关键词: 地震热异常, 偏度, 背景场, 时间域差值, 空间域差值

Abstract: The seismic thermal anomaly has been confirmed by a large number of earthquake case studies, but the weather factor is still the limiting factor for the fine extraction of thermal anomalies. In the past, in order to eliminate the short-term noise of meteorology, the spatial domain difference value was used to replace the temperature value in the original remote sensing image, but the cloud mask will affect the spatial domain mean value, which makes the spatial domain difference value not representative. This causes thermal anomalies due to differences in cloud masks and spatial distribution of temperature. In previous RST studies, the thermal anomaly thresholds were judged to be different values of 1.6, 2, 3, and 4, and there was no stable value. When the number of background fields changes under the same threshold, the thermal anomaly area will change. Therefore, when constructing the background field, the time domain difference is used first and then the spatial domain difference is calculated to remove the weather influencing factors. By introducing skewness abnormal data monitoring algorithm, different sample numbers correspond to different thresholds in thermal abnormality judgment. In this experiment, the areas of 32.70°~42.72°N and 96.37°~106.65°E in Gansu, Qinghai and Sichuan were used as the study area. The new algorithm is used to analyze the spatial and temporal distribution of thermal anomalies in Hutubi M_S6.2 earthquake on December 8, 2016, Jiuzhaigou M_S7.0 earthquake on August 8, 2017, and Jinghe M_S6.6 earthquake on August 9, 2017. Statistical analysis is made on the statistical relationship between earthquakes of magnitude 4 and above and thermal anomalies in the study area from January 1, 2003 to August 31, 2019. Experiments and results show that: 1)By analyzing the annual mean surface temperature map in 2004, it was found that the night temperature in Qinghai Province in the study area was lower than that in other areas, and the night temperature in the southern Gansu and desert areas in the northeast of the study area was relatively high. On June 26, 2004, most of the cloudless images were in the low-temperature region of Qinghai Province, which resulted in a low average value of the spatial domain. The temperature difference between the high-temperature and cloudless regions of Qinghai Province and Gansu Province was relatively large, so the spatial domain difference caused the false thermal anomalies. The difference in time domain is calculated before calculating the difference in spatial domain, and the thermal anomaly almost disappears. Using the difference in time domain and then calculating the difference in spatial domain will eliminate the false thermal anomaly caused by the difference in temperature and spatial distribution of clouds. 2)Through experiments, it is found that the skewness in different sample numbers used in this paper corresponds to different thresholds and the position and area of thermal anomalies are closest when the RST threshold is 4.5. When the RST is at a threshold of 4.5 and the number of background fields is 12, 14, and 16, respectively, the thermal anomaly area becomes smaller as the number of background fields increases, but for the skewness of different thresholds corresponding to different sample numbers in the background when the number of fields is 12, 14, and 16, respectively, the change in thermal anomaly area is small. It is obtained that under the premise that the number of samples satisfies the abnormality judgment, different sample numbers corresponding to different thresholds will basically keep the thermal anomaly area unchanged when the number of background fields changes, and the thermal anomaly threshold judgment stability gets stronger. 3)By extracting the thermal anomalies of Hutubi M_S6.2 earthquake on December 8, 2016, Jiuzhaigou M_S7.0 earthquake on August 8, 2017, and Jinghe M_S6.6 earthquake on August 9, 2017, it was found that the thermal anomaly before the Jiuzhaigou M_S7.0 earthquake moved around the epicenter from the northwest Jungong Fault to the northeast Longxian-Baoji Fault in a clockwise direction; Before the Hutubi M_S6.2 earthquake, the whole thermal anomaly moved from south to north to the epicenter; Before the Jinghe M_S6.6 earthquake, the overall thermal anomaly moved from the vicinity of the northwest Tuster-Ballake Fault to the epicenter and was “/” shaped. 4)There were 210 earthquakes with M_S≥4.0 in the study area, and a total of 98 thermal anomalies occurred. The total number of thermal anomalies corresponding to the earthquake is 46, accounting for 46.94%; the number of earthquakes with thermal anomalies found in 210 earthquakes is 53, accounting for 25.2%, among them, 73.6%were preceded by thermal anomalies. By analyzing the relationship between thermal anomalies over the years and earthquakes, it is found that the TPR value increases with the magnitude of earthquake. Although the number of strong earthquakes in the study area is small and the TPR value of strong earthquakes in this experiment is not representative, the overall trend shows that the probability of earthquake thermal anomalies increases with the increase of the earhtquake magnitude. The greater the probability of occurrence of seismic thermal anomalies during strong earthquakes, the greater the significance is for future earthquake prediction. In this paper, we use two types of data, cloud and surface temperature, to optimize the algorithm to remove the influence of weather factors on seismic thermal anomalies. However, the influence of topography and ground features on temperature still exists, and there is currently no good way to judge and rule it out.

Key words: seismic thermal anomaly, skewness, background field, time domain difference, spatial domain difference