A review of algorithms for detecting volcanic hot spots in satellite infrared data

Since the 1980s, application of thermal infrared satellite data for volcano monitoring has rapidly evolved to become a proven operational tool. Due to the large quantities of data provided by sensors in polar and geostationary orbits, as well as the sheer number of active volcanoes on earth, processing and managing such data sets requires an enormous amount of workforce. A number of algorithms have been developed to facilitate detection, location, and tracking of hot spots of active volcanoes. A collation and review of hot spot detection algorithms developed and applied by the volcanological community reveals three main types which have been applied to date: contextual, fixed threshold, and temporal. The founding algorithms for these three classes are VAST, MODVOLC, and RST, respectively. Through comparison with manually based detections, the performance of each algorithm was tested for sustained lava flows (Etna and Stromboli), strombolian activity (Stromboli), lava dome growth and collapse (Augustine), and fumarole fields (Vulcano). It is shown that, as the number of correctly identified anomalies increases, so too does the number of false positives. Although each of the algorithms operates well within the limits and criteria of their design requirements and application, under current data restraints, no algorithm can be expected to perform perfectly.