| The surface temperature is an important parameter in the earth’s life system and atmospheric circulation system.Combining all-weather,all-time,and large-scale thermal infrared remote sensing technology and other important means can characterize the thermal state of the earth’s surface in real time.It can also provide necessary research basic data for urban heat island effect,ocean temperature change,agroecology,natural disasters and global warming,etc.,and can also make corresponding countermeasures based on relevant research results.At present,many experts and scholars at home and abroad have shown that the earthquake occurrence area has a high correlation with the surface thermal infrared anomaly in the study of the surface thermal infrared anomaly.At the same time,it also indicates that the analysis of earthquakes with thermal infrared information is a more meaningful exploratory study.Therefore,in this paper,from the surface temperature inversion of MODIS data,after data preprocessing,temperature inversion and result analysis,the time series surface temperature field data of the study area is obtained,and on the basis of this data,from statistical description and field description In terms of depth,we must mine and use the time series temperature field data,and then analyze the correlation between the eigenvalue data set and the earthquake occurrence,and discuss the change of the temperature field before and after the earthquake.Because the rapid capture of the anomalous features of the surface temperature field before the earthquake has important reference value for earthquake prediction,traditional surface anomaly feature mining often requires a lot of manpower and material resources to complete,and this article hopes to analyze the complex scientific system and explore meaningful physical quantities from the aspects of space telemetry analysis,time autocorrelation analysis,and high-order tensor analysis.Therefore,this paper proposes a method for deep mining and analysis of anomalous features of surface temperature field based on the fusion of multiple mathematical statistical models of remote sensing data,this method comprehensively utilizes remote sensing data with high time resolution,feature information with high spectral resolution,and a logical mathematical statistical model.In this paper,the Qinghai-Tibet Plateau,which is located in the Himalaya earthquake zone in the western region of China,is the main research area.The relationship between the temporal and spatial eigenvalues of the time series temperature field and the spatial distribution of seismic frequency in the past 10 years is studied.The second typical earthquake is an individual case,focusing on key scientific issues such as split-window inversion time series surface temperature field,time series temperature field characteristic value selection analysis,temperature field diffusion model and time delay effect.Therefore,the main research contents and conclusions of the paper are as follows:(1)Confirmation of research ideas and basic understanding of inversion temperature field.By analyzing the research methods and achievements of domestic and foreign experts and scholars between thermal infrared remote sensing and earthquake,a research idea of feasibility and usability is sorted out,and the mathematical basic method for realizing the research idea is determined.At the same time,based on the improved split window algorithm,the temperature of MODIS data is retrieved,and the near-real time series surface temperature field is obtained.(2)Mining and analyzing time series temperature field data.Based on the tensor calculation and statistical signal system processing methods,the spatial and temporal eigenvalues with certain physical significance in the time series temperature field are mined,and the correlation between the eigenvalues and the earthquake is analyzed,and the second order differential diffusion equation of the temperature field is further solved to judge its Whether there is an abnormal heat source field,verify whether it has the ability of earthquake prediction.(3)Explore and analyze the temporal and spatial characteristics of time series temperature field in the Qinghai-Tibet Plateau in the past 10 years.Based on MODIS data and related auxiliary data,the surface temperature field data of the Qinghai-Tibet Plateau that lasted 10 years was retrieved,and the distribution of earthquake frequency in the year,month,longitude and latitude was analyzed.The results show that the distribution gap of earthquakes in space is obvious,mainly concentrated in the west of the Qinghai-Tibet Plateau,the margin of Xinjiang in the northwest,and the area of Yunguichuan in the east and southeast of the country.The frequency,especially the frequency of large earthquakes,is mainly distributed at the time when the heat information exchange is large in the late spring and early summer and the late summer and early autumn.In the analysis of the eigenvalues of the temperature field,the experimental results show that the periodic changes of the statistic eigenvalues of the time series,the gradient of the characteristic variable in the temperature field diffusion model,the Laplacian operator and the convergence of the curl mode have an effect on the anomaly The response is very significant,which proves that there is a relationship between the aggregation of earthquake occurrence and these eigenvalues.(4)Exploring and analyzing the time-series temperature field changes in the typical single earthquake area of Jiuzhaigou 7.0.Based on the split-window algorithm,the time-series surface temperature field in the area is inverted,and the spatiotemporal feature sequences in the data are also mined.The prediction models such as multiple linear regression,optimized neural network and optimized SVM are also established.The feature data is used as the input training set sample of the model,and the earthquake level occurring in the current area is used as the output training set to establish the internal correlation between the feature value and the earthquake level.The experimental results show that: within 1-2 months before the earthquake,the average temperature,temperature field information entropy and gradient in the earthquake zone all have obvious oscillation characteristics,and these characteristic values are important in earthquake prediction,and multivariate linear The prediction results of the regression model show that there is a correlation between the temperature field eigenvalue and the magnitude;at the same time,for the short-and medium-term earthquake prediction,the use of optimized neural network and SVM algorithm to analyze the temperature field eigenvalue can improve the accuracy of earthquake prediction to a certain extent.Among them,the SVM prediction model is significantly better than other methods in earthquake prediction.At the same time,the article mainly made some innovative contributions in the following aspects:1.Focus on the time-series surface temperature field of split-window inversion,optimize the temperature inversion code.Based on the consideration of limited computer configuration,the comprehensive performance of the split-window algorithm is extremely superior,which includes algorithm accuracy,algorithm complexity and operation Efficiency,etc.At the same time,it can also build a parallel structure during time-series temperature inversion;2.Focus on the deep-level utilization of surface temperature field data,based on time-series surface temperature field,excavate and select characteristic values with strong physical significance for later analysis;3.Focus on key scientific issues such as temperature field diffusion models and time delay effects,and break through the difficulties in this research based on the in-depth use of differential diffusion equations and Granger remote correlation tests;4.Fusion analysis of the relationship between time series surface temperature field data and seismic distribution in the study area,and then add relevant auxiliary data as constraints. |
PDF Full Text Request