Modeling soil erodibility and critical shear stress parameters to account for spatio-temporal variability

Authors: LEE, SANGHYUN - University Of Illinois; CHU, MARIA - University Of Illinois; GUZMAN, JORGE - University Of Illinois; Flanagan, Dennis

Submitted to: Soil and Tillage Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/15/2021
Publication Date: 12/15/2021
Citation: Lee, S., Chu, M.L., Guzman, J.A., Flanagan, D.C. 2021. Modeling soil erodibility and critical shear stress parameters to account for spatio-temporal variability. Soil and Tillage Research. 218:105292. https://doi.org/10.1016/j.still.2021.105292.
DOI: https://doi.org/10.1016/j.still.2021.105292

Interpretive Summary: Human activities, such as agricultural tillage and crop production, while very essential, can disturb the soil and accelerate rates of soil loss. Soil erosion by water, and off-site losses of sediment and agricultural chemicals can detrimentally affect soil productivity and off-site water quality. Often computer simulation models are used to estimate the amounts of potential erosion at a specific location that has unique climate, soil, and topographic characteristics, under various cropping and management systems. The alternative systems are then evaluated to select ones that minimize soil loss while minimally impacting farming practices and farm profitability. Internally within the computer models, adjustments are made to the erodibility (soil’s ability to resist detachment) of the soil through time as affected by tillage, cropping systems, and other factors. In this research, we attempted to determine if simpler, more widely available factors could be used to estimate the erodibility adjustments, so that spatial erosion estimates could be made by other erosion prediction systems developed within geographic information systems (GIS). Results showed that the regression equations developed predicted the erodibility parameter adjustments accurately when using the commonly available spatial data compared to determining the adjustments within the full stand-alone soil erosion computer simulation model. These findings have substantial impacts on geospatial modelers developing new soil erosion prediction tools, as well as soil conservation personnel who may have access to new tools in the near future.

Technical Abstract: The rate of soil erosion from agricultural fields is driven by climate erosivity and by soil erodibility or resistance, commonly represented through the interrill (Ki) and rill erodibility (Kr), and critical shear stress (TauC) parameters. These parameters are affected by factors showing high variability in time and space, such as soil properties, land use, and management practices. Estimating the site-specific time-varying values for Ki, Kr, and TauC can require laborious field or laboratory experiments and detailed datasets. The objective of this research was to develop a non-linear regression model for estimating temporal adjustments for Ki, Kr, and TauC considering different types of crops and management practices. One thousand sampling sites were randomly selected across the Kaskaskia watershed in Illinois, and the needed parameters to develop the regression models were extracted from the two-year simulation results of the Water Erosion Prediction Project (WEPP) model on a daily basis. The predicted values by the regression models showed good agreements with the sample data, capturing spatio-temporal variations of the three variables under different crops and management practices. In addition, daily soil loss estimations using the predicted adjustments showed good agreements with the WEPP simulations, confirming their possible use in soil erosion models. This is especially useful as the regression models require only a few explanatory variables, which are available at high resolution across the United States. Therefore, the models will facilitate estimating spatio-temporal variations of interrill and rill erodibility, and critical shear stress and thus a new platform of large-scale soil erosion models can be developed using readily available environmental datasets.