The Impact of Climate Change on Soil Water Content with Considering Machine-Learning Methods as a Downscaling Tool

Abstract Submission: Soil water content (SWC) is a crucial element of the hydrological cycle, affecting how precipitation is distributed among runoff, infiltration, and evapotranspiration. It plays a vital role in agriculture and ecosystem stability, particularly in regions with scarce water resources. Understanding SWC levels is also critical for predicting and preventing landslides and slope instability. Since precipitation is one of the primary factors affecting SWC, changes in its patterns and volume due to climate change can significantly impact hydrological and mechanical properties of soil on the slope. This study examines how changes in precipitation caused by climate change affect SWC in the Anacostia watershed in the District of Columbia and Maryland. Using General Circulation Model (GCM) outputs, we develop future SWC projections to support assessment of future landslide risks. We apply machine learning (ML) techniques to downscale GCM outputs and a hydrological model to simulate SWC based on rainfall. The results demonstrate that ML-based downscaling, in contrast to traditional statistical methods, helps account for the uncertainty between different models by integration of multiple GCM outputs. This makes ML approaches more robust for regional downscaling. Furthermore, calibration of the hydrological model using observed flow rate data from the Anacostia River (2015–2022) confirms the model’s reliability in simulating SWC. Finally, we generate a future SWC map reflecting anticipated precipitation changes due to climate change.

Learning Objectives/Expected Outcome (Optional) :