Long-term, process-based, continuous simulations for a small, nested rangeland watershed near Tombstone, AZ (USA): Extending model validity to include soil redistribution

Authors: ZHANG, H. - University Of Buffalo; RENSCHLER, C. - University Of Buffalo; Nichols, Mary; Nearing, Mark

Submitted to: Science of the Total Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/8/2021
Publication Date: 6/11/2021
Citation: Zhang, H., Renschler, C., Nichols, M.H., Nearing, M.A. 2021. Long-term, process-based, continuous simulations for a small, nested rangeland watershed near Tombstone, AZ (USA): Extending model validity to include soil redistribution. Science of the Total Environment. 792, Article 148403. https://doi.org/10.1016/j.scitotenv.2021.148403.
DOI: https://doi.org/10.1016/j.scitotenv.2021.148403

Interpretive Summary: Mathematical models that simulate runoff and erosion are important for understanding and predicting how watersheds respond over time. Although many models have been applied to model runoff and sediment yield at watershed outlets, accurately predicting soil redistribution within a watershed is a substantial challenge. In this study, long-term soil loss and deposition were estimated for semi-arid watersheds within the Walnut Gulch Experimental Watershed in Southeastern Arizona using the process-based Geo-spatial interface of the Water Erosion Prediction Project model, called GeoWEPP. Model results were compared to field measurements of cesium 137 that was collected in soil samples. Cesium 137 is an artificial radionuclide with a half-life of 30 years that was produced and deposited globally by atmospheric nuclear weapon tests beginning in 1945. Amounts of cesium 137 in soil indicate erosion or deposition. Watershed topography was represented in the model at three different resolutions. Predictions of soil redistribution were improved by using higher resolution topographic data and aggregated predictions compared well with measured values.

Technical Abstract: Soil redistribution prediction is considered to be one of the main challenges for the application of watershed erosion models, especially in semi-arid regions. In this study, long-term soil loss and deposition were estimated for semi-arid watersheds within the Walnut Gulch Experimental Watershed in Southeastern Arizona using the process-based Geo-spatial interface of WEPP (GeoWEPP). Channel parameters were adjusted and validated based on reference values of soil redistribution generated from fallout radionuclide 137Cs samples. Model inputs in three resolutions - 5m, 3m and 1m pixel sizes - were used to analyze the influence of model resolutions to the soil redistribution estimation. Two methods were applied for the soil redistribution analysis: (a) comparison between simulated soil loss/deposition rates at single pixels and reference values at the specific location of each sample site; and (b) a 5m buffer was created for each sample site, and the average value of simulated soil loss/deposition rates in pixels that fell in each buffer was calculated to compare with the reference value generated using 137Cs. Aggregated values of multiple pixels within the 5m buffer of each sample site had significant higher correlation with the reference value generated by 137Cs compared to the value at the specific location of each sample site. Models with input data of higher resolution were more representative, with higher R2 and Nash-Sutcliffe Efficiency for higher input resolutions. Thus, potential errors were eliminated by taking detailed micro-topography into consideration by using input data of higher resolutions and re-aggregating the output results.