نوع مقاله : ترویجی
نویسنده
استادیار پژوهشی بخش تحقیقات حفاظت خاک و آبخیزداری، مرکز تحقیقات، آموزش کشاورزی و منابع طبیعی خوزستان، سازمان تحقیقات، آموزش و
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسنده [English]
This study aimed to evaluate the performance of machine learning algorithms in predicting soil water repellency intensity in the loess soils of northern Iran, a region prone to water repellency due to its specific soil texture and climatic conditions. A total of 45 surface soil samples were collected from various locations across Golestan and Mazandaran provinces. For each sample, a set of physical and chemical properties—including organic carbon, organic matter, soil texture, aggregate stability, pH, and EC—were measured. The water drop penetration time (WDPT) test was used to quantify the degree of water repellency. Three machine learning algorithms, including Decision Tree (DT), Random Forest (RF), and XGBoost, were applied to model and predict the WDPT index. The results indicated that the XGBoost model outperformed the others, achieving an RMSE of 14.7 and an R² of 0.42. Moreover, organic carbon was identified as the most influential variable. These findings suggest that advanced machine learning algorithms can serve as effective tools for analyzing and predicting nonlinear phenomena in soil studies and watershed management.
کلیدواژهها [English]
 آبخوان', './data/abkhansj/coversheet/cover_fa.jpg'))