A Potential Hybrid Deep Learning Approach to Temperature Prediction Using MODIS Satellite Data and Historical Records

This study introduces a novel hybrid approach for forecasting daily temperatures by integrating satellite imagery from MODIS Terra and Aqua with historical temperature records using a ResNet Convolutional Neural Network (CNN). The focus is on Acquedolci and Caronia, Sicily, regions with extensive mango plantations where accurate temperature forecasts are essential for crop protection and yield optimization. By leveraging the ResNet architecture’s capability to extract spatial features, combined with XGBoost and ARIMA for tabular data integration and residual correction, the model excels at identifying spatial and temporal patterns. The results demonstrate the potential of this approach to improve agricultural forecasting and climate modeling, paving the way for enhanced environmental monitoring and precise agricultural planning.