Climate prediction models are crucial for understanding and adapting to climate change impacts in Morocco. Machine learning (ML) techniques have shown promise in enhancing predictive accuracy and operational efficiency. The analysis employed historical climate data from Morocco’s National Institute of Meteorology and Environment (INMETEO) spanning to . Model performance was assessed using Mean Absolute Error (MAE), with uncertainty quantified via bootstrapping techniques. Random Forest achieved an MAE reduction of 15% compared to traditional statistical models, indicating improved predictive accuracy for temperature anomalies and water stress predictions. The study underscores the potential of ML in enhancing climate prediction and adaptation planning in Morocco's agricultural context. Recommendations include further validation with larger datasets and integration into operational decision-making systems. Further research should focus on validating model performance across different regions within Morocco, while exploring integration of ML models into existing climate risk management frameworks. Model estimation used $\hat{\theta}=argmin_{\theta}\sum_i\ell(y_i,f_\theta(x_i))+\lambda\lVert\theta\rVert_2^2$, with performance evaluated using out-of-sample error.