Accurate prediction of environmental events is a result of the combination of climatic and environmental variables. The Random Forest model was employed in this study to examine the contribution of various climatic and environmental variables to the phenomenon in question under three scenarios: (1) all variables combined, (2) climatic variables only, and (3) environmental variables only. Training was conducted using 70% of the data and testing on 30%. Measures of performance like R-squared, MAE, MSE, RMSE, and MAPE were used to assess predictive accuracy, whereas the relative contribution of each variable was calculated by employing Gini importance index.

Results indicate that optimal prediction performance was achieved when all variables were included (R-squared = 0.98, MAE = 0.0271, RMSE = 0.0414), but the performance was lower when climatic variables alone were used (R-squared = 0.89). Environmental variables alone gave performance which was very close to that of the full-parameter (R-squared = 0.97) case, reflecting their better role in prediction of the model. Variable importance analysis revealed the top contributing drivers of the variables as urbanization (NDBI) and vegetation condition (VCI), then Albedo, PDSI, AOD, and LST. Soil moisture and actual evapotranspiration were climatic drivers with average contribution, while atmospheric pressure and wind speed contributed the least.

These findings demonstrate that Random Forest models can effectively capture complex interactions in environmental systems and emphasize the core role of environmental variables, particularly urbanization and vegetation condition, in predictive modeling. This research provides insights for future environmental monitoring and management strategies in arid and semi-arid regions.