduction of the SWAT model using multiple remotely sensed data. Heliyon. 10(10): Article e30923. https://doi.org/10.1016/j.heliyon.2024.e30923.
DOI: https://doi.org/10.1016/j.heliyon.2024.e30923

Interpretive Summary: Watershed modeling is an important tool for assessing impacts of agricultural practices on soil and water quality, but the lack of observational data can lead to high uncertainty in model predictions. To improve availability of data for model calibration and validation, we tested the use of remotely sensed evapotranspiration products and vegetation parameters to reduce prediction uncertainty. Our findings indicated that the use of multiple remotely sensed products as model constraints enabled model evaluations at finer scales, thereby increasing accuracy of calibrated models and improved ability to represent the spatial characteristics of hydrological variables. This study highlighted the utility of remotely sensed data to increase accuracy of hydrological models.

Technical Abstract: To improve the capacity of watershed modeling, remotely sensed products are frequently used to reduce the uncertainty resulting from data limitations. Although remotely sensed evapotranspiration (RS-ET) products are widely used, vegetation parameters that are responsible for key time/space variations in evapotranspiration (ET) are often calibrated without the use of suitable constraints. Recently, remotely sensed leaf area index (RS-LAI) products are becoming increasingly available. They provide an opportunity to assess vegetation dynamics and improve the calibration of associated parameters. This study aimed to assess the role of the two remotely sensed products (i.e., RS-ET and RS-LAI) in improving the accuracy of watershed model predictions. Specifically, it explored the contribution of RS-ET and RS-LAI products in 1) reducing parameter uncertainty and 2) improving the model capacity to predict the spatial distribution of ET and leaf area index (LAI) at the sub-watershed level. The degree of equifinality, defined as the tendency for different parameter sets to produce equally acceptable model outputs, was assessed for the parameter uncertainty. The results suggested that less than half of the parameter sets with acceptable performances for two constraints (streamflow and RS-ET; 14 parameter sets) were acceptable only for three constraints (streamflow, RSET, and RS-LAI; six parameter sets) at the watershed level. Relative to the watershed-level assessment, the number of parameter sets that satisfactorily characterized spatial patterns of ET and LAI at the sub-watershed level was reduced from six to three. This suggested that the use of multiple remotely sensed products as model constraints enabled model evaluations at finer scales, thereby constraining acceptable parameter sets and accurately representing the spatial characteristics of hydrological variables. Furthermore, this study highlighted the potential of remotely sensed data in increasing the predictability and utility of hydrological models.