Predictive Dredging Models for the Upper Mississippi River and Illinois Waterway

The Mississippi River and Illinois Waterways are high-traffic navigable channels that are used for commercial transportation. The U.S. Army Corps of Engineers (USACE) Rock Island District spends millions annually on dredging operations to maintain a 9-foot navigation channel. Currently, dredge coordinators rely on institutional knowledge and river forecasts to develop 10 to 14-day schedules, which is a reactive approach and vulnerable to personnel turnover and emergency groundings. This capstone developed machine learning models to extend forecasting windows and improve resource coordination. Using 25 years of river gage observations and shoaling rates derived from the Corps Shoaling Analysis Tool (CSAT), this study implemented ARIMA, PCA, xGBoost and LSTM approaches. The xGBoost model achieved the strongest regression performance with RMSE values of 0.77 ft/yr (Illinois Waterway) and 1.32 ft/yr (Mississippi River), representing 28.3% and 31.0% improvement over the baseline ARIMA model. At the pool level, LSTM models outperformed xGBoost in 11 of 17 pools. In classification tasks, xGBoost achieved 72.8% accuracy. Week of year and tributary gages were identified as strong predictors. These results provide a foundation for predictive modeling of dredging needs in inland waterways and can offer a framework for proactive, data-driven resource management in federal dredging operations.