PHILIPPINE VOLCANIC HAZARDS: A MODIFIED APPROACH OF THE VPI30 CALCULATION USING GIS EXTRACTION TOOLS AND SENTINEL-2 IMAGERY DATA

The Volcano Population Index at 30 kilometers (VPI30) is a key metric used in the National Volcano Threat Assessment (NVTA) to estimate the population at risk within a 30-kilometer radius of an active volcano. This metric is especially important to the Philippines which is home to more than 300 volcanoes, 24 of which are active. To support research and development on the mitigation of volcanic-related disasters, the Philippine Institute of Volcanology and Seismology (PHIVOLCS) uses a modified version of the VPI30 to fit the Philippines setting. The current Philippine VPI30 methodology assumes uniform distribution across barangay (village) boundaries. This study aims to improve the existing methods through the use of high-resolution Sentinel-2 satellite imagery and geospatial analysis techniques. This is investigated to enhance understanding of the population distribution which is especially important in rural areas where households are generally observed to be clustered. Using land classification models, the built-area pixels within village boundaries intersecting the 30-kilometer buffer zone are identified. These built-area pixels will be proxies for human settlement and population. The ratio of built-area pixels inside the buffer zone to the total built-area pixels for each village is then calculated, the result is multiplied by the village population to get a more refined estimate of population distribution. Using this modified VPI30 methodology, a representative village yielded at least a 70% decrease in calculated exposed population. While the method marks an improvement in identifying exposed populations, it has some limitations. The reliance on broad land classifications, such as "built areas," can impact precision, particularly in distinguishing between residential and non-residential structures. Future work will focus on refining land classification models to better identify housing clusters and validate consistency when applied to multiple villages.