Prediction of Ion-Type Rare Earth Mineralization Distribution in the Shitouping Area Based on DEM and Machine Learning

The Shitouping region in Gannan is rich in rare earth resources and stands as one of the most significant ion adsorption type rare earth mineral resources, both in China and globally. This study utilizes Digital Elevation Model data to extract various geomorphological factors, including elevation, slope, aspect, curvature, and terrain roughness, and employs the Weights of Evidence method to analyze the favorable geomorphological conditions for ion-adsorption rare earth mineralization. The Support Vector Machine, Random Forest, and BP Neural Network models are used to predict the mineralization distribution of ion-adsorption rare earth ores in Shitouping, Jiangxi Province. The results indicate that topographical factors such as elevation, slope, and aspect play significant roles in the mineralization process. The elevation range of 464m - 711m and slope range of 0°-25.84° represent the most favorable conditions for mineralization. Terrain roughness and other topographic factors show a negative correlation with rare earth ore enrichment. Among the models used for mineralization prediction in the Shitouping ion-adsorption rare earth deposit, the Random Forest model performed the best, with an accuracy of 0.96 and an AUC of 0.95. The prediction results indicate that ion-adsorption rare earth ores are less prevalent in higher elevation areas. The predicted results, validated through comparison with field survey data, demonstrate high accuracy. This study integrates terrain data into the prediction of ion-adsorption rare earth ore mineralization, showcasing its potential for predicting mineralization distributions and providing crucial scientific support for the exploration and development of rare earth resources in the Shitouping region.