Skip to main content
U.S. flag

An official website of the United States government

Validating machine learning hypothesis about soil thickness impacts on hydrologic fluxes using a process-based surface-subsurface model

Presentation Date
Tuesday, December 11, 2018 at 8:00am
Walter E Washington Convention Center Hall A-C (Poster Hall)



In this work we try to use a process-based model to test hypotheses generated by data mining. Data mining analysis suggested that a key physiographic difference between Appalachian basins and their southern Atlantic neighbors in the US is the soil thickness. Thicker soil in the southern Atlantic basins helped forming an extensive groundwater flow system that dominate the hydrologic functioning and has a significant influence on flood peaks. While this conclusion is statistically strong, there are weaker explanations such as slope that could alternatively explain the observed relationships between streamflow and storage. However, numerical experiments suggest the soil thickness is the main reason for the observed behavior. To test these competing hypotheses, we employed a process-based surface subsurface processes model. The model was able to successfully capture hydrologic responses in the Appalachian basins and the observed streamflow-storage relationships. This study implements the experiments of soil thickness alteration through the reliable process-based hydrologic model. The experiment results illustrate that soil thickness indeed impacts on the relationship between soil water storage and streamflow.

Funding Program Area(s)