Can meteorological satellite data accurately estimate monthly sunshine duration? This study explores the use of EUMETSAT CM SAF data, GMTED2010 data, month number, and daylength in an artificial neural network model to estimate monthly mean sunshine duration for Türkiye. The researchers built a multilayer perceptron model using data from 45 stations for training and 12 stations for testing and validating. Comparing the model's results with ground-measured values, they found a root mean square error (RMSE) of 0.7803 h, a mean absolute error (MAE) of 0.6206 h, a mean bias error (MBE) of 0.1751 h, and a coefficient of determination (R2) of 0.9387. The results demonstrate that using new generation cloud products, elevation data, and daylength, it is possible to predict sunshine duration in regions under satellite coverage, even without measured meteorological data. This approach offers a valuable alternative for estimating sunshine duration in areas where ground measurements are unavailable or unreliable, supporting applications in climate, renewable energy, and agriculture.
Published in Theoretical and Applied Climatology, this study aligns directly with the journal's focus on climate-related research and modeling. By exploring the use of satellite data to estimate sunshine duration, the paper contributes to the journal's ongoing efforts to improve climate models and enhance our understanding of climatological processes. The research offers valuable tools for climate scientists and researchers in related fields.