<![CDATA[Indonesia is a tropical archipelago located along the equator, where air temperature patterns exhibit seasonal trends and unstable fluctuations. This instability can impact several sectors, including agriculture—making it difficult for farmers to determine planting and harvesting times—electricity demand, which increases during hotter periods, and public health, as erratic weather may reduce productivity and elevate the risk of diseases such as dehydration, asthma, and respiratory infections. This study aims to evaluate the performance of the Trigonometric, Box-Cox Transformation, ARMA Errors, Trend, and Seasonal (TBATS) model and the Seasonal Autoregressive Integrated Moving Average (SARIMA) model in forecasting air temperature in Indonesia. The dataset used comprises 2-meter air temperature records in Indonesia from January 1940 to August 2024, obtained from ECMWF. The evaluation method applied is cross-validation with a rolling basis. The results show that the RMSE for the TBATS model is 21.3843%, while the SARIMA model has an RMSE of 21.2958%. These results indicate that SARIMA has a slightly better performance than TBATS. However, both methods perform well in forecasting air temperature in Indonesia, as their RMSE percentages are within an acceptable range. This research is expected to contribute to the scientific literature on air temperature forecasting in Indonesia and encourage further studies on hybrid models that integrate TBATS and SARIMA. Additionally, it may support efforts to mitigate the adverse impacts of air temperature changes in the country.]]>
Copyrights © 2025
