This study examines forecasting methods for rainfall prediction in drought-prone areas of Northeastern Thailand. Monthly rainfall data from January 1992 to December 2022 (372 months) were collected from seven meteorological stations across Loei, Udon Thani, Khon Kaen, Chaiyaphum, Nakhon Ratchasima, Surin, and Buriram provinces. A comparative analysis evaluated the accuracy of the combined forecasting and Box–Jenkins methods using mean squared error (MSE). The combined forecasting method was constructed from the integration of three approaches—Decomposition, Winter’s exponential smoothing, and the theta method. The study found that the combined forecasting method in Loei, Udon Thani, Khon Kaen, Chaiyaphum, Nakhon Ratchasima, and Buriram had MSE values of 9687.72, 3425.06, 2159.14, 10506.03, 3981.36, and 9366.54, respectively. These results were lower than for the Box–Jenkins method, for 6 meteorological stations, accounting for 85.71 percent, while the Box–Jenkins method in Surin meteorological station had an MSE value of 11 991.23, which in turn was lower than the combined forecasting method, accounting for 14.29 percent. Overall, the combined forecasting method was found to be appropriate for predicting the highest monthly rainfall data, followed by the Box–Jenkins method.
Thaithanan et al. (Sat,) studied this question.