This paper proposes a new approach to predicting dry season periods by using annual cumulative rainfall for the past 35 years to determine dry season parameters (onset, end, and duration) in Agam District, West Sumatra, Indonesia. After determining the dry season parameters, the trend of each parameter was determined by using the Mann-Kendall test and Sen's slope estimator. Finally, the parameters for future dry seasons were predicted by using the autoregressive integrated moving average (ARIMA) model. The results showed that (1) on average, the dry season period in Agam District starts on April 28 and ends on September 17 with a duration of 142 days, (2) dry season periods would be delayed in the future, as indicated by positive trends in the past 35 years for both the onset and end with probabilities of 78.83 and 84.42%, respectively, (3) the ARIMA model successfully predicted the dry season with good performance for the onset (NSE = 0.747) and very good performance for the end (NSE = 0.769), and (4) the dry season period in the next 5 years would start between April 29 and May 5 and end between September 19 and October 4. This study suggested that relevant stakeholders should reassess the schedule for water use and distribution of agricultural water, postpone transplanting to the first week of July in future dry seasons, and prepare additional irrigation water to prevent water deficit.
