Modeling and forecasting air temperature using the SARIMA approach (Case study: Juanda meteorological station)

Currently, climate condition, especially air temperature conditions, are abnormal and difficult to predict because of global warming. Temperature is a measure of the coldness or hotness of a condition or something else. Considering that temperature is one of the important elements of climate to be continuously observed, it is necessary to predict the temperature for each month. Predicting air temperature can help the government and health organizations take necessary preventive actions to protect the public from the adverse effects of extreme temperature changes. Some regions with extreme air temperatures in Indonesia are in Surabaya City and Sidoarjo Regency. Air temperature data is usually seasonal data that can be predicted using the Seasonal Autoregressive Integrated Moving Average (SARIMA) method. SARIMA utilizes past and present data to generate an accurate short-term forecast suitable for observations from interrelated time series. In the SARIMA method, assessing stationarity and determining the best model for a dataset is done through analysis of ACF and PACF plots. By using the SARIMA method in modelling and forecasting the monthly average air temperature from Juanda Meteorological Station for the period of January 2009 to December 2018, it was found that the best model is the probabilistic SARIMA (0, 0, 0) (0, 1, 1)12 model with a MAPE value of 1.658%. This means that the forecast results can be considered highly accurate, and the model is suitable as a reference for predicting the monthly average air temperature from Juanda Meteorological Station in the subsequent periods.