بحوث سكوبس — علي كريم عبيد الجحيشي

Climate change is a big global concern, so the prediction of environmental change is quite important. Deep learning approaches include LSTMs, GRUs, statistical models, and traditional forecasting systems, which can be difficult to manage complicated and always shifting climate data. By providing a fresh method for continuous-time modeling, Neural Ordinary Differential Equations (Neural ODEs) help make forecasts more accurate and flexible in dynamic environmental changes. This paper uses neural ODEs to investigate methods of climate pattern forecasting. It cleans and fills in any missing figures from climatic data gathered from satellites and meteorological stations. It creates and compares a neural ODE-based model versus traditional methods of predicting. Using error metrics including Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), methods including varying learning rates and avoiding overfitting are performed to increase accuracy. This helps to assess the models. Early results show that when detecting long-term climate patterns including temperature and rainfall, neural ODEs outperform traditional models. Changing and sophisticated data patterns are now better under control. This work shows how neural ODEs could enhance climate prediction, hence enhancing environmental monitoring and disaster readiness. Future research will concentrate on even more improving their accuracy by merging these models with physics-based approaches. © 2025 IEEE.