Hybrid Time Series and Neural Network Models for Sunshine Duration Modeling

Meteorological observation data used in this study were obtained from the Central Weather Administration (CWA), Taiwan. The datasets were collected in collaboration with the World Meteorological Organization (WMO), incorporating satellite observations and numerical weather prediction outputs. All data are acquired in real time with high spatial and temporal resolution, following standardized and scientifically validated procedures. The study period spans from March 1, 2013, to March 30, 2023, covering seven observation sites across Taiwan. The stations are located at altitudes ranging from 31 to 195 meters, with longitudes ranging from 120.2367\u00b0E to 121.2398\u00b0E and latitudes from 23.0384\u00b0N to 25.0067\u00b0N. To streamline data management and address potential naming ambiguities due to the similarity of station names, these seven stations are labeled sequentially as DF1 to DF7. The target variable for this study is Sunshine Duration, with the following ten factors selected as feature variables: Barometric Pressure, Temperature, Maximum Temperature, Minimum Temperature, Humidity, Wind Speed, Wind Direction, Precipitation, Solar Irradiance, and PM2.5.