Trend Analysis and Forecasting of Long-Term Temperature Data to Anticipate Local Warming: A Case Study of Jawa Timur, Indonesia

Abstract

Numerous studies have confirmed the significant rise in the mean land surface air temperature in recent decades at the global scale. However, studies on the land surface air temperature at local scales in Indonesia in recent decades are rare. Such studies are necessary to comprehend the imminence of the threat of climate change in regions with intensive agricultural industry, such as Jawa Timur, Indonesia. The first objective of this study is to analyse the trend of air temperature in six locations in Jawa Timur based on at least 30-year air temperature data retrieved from NOAA Global Summary of the Day data. After the significance or insignificance of the temperature trend is verified, the second objective is to compare the accuracies of two forecasting methods, namely Exponential Smoothing and ARIMA, on the long-term air temperature data in each location. The Mann-Kendall test along with sieve-bootstrapping is used in the air temperature trend analysis. Mean Absolute Error (MAE) metric is used to measure the accuracy of the methods. The trend analysis shows that statistically significant monotonic rises in air temperature are found in four locations, namely Bawean, Kalianget Madura, Surabaya Perak, and Banyuwangi. It is also found that ARIMA (3,1,3) is able to forecast the temperature data for 2021 with MAE < 1.30 degrees Celsius and it outperforms the Exponential Smoothing method.