<![CDATA[Modern challenges in climate prediction necessitate the adoption of advanced deep learning architectures for enhanced precision in temperature forecasting. This study undertakes a comparative evaluation of various neural network designs, particularly focusing on Deep Recurrent Neural Networks (DRNN) and their extension with Gated Recurrent Units (DRNN-GRU), chosen for their proven efficacy in sequential data analysis and long-term dependency capture. Leveraging a comprehensive meteorological dataset, collected from 1961 to 2023, which includes atmospheric temperature, pressure, and precipitation levels, the research unfolds a nuanced understanding of the climate variability. The evaluation framework rigorously applies Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE) metrics to quantify model performance. The DRNN and DRNN-GRU architectures are distinguished for their superior predictive accuracy, suggesting their high potential for real-world forecasting applications. These findings are not merely academic; they imply substantial practical implications, particularly for geographic information systems where they can enhance climate monitoring and resource management. The paper culminates with recommendations for dataset expansion and diversified analytical techniques, which are critical for refining the predictive prowess of these models. This research thereby sets a benchmark for future explorations in the field and directs towards innovative avenues to augment the scientific understanding of climate dynamics.]]>
