Integrating Machine Learning and Geospatial Analysis for Nitrate Contamination in Water Resources Management: A Case Study of Sinkholes in Winkler County, Texas

This study used machine learning methods and spatial analysis to examine groundwater quality in Winkler County, Texas, focusing on nitrate pollution. By analyzing 85 years of groundwater data from six aquifers, the study uses advanced machine learning models Random Forest, Decision Tree, Linear Regression, and XGBoost to predict contamination levels and explore spatial and temporal trends. These models were chosen because of their ability to handle larger and more complex datasets and their ability to capture nonlinear relationships between water quality parameters and environmental variables. These machine learning algorithms are particularly effective at identifying patterns and interactions that may not be obvious with traditional analytical methods, and get more reliable and accurate results. Our decadal analysis specifically identified systematic fluctuations in nitrate levels, with a notable increase since the early 2000s, driven by the synergistic effects of rising temperatures and intensified agricultural land use. Climate change, pressured by rising temperatures and lessened precipitation, along with natural factors such as the formation of sinkholes, has been identified as a key driver of groundwater quality fluctuations. Elevated nitrate levels were mostly related to agricultural irrigation and excessive use of synthetic fertilizers. The machine learning model also highlights how land cover changes and human activities are contributing to groundwater quality deterioration. This research reinforces the value of integrating machine learning and spatial analysis for groundwater management. This is especially true in areas affected by sinkholes. It provides important information to reduce man-made impacts to water quality in West Texas.