The Hunga-Tonga volcanic eruption occurred on January 15, 2022, triggering a complex interplay of thermochemical explosions, air–ocean interactions, and atmospheric physics on a planetary scale. This catastrophic event generated transoceanic tsunamis, seismic activity, and acoustic-gravity waves. The extensive scientific data collected during this event present exciting opportunities for interdisciplinary collaborative research. In addition to traditional fields such as seismology, volcanology, and tsunami simulation, the utilization of multiple satellites has provided innovative observations and unresolved insights into the mechanisms involved in the eruption's evolution. However, the presence of clouds and haze particles in the atmosphere makes it challenging for satellite sensors to accurately determine the exact start and end time of the eruption through observation alone.

To gain a profound understanding of the atmospheric evolution mechanism through optical observations, we introduced the non-stationary permutation entropy decision-tree algorithm. This algorithm reveals the inherent chaotic dynamics of space inhomogeneity and full-spectrum distributions. The chaotic nature of the atmospheric infrared radiation field implies that its long-term evolutionary process is unpredictable, thereby challenging classical physics theories in accurately depicting such intricate phenomena. In this study, we employ widely-used reservoir computing algorithms in chaotic processes to predict short-term variations in thermal infrared fields prior to volcanic eruptions (referred to as background fields). By utilizing the structural similarity index (SSIM), we can compare predicted data with real-time region measurements to determine the critical timing of abnormal events and forecast the termination of thermal anomalies.

Early warnings of thermal anomalies prior to volcanic eruptions has often been overlooked in previous research. However, early prediction not only captures timely and subtle indicators before a disaster occurs but also allows for more time in making decisions regarding disaster warnings, thereby minimizing economic and human losses.