BENGUET FOREST FIRE BURNED AREA TAAL ASH EXTENT ESTIMATION USING SUPPORT VECTOR MACHINE THRESHOLDING TECHNIQUES

摘要

Diwata-2 is the second Philippine microsatellite developed to provide Earth-observation data for disaster assessment and environmental monitoring. Since its launch in October 2018, the satellite has collected 25,811 images for various applications. In this study, the capability of Diwata-2 to monitor large-scale disasters is demonstrated through the identification of ash-stricken areas in the recent Taal Phreatic eruption and forest fire-affected areas of Benguet. Ashfall and burned area extent were classified from the Taal volcano eruption and Benguet forest fire, respectively. Diwata-2 images used were taken on January 6, 2020 and January 27, 2020 capturing the volcano's surrounding areas, and February 29, 2020 for areas in Benguet Province. The extent of both disasters' damages was delineated through index difference thresholding and Support Vector Machine (SVM) classification. For the index difference thresholding. Normalized Difference Vegetation Index (NDVI) and Burned Area Index (BAI) were used to identify ash stricken and burned areas, respectively. Similar methods were applied to Sentinel-2 for further evaluation using a higher resolution optical imagery. The ashfall extent measured from Diwata-2 was estimated at 20.359.64 ha and 55,669.24 ha for NDVI thresholding and SVM classification, respectively. The detected areas for both Diwata-2 and Sentinel-2 coincide with the major ash plume's general directions to the north and the subsequent ash plumes to the southwest direction of Taal volcano. In Benguet forest fire, the burned area extent was calculated to be around 2,714.59 ha using BAI thresholding. Identified fire occurrences for Diwata-2 and Sentinel-2 generally match with each other except for cloud-covered areas. Overall findings demonstrate the applicability of Diwata-2 satellite as a tool for large scale disaster monitoring and assessment in the Philippines. Future works include validation through field records and reports. The results underline the significance of remote sensing in disaster monitoring by providing spatial information to aid in emergency response, damage assessment, and future policymaking.