Artificial Intelligence in the NASA Volcano Sensorweb: Over a Decade in Operations

Volcanic activity can occur with little or no warning. Increasing numbers of space borne assets can enable coordinated measurements of volcanic events to enhance both scientific study and hazard response. We describe the use of space and ground measurements to target further measurements as part of a worldwide volcano monitoring system. We utilize a number of alert systems including the MODVOLC, GOESVOLC, US Air Force Weather Advisory, and Volcanic Ash Advisory Center (VAAC) alert systems. Additionally we use in-situ data from ground instrumentation at a number of volcanic sites, including Iceland. Artificial Intelligence Software plays a key role in the Volcano Sensorweb. First, several in-situ volcano monitoring networks use “intelligent” data interpretation software to trigger alerts that can then be used to allocate network resources, notify human agents, and even task space observations. Second, the Earth Observing One (EO-1) spacecraft uses Artificial Intelligence Software to automatically task the spacecraft to execute observations. Third, EO-1 also interprets thermal data onboard to allow for faster notifications of volcanic activity. Finally some data interpretation steps use intelligent software such as Random Decision Forest Methods used to automatically estimate volcanic plume heights in Worldview-2 Imagery.