<p>Accurate forecasting of key climate indicators such as atmospheric carbon dioxide (CO₂), methane (CH₄), and global surface temperature anomalies is essential for understanding the trajectory of climate change. This study presents a comprehensive data-driven forecasting framework that utilizes both machine learning (ML) and deep learning (DL) models for long-term prediction of these critical environmental variables. Monthly CO₂ and CH₄ concentrations, along with annual global temperature anomalies, are modeled using a combination of Linear Regression, Lasso, Random Forest, and XGBoost as classical ML approaches, and LSTM, GRU, Bi-LSTM, and Bi-GRU as advanced DL architectures. Model performance is assessed using standard metrics, including MSE, RMSE, MAE, and NSE. Comparative results reveal that DL models, particularly Bi-GRU and Bi-LSTM, exhibit superior performance in capturing long-term temporal dependencies, while ensemble-based ML models like XGBoost demonstrate competitive accuracy and robustness. These findings offer valuable insights for climate researchers, policy analysts, and environmental agencies engaged in monitoring and mitigating the impacts of climate change.</p>