<![CDATA[Forest and land fires (Karhutla) are a significant environmental threat in Riau Province with substantial ecological, health, and economic impacts. This research develops an integrated artificial intelligence (AI)-based application for Karhutla mitigation. The method uses a quantitative approach with a system development design. The dataset includes 87,600 spatiotemporal data items (2020-2024) from MODIS/VIIRS, BMKG, and Sentinel-2. Machine learning models (Random Forest and XGBoost) were trained on the data for real-time hotspot prediction. The XGBoost model achieved an accuracy of 91.2% (AUC 0.871), outperforming RF (88.1%, AUC 0.847). The results are integrated into a Geographic Information System through three main modules: (1) Prediction and Visualization, (2) Early Warning, and (3) Reporting and Analysis. A usability test involving 15 field users resulted in a System Usability Scale score of 82.5 (Excellent). A 4-week implementation pilot achieved a detection rate of 88.9% and a suppression rate of 86.7%, reducing the response time from 4.2 hours to 1.1 hour. The application integrates solutions for real-world AI challenges: model drift (automated retraining), black box (SHAP interpretability), and knowledge gap (training program). The research demonstrates AI technology for disaster mitigation operations with an ROI of 480% and an (investment) payback period of 10.3 months.]]>
Copyrights © 2025
