Prediction of Root Zone Temperature Dynamics at Effective Depth on Lettuce Production in Greenhouse Using Sensitivity and Feature Importance Analysis with XGBoost

Root-zone temperature (RZT) strongly affects plant growth, nutrient uptake and tolerance to environmental stress, making its regulation a key challenge in greenhouse cultivation in cold climates. This study aimed to assess the potential of passive techniques, namely black polyethylene mulch and row covers, for modifying RZT dynamics in lettuce (Lactuca sativa L.) production and to evaluate the predictive performance of the eXtreme Gradient Boosting (XGBoost) algorithm. Experiments were conducted in Iğdır, Türkiye, over a 61-day period, with soil temperature continuously monitored at depths of 1–30 cm under mulched and non-mulched conditions, alongside measurements of greenhouse air temperature both with and without row covers. The application of row covers increased internal air temperature by 5.8°C, while mulching raised RZT by 0.6–1.3°C, with effects diminishing at deeper layers. XGBoost modeling achieved high predictive accuracy, with RMSE values of 0.150–0.189°C and R² values above 0.99, and feature-importance analysis indicated that neighboring soil depths were the strongest predictors of RZT. These findings show that integrating row covers and mulching can stabilize the root-zone microclimate without active heating. The XGBoost model provides a robust tool for forecasting soil temperature and supports sustainable greenhouse production in cold regions.