<![CDATA[Global warming increases the urgency of accurate local temperature forecasting, particularly in Sleman, Yogyakarta, a region characterized by diverse topography and high exposure to climate-related risks such as volcanic activity, agricultural vulnerability, and rapid urbanization. Such conditions increase the urgency for localized predictive models that can support agricultural planning, energy management, and disaster preparedness. This research used quantitative approach with a comparative predictive modelling design to predict the weekly average air temperature in Sleman by comparing two models: the Fourier Series regression and the Markov Switching Autoregressive (MSAR) model. The Fourier Series was selected for its ability to capture smooth seasonal and periodic behavior typical of climatological data, whereas the MSAR model was employed to accommodate regime shifts and nonlinear structural variations. The dataset comprises 127 weekly observations from January 2023 to June 2025 (BMKG), the data were split into 70% training and 30% testing. Model performance was assessed using GCV, MSE, MAE, MAPE, and residual diagnostics. Results show that the Fourier Series model performs substantially better, achieving lower GCV (0.3520), MSE (0.00415 training; 0.00114 testing), and MAE (0.34015 training; 0.12940 testing), as well as lower MAPE (1.26% training; 0.47% testing). In contrast, the MSAR model yields higher errors with GCV (0.5747), MSE (0.9113 training; 0.4686 testing), MAE (0.8005 training; 0.5512 testing), and MAPE (1.96% training; 1.34% testing). These results indicate that Sleman’s temperature dynamics characterized by stable oscillatory patterns with minimal regime shifts are more effectively captured through harmonic decomposition. The study reinforces the importance of periodic modeling for mixed-topography regions like Sleman and recommends future research integrating additional climatic variables, hybrid statistical–machine-learning frameworks, and longer time spans to improve responsiveness to extreme events and nonlinear atmospheric behavior.]]>
