Machine Learning Models in Climate Prediction and Adaptation Planning for Democratic Republic of Congo

The Democratic Republic of Congo (DRC) is vulnerable to climate variability, which impacts agriculture, water resources, and health systems. A hybrid ensemble model combining Random Forest and Gradient Boosting Machines (GBM) was employed to predict temperature anomalies with an uncertainty of ±2°C over a spatial scale of 10 km². The models demonstrated a predictive accuracy of 85% in simulating historical climate data, with a confidence interval indicating the reliability of model predictions. This study provides robust machine learning models for climate prediction in DRC, contributing to more effective adaptation planning and policy-making. Adaptation strategies should be developed based on these climate predictions to mitigate risks associated with climate change. 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.