<p>Understanding the complex relationships between drought intensity (DI) and forest structure/environment is essential for understanding forest ecosystems. To explore these relationships, we calculate the DI using the medfate package for a set of 2681 circular permanent sample plots located throughout the Hyrcanian forest. Subsequently, we employed three machine learning models: random forest (RF), support vector regression (SVR), and generalized linear model (GLM) to predict DI using a comprehensive set of environmental variables. The performance of the models was evaluated by coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE), and Nash and Sutcliffe coefficient of efficiency (NSE). RF and SVR models outperformed the GLM, achieving the highest R² (0.76 for both RF and SVR), along with the RMSE (SVR = 0.064, RF = 0.065) and MAE (SVR = 0.045, RF = 0.047). The RF model had the highest NSE (0.94) and was evaluated as very good, whereas the GLM model showed the lowest NSE (0.28) and was considered unsatisfactory. We found that forest density, basal area, and precipitation were the most influential variables driving DI in the Hyrcanian Forest. Together, these factors emphasize the interplay between biotic structure and abiotic conditions in determining DI. Given the high predictive accuracy, these findings can help forest managers identify priority areas for intervention and develop plans to improve forest resilience. Although our research focused on the Hyrcanian Forest, our findings and methodologies have broad application in temperate forests facing similar drought threats worldwide.</p>