@article{article_1506012, title={Analyzing the Relationship Between Cosmic Rays and Total Cloud Cover with LSTM Networks}, journal={Journal of Anatolian Physics and Astronomy}, volume={3}, pages={19–26}, year={2024}, DOI={10.5281/zenodo.12174802}, author={Polatoğlu, Ahmet}, keywords={Kozmik Işın (CR), Toplam Bulut Örtüsü (TCC), Uzun Kısa-Dönemli Bellek (LSTM) Ağları, İklim Modelleme.}, abstract={Understanding the interactions between cosmic phenomena and terrestrial weather patterns, particularly the relationship between cosmic rays (CRs) and cloud cover, has been a significant scientific endeavor. CRs, high-energy particles originating from supernovae, can ionize air molecules upon entering Earth’s atmosphere, potentially influencing cloud formation. Cloud cover plays a vital role in Earth’s climate system by regulating energy balance through reflecting solar radiation and trapping infrared radiation. This study aims to analyze the relationship between CRs and Total Cloud Cover (TCC) globally using Long Short-Term Memory (LSTM) networks, a type of recurrent neural network suited for time series data. We used data from the University of Oulu’s Cosmic Ray Station and the Copernicus Climate Change Service’s ECMWF European Reanalysis V5 (ERA5). A correlation matrix was constructed to identify relationships between CRs and TCC across various regions, including the Antarctic, Arctic, Europe, and globally. The results indicated generally weak and negative correlations between CR and TCC, with weak positive correlations in the Southern Hemisphere and globally. Negative correlations were more pronounced in the Antarctic and Arctic regions, suggesting region-specific climate mechanisms. The LSTM model’s predictions of CR values did not closely follow actual values, indicating a significant gap in capturing dynamic changes and peaks, and suggesting the need for more data, additional features, or further tuning. The training process showed rapid initial learning but overfitting after several epochs. The final model’s performance, measured by test mean squared error (MSE), suggested inadequate generalization. These findings highlight the complexity of modeling the CR-TCC relationship using machine learning. Future research should focus on enhancing data quality, incorporating detailed cloud metrics, and exploring advanced models to better understand CRs influence on cloud formation and climate. This study contributes to the debate on CR role in climate systems, providing insights for improved climate models and predictions.}, number={1}, publisher={Atatürk Üniversitesi}