Machine Learning-Based Classification of Geothermal Hazard Potential by Characterizing AOT, SST, and NDVI Indexes Observed by Himawari-8 Satellite

This research investigated the classification of potential geothermal environmental hazards without needing to conduct manual field visits, it is based on remote sensing data from Himawari-8 satellite imagery. The Himawari-8 satellite image data is based on the parameters Aerosol Optical Thickness (AOT), Sea Surface Temperature (SST), and Normalized Difference Vegetation Index (NDVI). Environmental changes are observed in areas that become training data representing geothermal areas, namely Sarulla Geothermal and areas representing non-geothermal areas, namely Bangka Belitung. Classification of potential geothermal environmental hazards was done using machine learning by taking data on these parameters in the test data area representing other geothermal areas, namely Kamojang Geothermal and other non-geothermal areas, namely East Kalimantan. The five machine learning algorithms used are algorithms on the classification problem boundary that have been selected from various classification algorithms. Taking these five algorithms is expected to add a strong reason for the classification accuracy, where the accuracy generated from the five tends to be high. That is, the resulting accuracy is above 80% on average, with the Decision Tree algorithm as the best algorithm in this study, with an accuracy of 96%. The results show that the AOT, SST, and NDVI values in the Sarulla and the Kamojang geothermal areas tend to reach dangerous levels when compared to the AOT, SST, and NDVI values in nongeothermal areas such as Bangka Belitung and East Kalimantan. The indexes observed by the Himawari-8 Satellite are usually accumulated normatively. Still with the use of machine learning from remote sensing data, the observation of potential geothermal hazards becomes easier to do. The near real-time data obtained from the Himawari-8 satellite is essential to immediately identify and effectively mitigate potential hazards to minimize their impact.