Automated Identification of Volcanic Plumes using the Ozone Monitoring Instrument (OMI)

Abstract:

Volcanic eruptions are a global phenomenon which are increasingly impacting human populations due to factors such as the extension of population centres into areas of higher risk, expansion of agricultural sectors to accommodate increased production or the increasing impact of volcanic plumes on air travel. In areas where extensive monitoring is present these impacts can be moderated by ground based monitoring and alert systems, however many volcanoes have little or no monitoring capabilities. In many of these regions volcanic alerts are generated by local communities with limited resources or formal communication systems, however additional eruption alerts can result from chance encounters with passing aircraft. In contrast satellite based remote sensing instruments possess the capability to provide near global daily monitoring, facilitating automated volcanic eruption detection. One such system generates eruption alerts through the detection of thermal anomalies, known as MODVOLC, and is currently operational utilising moderate resolution MODIS satellite data. Within this work we outline a method to distinguish SO₂ eruptions from background levels recorded by the Ozone Monitoring Instrument (OMI) through the identification and classification of volcanic activity over a 5 year period. The incorporation of this data into a logistic regression model facilitated the classification of volcanic events with an overall accuracy of 80% whilst consistently identifying plumes with a mass of 400 tons or higher. The implementation of the developed model could facilitate the near real time identification of new and ongoing volcanic activity on a global scale.