A novel spectrally-derived method for detecting sea-water discoloration around submarine volcanoes by combining Sentinel 2A/B-MSI and Landsat 8/9-OLI data

Shallow eruptions from submarine volcanoes can hinder the navigation of ships and alter the biological response of marine ecosystems. Hydrothermal vents and ash-laden plumes can spread across the sea surface for weeks, affecting the water column's optical properties. Systematic in situ observations (i.e., underwater observations and hydro-acoustic and seismic arrays) are logistically complicated to deploy and usually expensive to carry out before and during an eruptive event. Conversely, satellite remote sensing can provide timely and continuous information about volcanic activity around dangerous sites, contributing to the assessment of pre-, syn-, and post-eruptive phenomena. Sea-water discoloration is one of the most significant indicators of underwater volcanic activity, as its accurate,  timely and continuous detection can help revealing possible precursor processes of submarine volcanic eruptions and tracking their evolution. Most published studies have characterized discolored water patches after major eruptions by assessing their reflectance patterns using multispectral ocean color data acquired by MODIS, VIIRS, and Sentinel-3 OLCI. While these sensors may enable the timely detection of submarine eruption features, their coarse spatial resolution makes them unsuitable for mapping discolored patches whose size and spatial dynamics are on a ten- or hundred-meter scale. The improved spatial resolution offered by Sentinel 2-MSI and Landsat 8/9-OLI data (10–60 m) ensures accurate mapping of sea-water discoloration, even for small and weak plumes. In this framework, we have proposed a novel, spectrally-derived method to accurately detect and map discolored plumes around submarine volcanoes in oligotrophic oceans by integrating Sentinel 2A/B-MSI and Landsat 8/9-OLI satellite data. The developed method, which combines two discoloration algorithms, was tested using a yearly (2022) MSI-OLI integrated dataset around a representative test case, namely the Kavachi Volcano (Solomon Islands, Southwest Pacific Ocean). It exhibited satisfactory validation metrics, recording overall accuracies (OAs) close to 90% for both single and integrated (multi-sensor) configurations. Despite omission errors (OEs) ranging from 18% to 20%, the very low (around 2%) commission errors (CEs) demonstrated its high level of reliability in mapping discolored waters of volcanic origin. Furthermore, the proven exportability of this method to the Kaitoku Volcano (Japan, Western Pacific Ocean) confirms its capability in detecting underwater volcanic activity regardless of geographic location or the chemical composition of discolored seawater. This method could serve as an automated early warning tool to support the operational monitoring of submarine volcanoes arranged by maritime surveillance systems.