By Andi Anderson
Gully erosion is one of the most damaging forms of soil erosion, often occurring after heavy rainfall and leaving deep channels in farmlands. These gullies can cause severe sediment loss and nutrient runoff into nearby water bodies.
Researchers from the University of Illinois Urbana-Champaign have developed a new AI-based method to predict and interpret gully erosion.
The method uses machine learning models and an explainable AI technique called SHAP—SHapley Additive exPlanations. This combined approach allows farmers and land managers to better understand where erosion is likely and why it happens.
The team studied land in Jefferson County, Illinois, which is part of the Big Muddy River watershed. The area mainly grows corn and soybeans. Using maps created from elevation data between 2012 and 2015, the researchers tracked gully formation.
They then identified 25 environmental factors such as soil type, slope, vegetation, and rainfall that influence erosion risk.
Rather than using a single AI model, the team used stacking—a method that combines multiple models for better results. They tested 44 stacking models and found the best one achieved 91.6% accuracy, higher than the 86% from the best individual model.
To make these results more understandable, the researchers applied SHAP to the models. SHAP helped identify the annual leaf area index—how much crop leaves cover the ground—as the most influential factor in erosion control.
“Explainable AI provides metrics that allow you to understand how different variables influence model predictions,” said Dr. Jorge Guzman, a co-author of the study.
This transparent AI system gives farmers a clearer picture of erosion threats and guides better land-use decisions. It also supports broader environmental policies by offering a reliable way to assess erosion risks.
The research was funded by the USDA’s National Institute for Food and Agriculture and is published in the Journal of Environmental Management.
Photo Credit: gettyimages-sasiistock
Categories: Illinois, Education, Sustainable Agriculture