Detection of Subtle Thermal Anomalies: Deep Learning Applied to the ASTER Global Volcano Dataset

Twenty-one years of Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) global thermal infrared (TIR) acquisitions provide a large amount of data for volcano monitoring. These data, with high spatial and spectral resolution, enable routine investigations of volcanoes in remote and inaccessible regions, including those with no ground-based monitoring. However, the dataset is too large to be manually analyzed on a global basis. Here, we systematically process the data over several volcanoes using a deep learning algorithm to automatically extract volcanic thermal anomalies. We explore the application of a convolutional neural network (CNN), specifically UNET, to detect subtle to intense anomalies exploiting the spatial relationships of the volcanic features. We employ a supervised UNET network trained with the largest (1500) labeled dataset of ASTER TIR images from five different volcanoes, namely, Etna (Italy), Popocatépetl (Mexico), Lascar (Chile), Fuego (Guatemala), and Kliuchevskoi (Russia). We show that our approach achieves high accuracy (93%) with excellent generalization capabilities. The effectiveness of our model for detecting the full range of thermal emission is shown for volcanoes with very different styles of activity and tested at Vulcano (Italy). The results demonstrate the potential applicability of the proposed approach to the development of automated thermal analysis systems at the global scale using future TIR data such as the planned NASA Surface Biology and Geology (SBG) mission.