Time series trend analysis and forecasting of climate variability using deep learning in Thailand

Climate variability, trend analysis, and accurate forecasting are vital in a country's effective water resource management and strategic planning. Precipitation and temperature are critical indicators for assessing the effects of climate change (CC) variability. Thailand is sensitive to climatic variations, affecting the socio-economic conditions. This study quantifies climate variability and trends analysis based on precipitation, mean temperature (T-mean), and daily temperature range (DTR) across five climatic regions of Thailand. The results indicate regional variations: in the Central and Southern regions, there are increases in precipitation and warming temperatures, with substantial upward trends in annual precipitation (0.093 mm/year and 0.148 mm/year) and T-mean (0.002 degrees C/year). The Eastern and Northeastern regions display complex patterns with increased precipitation and temperatures. Also, DTR trends across regions show a decrease in temperature variability. The study offers new insights into forecasting climate variables for the different regions of Thailand between 2023 and 2028 b y utilizing two deep learning (DL) algorithms: Wavelet-CNN-LSTM and Wavelet-LSTM, which reveals high predictive accuracy. For precipitation forecasting, Wavelet-CNN-LSTM showed higher performance in the eastern region (R-2 = 0.83) and comparative efficiency in other regions. Both models faced challenges in precipitation forecasting in the northeastern and southern regions. These models performed efficiently for the DTR forecast, especially in the northern region (R-2 = 0.87 and 0.86). For T-mean, both models perform similarly with high R-2 (0.57-0.87) across all regions, suggesting a substantial model accuracy. Wavelet-CNN-LSTM provides consistent performance for DTR and T-mean forecasting. These findings underscore the importance of climate analysis and refined forecasting models.