Remote Detection of Pumice Rafts and Discolored Water from Submarine Volcanic Eruptions Using Satellite Imagery

Abstract:

Up to 80% of volcanic eruptions occur underwater, and these submarine volcanoes can have an impact on large-scale earth systems (e.g. nutrient input via discharged pumice rafts). Eruptions from volcanoes that are only partially underwater or are close to the surface are easier to detect due to their sub-aerial plumes of ash and SO₂. However, deep water volcanic eruptions are difficult to detect as they may only release floating pumice clasts or hydrothermal fluids (observed as discolored water) into the surrounding area. The short-lived nature of many eruptions can exacerbate the issue of detection. To improve eruption detection, we can utilize extensive collections of global satellite imagery of 10-30 m resolution (e.g. Landsat-8, Sentinel-II) from the past decade. In this study, we utilize Top-of-Atmosphere Reflectance data collected from Sentinel-II, which offers wide-swath, high-resolution imaging and data sampling of 13 spectral bands. In analyzing this satellite imagery collection, one of the difficulties we face is the large amount of data that must be processed. We addressed this challenge by developing a new methodology to detect submarine eruptions through the supervised classification of satellite images on Google Earth Engine (GEE). GEE is a freely-available web-based platform for planetary-scale scientific analysis and visualization of geospatial datasets. Using GEE, we trained a Random Forest classifier on areas with known submarine volcanoes and recorded sightings of pumice rafts and discolored waters. Once trained, we applied this classifier to Sentinel-II images to detect submarine eruptions globally. In this study, we will present results from the analysis of satellite imagery in the Kermadec-Tonga subduction zone to assess the presence of any new submarine eruptions over the 4-year Sentinel-II coverage period. Additionally, with our algorithm, we have found new raft detections from Rabaul and Anak Krakatau that are not directly associated with new submarine eruptions, but with older, sub-aerial eruptions. These observations suggest that beached pumice rafts may be re-suspended by major weather events or large-scale erosion events (e.g. landslides, tsunamis). Thus, in some near-shore regions, the presence of a pumice raft may not necessarily indicate a new submarine eruption.