Combining high temporal and high spatial resolution thermal infrared datasets to detect volcanic precursors

Most volcanoes produce some level of precursory activity prior to an eruption. In certain cases, the precursors can be interpreted to make forecasts about the time and magnitude of the impending eruption. Furthermore, subtle changes in this activity can be used to determine the style of an eruption at volcanoes known for differing eruption types. This study focuses on the effectiveness of using thermal infrared (TIR) spaceborne sensors to track volcanic precursors from thermal changes at the summit of volcanoes using data from both the ASTER and AVHRR sensors by focusing on the Kamchatka (Russia) region. At Bezymianny volcano, a clear increase in activity commonly occurs before an eruption, which has allowed predictions to be made months ahead of time. In other cases, such as the Tolbachik eruption in 2012, no precursors were detected before the large scale effusive eruption. However, most volcanoes do not fall into these extremes, as is the case with Kliuchevskoi volcano. Three large eruptions (e.g., 2005, 2007, 2009) have occurred at Kliuchevskoi in the past decade. Before each of these eruptions, AVHRR and ASTER detected thermal anomalies (pixels with radiant energy significantly above the background). High temporal but low spatial resolution (i.e., hours; 1 km) AVHRR data are ideal for detecting high energy events occurring over short time periods. In contrast, high spatial but low temporal resolution (i.e., days to weeks; 90 m) ASTER data enables the detection of much lower levels of activity. These smaller thermal anomalies are more commonly associated with the longer time scale pre-eruptive phase at Kliuchevskoi. The problem occurs where attempting to detect patterns in these anomalies over time. Only volcanic activity that produces a large amount of radiant energy can be seen with the relatively poor spatial resolution of AVHRR, and ASTER will commonly miss shorter duration activity. In this study, we use ASTER data to track low-level anomalies months prior to an eruption. By combining these with data from AVHRR, a methodology of monitoring these seemly unpredictable eruptions can be established. Data from the 2005, 2007, and 2009 Kliuchevskoi eruptions were analyzed. Results show that large strombolian eruptions are the only style to produce enough energy in the pre-eruptive phase to trigger an AVHRR detection. However, when paired with ASTER data, results can be extended further back in time to develop a precursory timeline. Both the 2009 and 2005 eruptions behaved similarly, with a VEI 1 eruption followed by inactivity for 6-8 months before the onset of a VEI 2 eruption. In contrast, the 2007 eruption had a relatively long period of inactivity before a VEI 2 eruption. However, the TIR data show summit temperatures have clear fluctuations that begin 6 to 8 months before the eruption as the result of strombolian eruptions. A significant increase in the precursory intensity occurs 2 months before the larger eruption and is easily detect with ASTER. Merging these datasets produces a product that extends the precursory lead time, thereby capturing subtle changes in volcanic activity that would go unnoticed in a single data set. Although some of these precursors are likely volcano-specific, this methodology can be repeated at other volcanoes to establish a greater understanding of their precursory baseline.