Soils play a critical role in sustaining ecosystems and ensuring food security. Effective soil water management, particularly through drainage, is crucial for both agricultural productivity and environmental sustainability. Given the importance of soil in food security, numerous studies have focused on soil classification. However, many rely on satellite imagery, making them inaccessible to most farmers. This study explores an alternative approach to soil classification using easily accessible soil properties, such as soil moisture, soil temperature, rainfall, and air temperature. The developed classification system categorizes soils into three drainage classes: Excessively Drained, Poorly Drained, and Well Drained. The analysis utilized a dataset from the Soil Climate Analysis Network (SCAN), comprising 15 years of environmental data from over 200 U.S. sites. Various machine learning models were employed for the classification task, including Logistic Regression, Random Forest, XGBoost, LightGBM, ANN, CatBoost, CNN, and SVM. Among these, CatBoost emerged as the best performer, achieving an accuracy of 89.77% in drainage classification, along with high precision, recall, and F1-score. These results provide valuable insights for enhancing agricultural decision-making and promoting sustainable land management, supporting data-driven agricultural and conservation policies.