Coarse-graining research of the thermal infrared anomalies before earthquakes in the Sichuan area on Google Earth engine

Abstract

Seismo-induced Thermal infrared (TIR) anomalies has been proposed as a significant precursor of earthquakes. Several methods have been proposed to detect Thermal infrared anomalies that may be associated with earthquakes. However, there is no comparison of the influence for Thermal infrared extraction methods with a long time statistical analysis. To quantify the effects of various techniques used in Thermal infrared anomaly extraction, in this paper, we offer a complete workflow of their comparative impacts. This study was divided into three parts: anomaly detection, statistical analysis, and tectonic factor research. For anomaly detection, daily continuous nighttime surface temperature (ConLST) data was obtained from the Google Earth Engine (GEE) platform, and each different anomaly detection method was used to detect Thermal infrared outliers in the Sichuan region (27°-37°N, 97°-107°E). During statistical analysis, The heated core model was applied to explore Thermal infrared anomalies which is to filter anomalies unrelated to earthquakes by setting time-space-intensity conditions. The 3D error diagram offers scores to assume the best parameter set using training-test-validation steps. In the final part, we considered information on stresses, active faults, and seismic zones to determine the optimal parameters for extracting the Thermal infrared anomalies. The Kalman filter method detected the highest seismic anomaly frequency without considerating the heating core condition. The Autoencoder and Isolation Forest methods obtain the optimal alert type and parameter set to determine if the anomaly is likely earthquake-related. The RST method performs optimally in the final part of the workflow when it considers physical factors such as active faults, seismic zones, and stresses. However, The six methods we have chosen are not sufficient to contain the entire Thermal infrared anomaly extraction. The consideration of tectonic factors in the research remains poorly developed, as statistical methods were not employed to explore the role of constructive factors. Nevertheless, it is a significant factor in comparing anomaly extraction methods and precursor studies.