<p>Meteorological conditions have the greatest immediate influence on fire occurrence and behavior; this is why they are commonly used in forest fire prediction and detection systems, such as forest-fire forecasting models based on the machine learning (ML) algorithms. In this paper, we assess the role of the weather/climate in forest fire occurrence using machine learning and the SHapley Additive exPlanations (SHAP) technique. The aim of this study is to determine the contribution of various weather variables and indices as drivers of fire activities in Algeria. To perform this assessment we address two issues: first, we have evaluated and compared the performance of different ML models using the Algerian forest fires dataset; second, we apply the SHAP method to assess the importance of variables for each model. We focus on analyzing the influence of the weather/meteorological conditions to provide insights into their role in the ML-models predictions. The SHAP analysis reveals that the meteorological conditions contribute more strongly to the prediction when combined in the computation of the Fire Weather Index (FWI) system components (calculated from the meteorological elements) rather than considered separately. Temperature is identified as a relevant feature across all models, while relative humidity (RH) shows no effect in all models, except for the KNN model. Meteorological features are identified as less influential factors for ‘predicting the “not fire” class. The feature importance analysis using the SHAP method, which improves the interpretability of the considered ML models, is the main highlight of our investigation. In fact, the comparison of different models for the same phenomenon and the assessment of the role of weather/climate in forest fire occurrence are important phases in the process of comprehending the fire regime and its drivers. This approach ensures the choice of appropriate methodologies for future work on the topic.</p>