Detecting and Modeling Long-Term Volcanic Thermal Unrest Captured by MODIS Data Years Before Eruption

Timely detection of early warning signals prior to volcanic eruptions heavily relies on remote sensing, particularly for volcanoes which are not easy to access, such as the Sunda Arc volcanoes. Girona et al. (2021) have introduced a novel method using MODIS infrared radiation data to identify a long-term, large-scale, and subtle increase in the surface temperature preceding eruptions by several months to years. This finding suggests the presence of enhanced hydrothermal activities before volcanic eruptions, complementing other monitoring data like surface deformation, gas flux, and thermal infrared hotspot. Nevertheless, when employing this methodology in volcanoes situated in regions with a variety of landuse and pronounced cloud cover, the detected surface thermal anomalies may not be controlled by volcanic activities. Our strategy involves detailed pixel labeling to improve the precise selection of pixels for background temperature, boosting volcanic thermal signal recognition.We used MODIS cloud mask data to filter out cloudy pixels. Then, we exclude the regions that may be sensitive to climate, like water bodies and urban areas. After cleaning the data, the correlation between the surface temperature evolution and the volcanic activities becomes stronger, especially for those volcanoes with more frequent eruptions. In addition, the correlation between the thermal signal and the eruptions with significant precursors, such as surface deformation and seismic activity, is stronger than those eruptions without early warning signals. One possible explanation is that in those sealed volcanic systems, the gas is accumulated and pressurized to trigger surface displacements or seismicity. At the same time, the gas can only slowly percolate through the volcanic edifice to generate long-term, large-scale thermal anomalies. To explain the origination of these large-scale thermal anomalies, we further built numerical models to explore a wide range of processes that can generate surface warming, including magma intrusion, intensified degassing, and redistribution of pore fluids due to rock permeability changes. The model and data support the hypothesis that within relatively “sealed/closed” volcanic systems, volcanic gases exsolved from magma reservoirs ascend to the surface via volcanic flanks, triggering extensive surface warming and uplift. The integration between the thermal data and the numerical model allows us to assess the practical viability of these adjustments, thereby deepening our comprehension of subsurface mechanisms and improving the predictive precision for forthcoming volcanic events.