Joint use of Sentinel-2 and Sentinel-1 data for rapid mapping of volcanic eruption deposits in Southeast Asia

Indonesia and Papua New Guinea (PNG) have 97 active volcanoes with high concentration of human life in the very close proximity to them. In case of a volcanic eruption, provision of detailed information on affected regions is very crucial to support rescue and humanitarian relief organizations. In this paper, we present a semi-automated unsupervised knowledge-based region growing procedure that utilizes Synthetic Aperture Radar (SAR) data, from Sentinel-1, and optical data, from Sentinel-2, for mapping land surface changes after volcanic eruptions. With initial seed points, being placed on active volcano vents and areas affected by thermal anomalies (derived from Sentinel-2), the region growing procedure considers interferometric coherence data in unvegetated sites, and radar brightness and polarimetric decomposition features at vegetated sites. We selected five eruptive events that occurred between 2018 and 2021 at the Indonesian volcanoes of Karangetang, Semeru, Sinabung and at Ulawun Volcano on PNG. The eruption patterns varied with respect to duration, spatial extent and ejected volcanic materials. The results indicated that usage of radar brightness features with interferometric coherence already gives good change delineation. However, in the Ulawun test case, where heavy ash and scoria fall occurred, the addition of polarimetric decomposition features substantially improved the output accuracy due to the improved detection of ash deposits. The presented change detection method is straight forward to implement, and will strongly improve rapid mapping activities during as well as after major volcano eruptions.