Predicting eruptions from precursory activity using remote sensing data hybridization

Many volcanoes produce some level of precursory activity prior to an eruption. This activity may or may not be detected depending on the available monitoring technology. In certain cases, precursors such as thermal output can be interpreted to make forecasts about the time and magnitude of the impending eruption. Kamchatka (Russia) provides an ideal natural laboratory to study a wide variety of eruption styles and precursory activity prior to an eruption. At Bezymianny volcano for example, a clear increase in thermal activity commonly occurs before an eruption, which has allowed predictions to be made months ahead of time. Conversely, the eruption of Tolbachik volcano in 2012 produced no discernable thermal precursors before the large scale effusive eruption. However, most volcanoes fall between the extremes of consistently behaved and completely undetectable, which is the case with neighboring Kliuchevskoi volcano. This study tests the effectiveness of using thermal infrared (TIR) remote sensing to track volcanic thermal precursors using data from both the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and Advanced Very High Resolution Radiometer (AVHRR) sensors. It focuses on three large eruptions that produced different levels and durations of effusive and explosive behavior at Kliuchevskoi. Before each of these eruptions, TIR spaceborne sensors detected thermal anomalies (i.e., pixels with brightness temperatures N2 °C above the background temperature). High-temporal, lowspatial resolution (i.e., ~ hours and 1 km) AVHRR data are ideal for detecting large thermal events occurring over shorter time scales, such as the hot material ejected following strombolian eruptions. In contrast, highspatial, low-temporal resolution (i.e., days to weeks and 90 m) ASTER data enables the detection of much lower thermal activity; however, activity with a shorter duration will commonly be missed. ASTER and AVHRR data are combined to track low-level anomalies months prior to an eruption and higher-energy events prior to large eruptions to develop a monitoring approach for this eruption style. Results show that strombolian eruptions produce enough energy in the pre-eruptive phase to trigger an AVHRR detection. Paired with ASTER data, the results can be extended back in time to develop a precursory timeline, which captures subtle changes in volcanic activity that would commonly go unnoticed in a single data set. Although these precursors may be volcano and eruption specific, the now sixteen-year-old database from ASTER allows this methodology to be repeatable at other volcanoes to establish a quantitative precursory baseline, which would be an improvement over current eruption classifications.