Application Of Detection Of Ionospheric Total Electron Concentration And Infrared Long-wave Radiation Anomalies Using Time Series Based On Forecasting Model

In recent decades,satellite remote sensing technology has become an important means of seismological industry development,and has been widely used in the field of seismic study.However,due to the lack of effective methods of data processing technology for earthquake ionosphere and infrared remote sensing data,some useful information obtained from earthquake ionosphere and infrared remote sensing data has not been fully utilized.The study of earthquake ionosphere and infrared remote sensing data processing technology can fully utilize and excavate ionosphere and infrared remote sensing information,and improve the identification and extraction of seismic ionosphere and infrared remote sensing precursor related to large earthquakes.It opens up a new way to enhance the level of seismic scientific research in China and solve the important problems of seismic monitoring and prediction research in the world at present,and then these new ways are provided to enhance theoretical research.In the traditional methods of detecting pre-earthquake ionosphere electron concentration?TEC?and infrared long wave radiation?OLR?anomalies,the accuracy of the detection results depends on the rationality of anomaly detection methods.Although it is reasonable to some extent,it only both considers the inherent properties of the respective sequence data,and without considering its the uncertainty component.Therefore,this paper analyzed and processed the ionosphere TEC data inverted from the GNSS observation data of Crustal Movement Observation Network of China?CMONOC?base stations and OLR product data of NOAA satellite by using various time series forecasting models such as the Autoregressive Integrated Moving Average model?ARIMA?and the Prophet model.And combined with the earthquake events such as the Jiuzhaigou earthquake of August 8,2017 and the Mexico earthquake of September 20,2017,the effectiveness and applicability of the two time series forecasting models in sesmic anomalies detection of ionosphere TEC and infrared OLR product are verified by a series of comparative analysis experiments.1?Application of detection of ionosphere TEC anomalies using time series based on Prophet forecasting modelBased on the characteristics of ionospheric TEC with seasonal and periodic variations at various time-space scales,such as year,month and day,and combined with the Jiuzhaigou earthquake in the Sichuan province.With the epicenter of the Jiuzhaigou earthquake?33.20°N,103.82°E?as the center,and Sichuan Songpan Station SCSP?32.65°N,103.58°E?,which is the nearest GPS base station from the epicenter,is selected as the study object of building model sequence data.Through a series of steps,such as processing of building model sequence data,pattern recognition,model diagnostic test,evaluation and analysis of building model prediction,the optimal time series forecasting model method is selected,and the optimal time series forecasting model method is used to detect the earthquake anomalies in ionosphere TEC data.The experimental results are as follows:1)For the building model ionosphere TEC sequence data before the Jiuzhaigou earthquake,after six kinds of time series based on forecasting models such as Prophet model are predicted and processed through a series of building model steps,the six time series based on forecasting models all both show certain prediction effects.However,the precision of the Prophet model for predicting the background values of the building model is significantly higher than that of the other five forecasting models,while the precision of prediction modeling of the former is about 2.55 times higher than that of the ARIMA models,and about 10.74 times higher than that of the Inter Quartile Range?IQR?method.2)After the time series based on forecasting model methods is processed by the building model steps,the RMSE precision value of the predicted to build model is both characterized by the decreasing the precision of build model with the increase of the prediction time,and the forecasting model methods all both show the best forecasting effect when the training data set length is 38 days.When the best prediction model interval is established,the comparison of the total RMSE precision values is=10.5841>=3.2780>=0.8469.3)After detecting process using time series based on Prophet forecasting model,obvious anomalies can be detected.The results show that obvious ionosphere TEC anomalies appeared on the 10^th?7^th and2^nd days before the earthquake,also on the 1^st,6^th and 7^th days after the earthquake.4)Both the time series based on Prophet forecasting model and the IQR method can detect obvious anomalies,but the detection results are quite different,and the Prophet method may reduce the occurrence of false anomalies.2?Application of NOAA satellite OLR product anomalies detection using ARIMA time series forecasting modelBased on the physical characteristics of the periodic variation of infrared OLR under the influence of the physical activity of the Sun and the Earth,and combined with the Jiuzhaigou earthquake in Sichuan province and the Mexican earthquake.With the epicenter of the Jiuzhaigou earthquake?33.20°N,103.82°E?and the Mexican earthquake?18.58°N,98.47°W?as the center,and the non-seismic data points?33.50°N,103.50°E and 18.50°N,98.50°W?of infrared OLR,which is the closest to the epicenter of the Jiuzhaigou and the Mexican earthquake,is selected as the study object of building model sequence data.Through a series of steps,such as stationarity test,pattern recognition,model diagnostic test,evaluation and analysis of building model prediction,the optimal time series forecasting model method is selected,and the optimal time series forecasting model method is used to detect the earthquake anomalies in infrared OLR data.Preliminary understanding and conclusions are as follows:1)For the building model infrared OLR sequence data before the Jiuzhaigou and Mexico earthquakes,after six kinds of time series based on forecasting models are predicted and processed through a series of building model steps,the time series based on ARIMA forecasting model is generally superior to the other five models to build model and predict the total RMSE precision of background value.2)Through the building model and predicting by infrared OLR sequence data using time series based on ARIMA forecasting model,the model has good predictive performance for both Jiuzhaigou and Mexico earthquakes.And achieves the best combination of training and prediction time length when Jiuzhaigou=12 months?Jiuzhaigou=20 days and Mexico=12 months?Mexico=30 days.3)After detecting process using time series based on ARIMA forecasting model,the results show that obvious infrared OLR anomalies appeared on the 12th and 3rd days before the Jiuzhaigou earthquake,and the intensity of infrared OLR anomalies reached the maximum?39.44 W/m²?on the 3rd day before the earthquake;While obvious infrared OLR anomalies appeared on the 1st,2nd and 6th days before the Mexico earthquake,and the intensity of infrared OLR anomalies reached the maximum?50.83W/m²?on the 6th day before the earthquake.