Location: Hydrology and Remote Sensing Laboratory
Title: Utility of remotely sensed evapotranspiration products on assessing an improved model structureAuthor
![]() |
LEE, S. - University Of Maryland |
![]() |
QI, J. - University Of Maryland |
![]() |
KIM, H. - University Of Virginia |
![]() |
McCarty, Gregory |
![]() |
Moglen, Glenn |
![]() |
Anderson, Martha |
![]() |
ZHANG, X. - South Dakota State University |
![]() |
DU, L. - US Department Of Agriculture (USDA) |
Submitted to: Sustainability
Publication Type: Peer Reviewed Journal Publication Acceptance Date: 2/18/2021 Publication Date: 2/23/2021 Citation: Lee, S., Qi, J., Kim, H., McCarty, G.W., Moglen, G.E., Anderson, M.C., Zhang, X., Du, L. 2021. Utility of remotely sensed evapotranspiration products on assessing an improved model structure. Sustainability. 13(4):2375 .https://doi.org/10.3390/su13042375. DOI: https://doi.org/10.3390/su13042375 Interpretive Summary: Hydrologic models such as the Soil Water Assessment Tool (SWAT) include many parameters to represent physical and biogeochemical processes with high spatial and temporal heterogeneity. Most of these parameters are not directly measurable which leads to increased uncertainty associated with predictions. Model structure improvements that allow incorporation of new data sources can address this modeling limitation. We investigated the usefulness of remotely sensed evapotranspiration products (RS-ET) to provide improved model predictions by the SWAT model after structural improvements. Our results for watershed-level comparisons showed that daily time series of model outputs for the modified SWAT model were not greatly improved with use of RS-ET. However, an alternative metric indicated that the modified SWAT model using RS-ET better represented low-flow conditions. This study demonstrates the utility of both structural model improvements and using additional datasets in these models to reduce model prediction uncertainties. Technical Abstract: Predictive uncertainty is inevitable when applying a hydrologic model for operational purposes. Model structure improvements are often conducted to address this modeling limitation. However, structure improvements are not well quantified when used only with commonly available observations taken at the watershed outlet (e.g., streamflow). We investigated the usefulness of remotely sensed evapotranspiration products (RS-ET) to represent improved model predictions made by structural improvements. Two versions of the Soil and Water Assessment Tool (SWAT) were calibrated against streamflow and RS-ET. We compared simulations from an original SWAT model with those from a modified SWAT model that integrated a physically-based soil moisture module (referred to as RSWAT). First, we analyzed model outputs from the two SWAT versions to see how well two versions replicated streamflow and watershed-level RS-ET. We further compared simulated ET from SWAT and RSWAT at the subwatershed level. Our results for watershed-level comparisons showed that daily time series of model outputs from SWAT and RSWAT appeared similar, showing only minimal differences in performance metrics (Kling-Gupta Efficiency [KGE]) of streamflow (0.56 and 0.58, respectively) and evapotranspiration (ET) (0.53 and 0.54, respectively); but examination of the flow duration curve (FDC) indicated that RSWAT outperformed SWAT in its representation of low-flow conditions. Subwatershed-level comparisons indicated more clear differences between SWAT and RSWAT. RSWAT provided more accurate ET than SWAT in most subwatersheds with average KGE values of 0.47 and 0.53 for SWAT and RSWAT, respectively. Our study demonstrates that RS-ET is useful to assess the prediction capability of an improved SWAT structure, and improved predictions are more observable at a finer spatial scale. |