Hotspots Forecasting Using Autoregressive Integrated Moving Average (ARIMA) for Detecting Forest Fires

Abstract

Hotspots are indicators of forest fires that detect a location that has a relatively higher temperature than the surrounding temperature. Hotspots are usually used as indicators or forest and land fires in an area, so the more hotspots, the more potential land fires occur in an area. Although not always more and more hotspots in an area the more the potential for fires. But the hotspot point can indeed be used to identify the initial occurrence of forest and land fires. Pekanbaru Meteorological Station owned by the Meteorology, Climatology and Geophysics Agency (BMKG) via NASA's Terra and Aqua remote sensing satellite detected 103 hotspots indicating forest and land fires in Riau Province, 97 of which are located in Rokan Hilir District. Therefore, based on these problems, the hotspots forecasting model research will be conducted to prevent the expansion of forest and land fires. Forecasting time series uses historical data to predict future data. The time series forecasting model used in this study applies the Box-Jenkins procedure to build ARIMA models that are compatible with the data and applies the ARIMA models obtained to make predictions of the occurrence of hotspots in July - December 2019. Based on the research results, the best model is obtained. ARIMA (2,0,2) with non-zero mean with AICc value of 408.62. Model evaluation is used to determine the accuracy of the ARIMA model (2,0,2) in forecasting by looking at the error value. Obtained the Mean Absolute Error of 3.766891 from the test results, so that the accuracy of the model in forecasting is 96.23%.