Earthquake monitoring and prediction method based on electromagnetic satellite plasma observation data anomaly

Abstract

The application discloses a method for monitoring and predicting earthquakes based on abnormal electromagnetic satellite plasma observation data, and relates to the technical field of earthquake ionospheric plasma abnormal signal identification, and comprises the following steps: constructing a main earthquake data set and a plasma observation abnormal data set, establishing the space-time correlation characteristics of earthquake events and abnormal disturbance events based on the main earthquake data set and the plasma observation abnormal data set, training a multi-dimensional earthquake monitoring and prediction model based on the space-time correlation characteristics, inputting plasma observation data obtained in real time by an electromagnetic satellite into the trained multi-dimensional earthquake monitoring and prediction model, and obtaining potential dangerous earthquake prediction information. The application establishes the space-time correlation characteristics of earthquake events and ionospheric plasma abnormal disturbance events, trains a multi-dimensional earthquake monitoring and prediction model, and realizes effective identification and limitation of the occurrence time, direction and magnitude of a potential dangerous earthquake based on electromagnetic satellite plasma observation data.

Description

Technical Field

[0001] This invention relates to the field of earthquake ionospheric plasma anomaly signal identification technology, specifically to an earthquake monitoring and prediction method based on anomalies in electromagnetic satellite plasma observation data. Background Technology

[0002] Earthquakes cause enormous casualties and property damage, making the study of earthquake precursor anomalies a major focus of the scientific community. In recent years, earthquake prediction has gained increasing attention as a crucial means of disaster prevention and mitigation. However, the complexity and randomness of earthquakes mean that their prediction remains a global technological challenge. Traditional earthquake monitoring methods primarily rely on various ground-based observation devices, but these devices have limitations in terms of spatial coverage and effectiveness.

[0003] With the development of space information technology, electromagnetic satellites have become an important technical means for earthquake monitoring research. The Zhangheng-1 electromagnetic monitoring experimental satellite is my country's first geophysical field detection satellite with the core mission of ionospheric environment detection and earthquake precursor information exploration. One of its scientific objectives is to explore new approaches to earthquake monitoring and prediction. Research shows that during the gestation and occurrence of earthquakes, crustal tectonic activity can induce abnormal disturbances in ionospheric parameters, i.e., seismic ionospheric anomalies. Ionospheric plasma parameters are a type of observation that is highly sensitive to seismic activity in electromagnetic satellite observations, and their data structure is simple and easy to process, making them the preferred data source for monitoring seismic ionospheric anomalies.

[0004] However, current earthquake monitoring and prediction technologies utilizing electromagnetic satellite plasma observation data are mostly limited to post-earthquake retrospective studies and have not yet developed into earthquake monitoring and prediction technologies suitable for operational applications. Secondly, the spatiotemporal correlation between earthquakes and plasma anomalies lacks scientific quantitative standards, making it difficult to effectively predict the time, location, and magnitude of potentially hazardous earthquakes based on observed anomalies. This results in insufficient practicality of ionospheric observation data in the field of earthquake monitoring and prediction. Summary of the Invention

[0005] The purpose of this invention is to provide an earthquake monitoring and prediction method based on anomalies in electromagnetic satellite plasma observation data, aiming to solve at least one problem existing in the prior art.

[0006] The technical solution of this invention is: an earthquake monitoring and prediction method based on anomalies in electromagnetic satellite plasma observation data, the method comprising the following steps: Acquire historical earthquake data and construct a mainshock dataset based on the historical earthquake data; Acquire plasma observation data accumulated by electromagnetic satellites, and construct a plasma observation anomaly dataset based on the accumulated plasma observation data; Based on the mainshock dataset and the plasma observation anomaly dataset, the spatiotemporal correlation characteristics between seismic events and anomalous disturbance events are established; Based on the aforementioned spatiotemporal correlation characteristics, a multi-dimensional earthquake monitoring and prediction model based on plasma observation data is trained. Real-time plasma observation data from the electromagnetic satellite is acquired, and the real-time plasma observation data is input into the trained multi-dimensional earthquake monitoring and prediction model to obtain potential hazardous earthquake prediction information.

[0007] In some embodiments of the present invention, constructing the mainshock dataset based on the historical earthquake data includes: Obtain the earthquake center of each historical earthquake event in the historical earthquake data, and delineate the target spatial region based on the earthquake center and a first spatial range. The earthquake event with the largest magnitude within the first time range in the target spatial region is identified as the mainshock event. The mainshock events are classified and stored according to their magnitude ranges to obtain the mainshock dataset.

[0008] In some embodiments of the present invention, constructing a plasma observation anomaly dataset based on the plasma observation data includes: For each trajectory data to be judged, several revisited trajectory data are selected as background trajectory data. Based on the background trajectory data, a plasma background value is constructed and the upper and lower limits of the background are determined. The trajectory data to be judged is compared with the upper and lower boundaries of the background, and data segments exceeding the upper and lower boundaries of the background are identified as candidate abnormal disturbances. The candidate anomalous perturbations are screened to obtain valid anomalous perturbations, which are then stored as the plasma observation anomalous dataset. The plasma observation anomalous dataset is continuously updated based on the accumulation of observation data.

[0009] In some embodiments of the present invention, the step of screening the candidate abnormal perturbations to obtain effective abnormal perturbations specifically includes: If the center of the candidate abnormal disturbance is located within a preset seismic zone and the duration of the candidate abnormal disturbance is within a preset time threshold, then the candidate abnormal disturbance is determined as a valid abnormal disturbance. Calculate the disturbance amplitude of the effective abnormal disturbance and determine whether the disturbance type of the effective abnormal disturbance is an increasing anomaly or a decreasing anomaly.

[0010] In some embodiments of the present invention, establishing the spatiotemporal correlation features between seismic events and anomalous disturbance events specifically includes: The spatial impact range of each mainshock event is determined based on the correspondence between earthquake magnitude and spatial influence radius, and a preset time window is defined based on the occurrence time of each mainshock event. Effective abnormal disturbances located within the spatial influence range and occurring within the preset time window are identified as abnormal disturbance events that have a spatiotemporal correlation with the main shock event. The number of anomalous disturbance events associated with each mainshock event and the number of mainshock events associated with each anomalous disturbance event are counted to obtain the correlation distribution between anomalous disturbance events and mainshock events, forming the spatiotemporal correlation characteristics.

[0011] In some embodiments of the present invention, training the multi-dimensional earthquake monitoring and prediction model specifically includes: For different magnitude ranges, a temporal distribution model, a spatial distribution model, and a disturbance amplitude correlation model are constructed. The temporal distribution model represents the temporal correlation between anomalous disturbance events and the main shock event, the spatial distribution model represents the spatial correlation between anomalous disturbance events and the main shock event, and the disturbance amplitude correlation model represents the correlation between disturbance amplitude and magnitude. The multi-dimensional earthquake monitoring and prediction model is obtained by jointly training the time distribution model, the spatial distribution model, and the disturbance amplitude correlation model.

[0012] In some embodiments of the present invention, constructing the time distribution model includes: Using the occurrence time of the abnormal disturbance event as the baseline zero point, a preset time range before and after the occurrence of the abnormal disturbance event is defined. The preset time range is divided into multiple consecutive time units according to a preset time interval. The cumulative number of main shock events in different magnitude intervals within each time unit is counted. The distribution function of the probability of the main shock event occurring after the occurrence of the abnormal disturbance event is obtained by fitting the data. Constructing the spatial distribution model includes: For each magnitude interval, the azimuth and spatial distance of the earthquake center of the main shock event relative to the center of the anomalous disturbance event within the magnitude interval are statistically analyzed. The center of the anomalous disturbance event has a spatiotemporal correlation with the main shock event. The theoretical influence radius is determined based on the correspondence between earthquake magnitude and spatial influence radius. The theoretical influence radius is used as a constraint condition for spatial attenuation to construct the probability density distribution function between the center of the anomalous disturbance event and the earthquake center. Constructing the disturbance amplitude correlation model includes: For each magnitude interval, the amplitude distribution of the effective anomalous disturbances corresponding to the mainshock event within the magnitude interval is statistically analyzed; A regression model is established between the disturbance amplitude and the earthquake magnitude. The regression model uses the law represented by the correspondence between the earthquake magnitude and the spatial influence radius as a constraint condition.

[0013] In some embodiments of the present invention, the step of inputting the current plasma observation data into the trained multi-dimensional earthquake monitoring and prediction model to obtain earthquake prediction information specifically includes: The real-time plasma observation data is compared with the plasma background value to identify abnormal disturbances in the real-time observation data, and the occurrence time, center location, and disturbance amplitude of the abnormal disturbances in the real-time observation data are extracted. The occurrence time, center location, and amplitude of the abnormal disturbances in the real-time observation data are input into the multi-dimensional earthquake monitoring and prediction model; The predicted occurrence time output by the time distribution model, the predicted occurrence location and distance output by the spatial distribution model, and the predicted magnitude range output by the disturbance amplitude correlation model are obtained.

[0014] This invention provides an earthquake monitoring and prediction system based on electromagnetic satellite plasma observation anomalies, the system comprising: The mainshock dataset construction module is used to acquire historical earthquake data and construct the mainshock dataset based on the historical earthquake data. An anomaly dataset construction module is used to acquire plasma observation data accumulated by electromagnetic satellites and construct an anomaly dataset of plasma observations based on the accumulated plasma observation data. The association construction module is used to establish the spatiotemporal correlation characteristics between earthquake events and anomalous disturbance events based on the mainshock dataset and the plasma observation anomaly dataset. The training module is used to train a multi-dimensional earthquake monitoring and prediction model based on plasma observation data based on the spatiotemporal correlation features. The prediction module is used to acquire real-time plasma observation data from the electromagnetic satellite, input the real-time plasma observation data into the trained multi-dimensional earthquake monitoring and prediction model, and obtain potential hazardous earthquake prediction information.

[0015] This invention establishes the spatiotemporal correlation characteristics between earthquake events and anomalous disturbance events by constructing a standardized mainshock dataset and a plasma observation anomaly dataset, and trains a multi-dimensional earthquake monitoring and prediction model. This enables the effective identification and limitation of the time, location, and magnitude of potentially hazardous earthquakes based on electromagnetic satellite plasma observation anomalies, significantly improving the practicality and accuracy of earthquake monitoring and prediction. Attached Figure Description

[0016] Figure 1This is a flowchart of the steps of the earthquake monitoring and prediction method based on electromagnetic satellite plasma observation data anomalies provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the structure of an earthquake monitoring and prediction system based on anomalies in electromagnetic satellite plasma observation data provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0018] Reference Figure 1 This paper illustrates an earthquake monitoring and prediction method based on anomalies in electromagnetic satellite plasma observation data, provided by an embodiment of the present invention.

[0019] This method may specifically include the following steps: Step 101: Obtain historical earthquake data and construct a mainshock dataset based on the historical earthquake data.

[0020] The mainshock dataset refers to the collection of data that retains only the mainshock event after removing foreshocks and aftershocks from historical earthquake data and stores it according to magnitude intervals; historical earthquake data refers to earthquake event data recorded by earthquake catalogs published by seismic networks, which includes information such as the time of occurrence, epicenter latitude and longitude, magnitude, and focal depth; the mainshock event refers to the earthquake event with the largest magnitude within a specific time and space range; and the magnitude interval refers to a continuous range of intervals divided according to the magnitude of the earthquake, used for classifying and managing earthquakes of different intensities.

[0021] In the specific implementation of this invention, historical earthquake data is first obtained from the earthquake catalog published by the seismic network. Then, for each historical earthquake event, a target spatial region is delineated centered on its epicenter. This target spatial region is specifically a rectangular area spanning 2° latitude, extending 1° south and 1° north, and 2° longitude, extending 1° west and 1° east. Simultaneously, a 15-day time window is defined based on the event's occurrence time. Within this spatiotemporal range, all earthquake events are identified, and the event with the largest magnitude is determined as the mainshock event; the remaining events are discarded as foreshocks or aftershocks. Finally, the identified mainshock events are classified and stored according to three magnitude intervals: 5.0 ≤ M < 6.0, 6.0 ≤ M < 7.0, and M ≥ 7.0, forming a mainshock dataset.

[0022] In some embodiments of the present invention, step 101, "constructing a mainshock dataset based on historical earthquake data," may specifically include the following sub-steps: Sub-step 1011: Obtain the earthquake center of each historical earthquake event in the historical earthquake data, and delineate the target spatial region based on the earthquake center and the first spatial range.

[0023] Among them, the earthquake center refers to the location of the earthquake epicenter recorded in the earthquake catalog, usually represented by latitude and longitude coordinates; the first spatial range refers to the spatial scale parameters used to delineate the target spatial region; the target spatial region refers to the spatial range used for foreshock and aftershock identification, which is centered on the earthquake center and bounded by the first spatial range.

[0024] Sub-step 1012: Determine the earthquake event with the largest magnitude within the first time range in the target spatial region as the main shock event.

[0025] The first time range refers to the time window defined based on the time of the earthquake event, which is 15 days before and after in this embodiment; the earthquake event with the largest magnitude refers to the event with the largest magnitude value among all earthquake events in the target spatial area and the first time range.

[0026] Sub-step 1013: Classify and store the mainshock events according to their magnitude ranges to obtain the mainshock dataset.

[0027] Among them, the magnitude interval refers to the continuous range of intervals divided according to the magnitude of the earthquake. In this embodiment, it includes three intervals: 5.0≤M<6.0, 6.0≤M<7.0, and M≥7.0. Categorized storage means storing the mainshock event into different data sets according to its magnitude interval.

[0028] In some embodiments of the present invention, the first spatial range and the first temporal range can be adjusted according to the magnitude range. For example, for earthquake events of 5.0 ≤ M < 6.0, a spatial range of 2° × 2° and a time window of 10 days before and after are used; for earthquake events of 6.0 ≤ M < 7.0, a spatial range of 3° × 3° and a time window of 15 days before and after are used; and for earthquake events of M ≥ 7.0, a spatial range of 4° × 4° and a time window of 20 days before and after are used. By using different spatial ranges and time windows for different magnitude ranges, the embodiments of the present invention can more accurately identify foreshocks and aftershocks of different magnitudes, improving the purity and representativeness of the mainshock dataset.

[0029] In some embodiments of the present invention, the method of dividing magnitude intervals can be adjusted according to actual application needs. For example, for refined earthquake prediction research, the magnitude intervals can be further subdivided into smaller intervals such as 5.0≤M<5.5, 5.5≤M<6.0, 6.0≤M<6.5, 6.5≤M<7.0, and M≥7.0; for rapid earthquake trend judgment, it can also be simplified to two intervals: M<6.0 and M≥6.0. In addition, for seismic activity research in a specific region, a non-uniform magnitude interval division method can be set according to the historical earthquake distribution characteristics of that region.

[0030] Step 102: Obtain plasma observation data accumulated by electromagnetic satellites, and construct a plasma observation anomaly dataset based on the accumulated plasma observation data.

[0031] Among them, electromagnetic satellites are artificial satellites that carry detection payloads such as electric fields, magnetic fields, plasma, and high-energy particles to observe physical quantities such as electromagnetic fields, electromagnetic waves, ionosphere, and high-energy particles of the Earth and near-Earth space in space orbit, including but not limited to the Zhang Heng-1 electromagnetic monitoring satellite; plasma observation data accumulated by electromagnetic satellites refers to orbital observation data containing plasma physical quantities such as electron density and electron temperature collected from the ionosphere by electromagnetic satellites during long-term operation; plasma observation anomaly datasets refer to the collection of plasma anomaly disturbance data related to earthquakes identified from three-dimensional observation data such as electromagnetic satellites by comparing each orbital observation data with its revisited orbit background and screening under limited conditions.

[0032] In a specific implementation of this invention, plasma observation data accumulated by electromagnetic satellites are first acquired. Utilizing the orbit revisit characteristics of electromagnetic satellites, for each orbital data point to be judged, several preceding revisited orbital data points are selected as background orbital data. Statistical analysis is performed on the background orbital data at latitudinal intervals, calculating the median, first quartile, third quartile, and interquartile range within each latitudinal interval. Upper and lower boundaries of the background data are then set based on the interquartile range method. The orbital data to be judged is compared with these background upper and lower boundaries, identifying data segments continuously exceeding these boundaries as candidate anomalous disturbances. These candidate anomalous disturbances are then screened based on spatial correlation, duration, and disturbance amplitude, and those meeting the criteria are determined as valid anomalous disturbances. Finally, the valid anomalous disturbances and their related attribute information are stored as a plasma observation anomaly dataset.

[0033] In some embodiments of the present invention, step 102, "constructing a plasma observation anomaly dataset based on accumulated plasma observation data," may specifically include the following sub-steps: Sub-step 1021: For each orbital data to be judged, select several revisited orbital data as background orbital data, construct plasma background values ​​based on the background orbital data, and determine the upper and lower limits of the background.

[0034] The orbital data to be judged refers to the observation data of a single satellite orbit that needs to be anomaly detected. In the specific implementation of this invention, the orbital data to be judged includes a sequence of plasma physical quantities sampled along the orbit. The revisit orbital data refers to the observation data of multiple orbits that pass through the same spatial location as the orbital data to be judged. The global orbit of the Zhangheng-1 satellite is divided into 76 groups of revisit orbits, and each group of orbits has the same spatial trajectory. Therefore, for each orbit, the first few strictly revisit orbits can be obtained as background orbital data, which serves as the data basis for constructing the plasma background value. For example, the first 5 strictly revisit orbits can be selected as background orbital data.

[0035] In some embodiments of the present invention, the number of background track data is not fixed. Specifically, the number of historical revisited tracks used to construct the background value can be adjusted according to the data quality of different tracks. For example, for tracks with stable data quality, the first 5 historical revisited tracks are selected as background track data; for tracks with large fluctuations in data quality, the first 7 historical revisited tracks are selected as background track data to increase statistical stability; for tracks with many missing data, the first 3 historical revisited tracks can be selected as background track data, and a confidence threshold is set to evaluate the reliability of the background value.

[0036] Plasma background value refers to the normal fluctuation range of plasma physical quantities obtained based on background orbit data statistics, while the upper and lower boundaries of the background refer to the anomaly discrimination thresholds set according to the plasma background value. In the specific implementation of this invention, the plasma background value includes the median, the first quartile Q1, and the third quartile Q3 statistically calculated at latitudinal intervals, and the upper boundary Ne. up =Q3+1.5×IQR, background lower bound Ne down =Q1-1.5×IQR, where IQR=Q3-Q1.

[0037] In other specific embodiments of the present invention, the method for determining the upper and lower boundaries of the background can be adjusted according to the data distribution characteristics. For example, for non-normally distributed plasma observation data, the percentile method can be used instead of the interquartile range method, with the 5th percentile as the background lower bound and the 95th percentile as the background upper bound. Taking the electron density data observed by the Zhang Heng-1 satellite as an example, if the 5th percentile of the historical revisit orbit data within a certain latitudinal interval is... The 95th percentile is Then the upper and lower boundaries of the background are set to . and The orbital data to be judged is lower than or higher All data segments were identified as candidate anomalies. Furthermore, to avoid plasma anomalies caused by non-seismic factors, such as strong solar flares or geomagnetic storms, this application introduces the geomagnetic activity index as an auxiliary parameter. The geomagnetic activity index is a core indicator used to describe the degree of global disturbance in the Earth's magnetic field, quantifying the average disturbance of the Earth's magnetic field every three hours using a value from 0 to 9; the higher the value, the stronger the geomagnetic disturbance. Therefore, in specific implementations of this invention, anomaly detection is suspended when the geomagnetic activity index > 5 to eliminate false anomalies caused by space weather interference.

[0038] Sub-step 1022: Compare the trajectory data to be judged with the upper and lower boundaries of the background, and identify data segments that exceed the upper and lower boundaries of the background as candidate abnormal disturbances; After determining the upper and lower boundaries of the background, if the plasma physical quantity values ​​of multiple consecutive sampling points in the orbital data are lower than the background lower boundary or higher than the background upper boundary, they can be identified as candidate anomalous disturbances. For the identified candidate anomalous disturbances, this embodiment of the invention records their start time, end time, center position and other basic information.

[0039] Sub-step 1023: Filter the candidate anomalous perturbations to obtain the effective anomalous perturbations, and store the effective anomalous perturbations as a plasma observation anomalous dataset.

[0040] Among them, effective anomalies refer to plasma anomaly data segments that have been identified as potentially related to earthquakes after multiple screenings.

[0041] In some embodiments of the present invention, sub-step 1023, "screening candidate abnormal perturbations to obtain effective abnormal perturbations," may specifically include the following sub-steps: Sub-step 10231: If the center of the candidate abnormal disturbance is located within the preset seismic zone and the duration of the candidate abnormal disturbance is within the preset time threshold, then the candidate abnormal disturbance is determined as a valid abnormal disturbance. Sub-step 10232: Calculate the disturbance amplitude of the effective abnormal disturbance and determine whether the disturbance type of the effective abnormal disturbance is an increasing anomaly or a decreasing anomaly.

[0042] In a specific implementation of this invention, firstly, it is determined whether the center of the candidate anomaly is located within a preset seismic zone, which refers to the area covered by extending 1200 km outward from the epicenter of a global earthquake; secondly, it is determined whether the duration of the candidate anomaly is within a preset time threshold, for example, a preset time threshold of 15 seconds to 2 minutes; then, the disturbance amplitude of the candidate anomaly is calculated. If the current value is greater than the background upper bound, then the disturbance amplitude = (current value - background upper bound Ne) / (current value - background upper bound Ne). upIf the current value is less than the background upper bound, then the perturbation amplitude = (background upper bound - current value) / background upper bound × 100%, where the background upper bound is Ne in sub-step 1021. up In the context of the boundary being Ne in sub-step 1021 down Furthermore, this application also determines the perturbation type of valid anomalous perturbations. If the peak plasma observation value in the anomalous data segment is greater than the background upper limit, it is determined to be an increasing anomaly; if the peak plasma observation value is less than the background lower limit, it is determined to be a decreasing anomaly.

[0043] Taking a certain observation by the Zhang Heng-1 satellite as an example, assuming that within a certain latitudinal interval, the upper bound of the plasma background constructed based on the data from the first 5 historical revisit orbits is... The corresponding current value, with the background boundary as The corresponding current value. If the current value increases for 45 consecutive seconds in the current orbit data, and the peak current value reaches... If the corresponding current level exceeds the upper background limit, the abnormal disturbance is identified as an increasing anomaly; if the current value decreases for 30 consecutive seconds in the current orbit data, and the peak current value drops to... If the corresponding current level is lower than the background threshold, then the abnormal disturbance is identified as a reduced anomaly.

[0044] In some embodiments of the present invention, a multi-level confidence mechanism is employed for screening candidate anomalies. Specifically, for data segments exceeding the upper and lower limits of the background, a confidence score is calculated based on the magnitude and duration of the exceedance. For example, data segments exceeding the background value by more than 30% and lasting for more than 60 seconds are assigned a high-confidence label; data segments exceeding the background value by 15%-30% and lasting for 30-60 seconds are assigned a medium-confidence label; and data segments exceeding the background value by 5%-15% and lasting for 15-30 seconds are assigned a low-confidence label. In subsequent earthquake correlation analysis, different weights are assigned to anomalies with different confidence levels. High-confidence anomalies have a higher influence in model training, thereby improving the accuracy of the prediction model.

[0045] After the above screening steps, the latitude and longitude of the center point of the effective anomaly, the duration, the intensity and amplitude of the anomaly change, and the earthquake data corresponding to the effective anomaly are stored to obtain the plasma observation anomaly dataset, as shown in Table 1, which is an effective anomaly information table in the plasma observation anomaly dataset.

[0046] Table 1. Valid Abnormal Disturbance Information Table

[0047] In some embodiments of the present invention, to improve the reliability of anomalous disturbances, a multi-source data fusion verification mechanism is employed to cross-verify valid anomalous disturbances. Specifically, the identified valid anomalous disturbances are compared and verified with physical quantities detected by other payloads on the electromagnetic satellite or plasma observation data from other satellites in orbit during the same period. For example, when the Zhangheng-1 satellite detects a plasma anomalous disturbance in a certain region, observation data from other payloads on the same satellite in the same time period and region are retrieved simultaneously, such as electromagnetic fields and high-energy particles, or plasma observation data from different satellites in orbit during the same period, such as plasma observation data from the Swarm satellite. If at least two or more independent data sources simultaneously observe similar anomalies in the region, the confidence level of the plasma anomalous disturbance is increased by one level; if only a single data source observes the anomaly, the original confidence level is maintained or appropriately downgraded. This multi-source data fusion verification mechanism can effectively eliminate instrument errors or local interference that may exist from a single observation source, significantly improving the reliability of anomalous disturbances and providing a more reliable data foundation for subsequent seismic correlation analysis.

[0048] Step 103: Based on the mainshock dataset and the plasma observation anomaly dataset, establish the spatiotemporal correlation characteristics between earthquake events and anomalous disturbance events.

[0049] Among them, spatiotemporal correlation features refer to the set of features that describe the temporal and spatial correspondence between the main shock event and the anomalous disturbance event.

[0050] In a specific implementation of this invention, the magnitude and occurrence time of the mainshock event are first read from the mainshock dataset, and information such as the center location, occurrence time, and disturbance amplitude of the anomalous disturbance event are read from the plasma observation anomaly dataset. For each mainshock event, the spatial influence radius is calculated based on the correspondence between its magnitude and spatial influence radius, and the spatial influence range centered on the epicenter is determined. At the same time, a preset time window is delineated based on the occurrence time of the mainshock event.

[0051] By traversing all anomalous disturbance events in the plasma observation anomaly dataset, anomalous disturbance events whose center location is within the spatial influence range and whose occurrence time is within a preset time window are identified as anomalous disturbance events that have a spatiotemporal correlation with the main shock event.

[0052] After completing the correlation matching of all mainshock events and anomalous disturbance events, the number of anomalous disturbance events associated with each mainshock event and the number of mainshock events associated with each anomalous disturbance event are counted to obtain the correlation distribution between anomalous disturbance events and mainshock events, forming a spatiotemporal correlation feature. This spatiotemporal correlation feature will serve as the basic input for training the subsequent multi-dimensional earthquake monitoring and prediction model.

[0053] In some embodiments of the present invention, step 103, "establishing the spatiotemporal correlation characteristics between seismic events and anomalous disturbance events," specifically includes the following sub-steps: Sub-step 1031: Determine the spatial influence range of each mainshock event based on the correspondence between earthquake magnitude and spatial influence radius, and define a preset time window based on the occurrence time of each mainshock event; Sub-step 1032: Identify effective anomalous disturbances located within the spatial influence range and occurring within a preset time window as anomalous disturbance events that have a spatiotemporal correlation with the main shock event; Sub-step 1033: Count the number of anomalous disturbance events associated with each mainshock event, and the number of mainshock events associated with each anomalous disturbance event, to obtain the correlation distribution between anomalous disturbance events and mainshock events, forming spatiotemporal correlation characteristics.

[0054] The correspondence between earthquake magnitude and spatial influence radius refers to the quantitative relationship between earthquake magnitude and the spatial influence range of ionospheric anomalies. In this embodiment, the correspondence between earthquake magnitude and spatial influence radius is specifically a logarithmic linear relationship, where the logarithm of the spatial influence radius is directly proportional to the earthquake magnitude. The spatial influence range refers to a circular area centered on the epicenter of the mainshock event and with the spatial influence radius as its radius. The preset time window refers to the time range within which anomalous disturbances may be observed, defined based on the occurrence time of the mainshock event. Valid anomalous disturbances refer to anomalous disturbance events identified from electromagnetic satellite plasma observation data and stored in the plasma observation anomalous dataset in the aforementioned sub-step 10231. Anomalous disturbance events that simultaneously meet the spatial range and time window conditions can be determined to have a spatiotemporal correlation with the mainshock event. The correlation distribution refers to the statistical distribution characteristics of the correspondence between the mainshock event and anomalous disturbance events, including the distribution of the number of anomalous disturbance events associated with each mainshock event and the distribution of the number of mainshock events associated with each anomalous disturbance event.

[0055] In some embodiments of the present invention, the spatial influence radius can be determined by calculating the magnitude of the earthquake. The specific calculation formula is as follows: , where M represents the earthquake magnitude and D is the calculated spatial influence radius.

[0056] Taking the observation data from the Zhangheng-1 satellite as an example, for an earthquake event with a magnitude of M=6.0, according to... Calculation yields the spatial influence radius. Kilometers; for an earthquake event with magnitude M=7.0, the spatial radius of influence. km.

[0057] In some embodiments of the present invention, different preset time windows of varying lengths are used for correlation matching of anomalous disturbances for mainshock events in different magnitude ranges. For example, for earthquake events with magnitudes of 5.0 ≤ M < 6.0, a time window of 15 days prior to and 5 days following is used; for earthquake events with magnitudes of 6.0 ≤ M < 7.0, a time window of 20 days prior to and 5 days following is used; and for earthquake events with magnitudes ≥ 7.0, a time window of 30 days prior to and 5 days following is used. This magnitude-adaptive preset time window can better match the ionospheric anomaly response time characteristics of earthquakes of different intensities.

[0058] In some embodiments of the present invention, a confidence-weighted strategy is used to match the association between anomalous disturbance events and the mainshock event. Specifically, for anomalous disturbance events that have been assigned a confidence label in sub-step 1023, different weights are assigned to anomalous disturbance events with different confidence levels when establishing spatiotemporal correlation features. For example, anomalous disturbance events with high confidence have a weight of 1.0, anomalous disturbance events with medium confidence have a weight of 0.6, and anomalous disturbance events with low confidence have a weight of 0.2. The correlation distribution obtained through weighted statistics can more accurately reflect the true correlation strength between anomalous disturbances and earthquakes, reducing the interference of false anomalies on the correlation features.

[0059] Assuming that after the correlation matching in sub-steps 1031-1032, 5 mainshock events and 6 anomalous disturbance events are obtained, namely: mainshock event EQ-001 (magnitude 6.3, occurring on May 10, 2019) is associated with anomalous disturbance events ANOM-002 and ANOM-005; mainshock event EQ-002 (magnitude 5.7, occurring on July 22, 2019) is associated with anomalous disturbance events ANOM-001, ANOM-003, and ANOM-006; mainshock event EQ... ANOM-003 (magnitude 7.1, occurring on September 15, 2019) is associated with anomalous disturbance events ANOM-002, ANOM-004, ANOM-005, and ANOM-006; mainshock event EQ-004 (magnitude 5.2, occurring on November 3, 2019) is associated with anomalous disturbance event ANOM-001; mainshock event EQ-005 (magnitude 6.8, occurring on January 20, 2020) is associated with anomalous disturbance events ANOM-003 and ANOM-005.

[0060] From the perspective of anomalous disturbance events, the following conclusions can be drawn: Anomalous disturbance event ANOM-001 (occurring on July 10, 2019, centered at 35.2°N, 118.5°E) is associated with mainshock events EQ-002 and EQ-004; anomalous disturbance event ANOM-002 (occurring on April 25, 2019, centered at 34.8°N, 119.2°E) is associated with mainshock events EQ-001 and EQ-003; anomalous disturbance event ANOM-003 (occurring on July 5, 2019, centered at 36.1°N, 117.9°E) is associated with mainshock event E Q-002 and EQ-005; anomalous disturbance event ANOM-004 (occurred on August 30, 2019, centered at 35.5°N, 118.8°E) is associated with mainshock event EQ-003; anomalous disturbance event ANOM-005 (occurred on May 1, 2019, centered at 35.0°N, 118.3°E) is associated with mainshock events EQ-001, EQ-003, and EQ-005; anomalous disturbance event ANOM-006 (occurred on September 1, 2019, centered at 34.5°N, 119.5°E) is associated with mainshock events EQ-002 and EQ-003.

[0061] Based on the above correlations, the number of anomalous disturbance events associated with each mainshock event is as follows: EQ-001 is associated with 2, EQ-002 with 3, EQ-003 with 4, EQ-004 with 1, and EQ-005 with 2. The number of mainshock events associated with each anomalous disturbance event is as follows: ANOM-001 with 2, ANOM-002 with 2, ANOM-003 with 2, ANOM-004 with 1, ANOM-005 with 3, and ANOM-006 with 2.

[0062] Based on the above statistics, the correlation distribution between anomalous disturbance events and the mainshock event is obtained, forming a spatiotemporal correlation feature. This spatiotemporal correlation feature will serve as the basic input for training subsequent multi-dimensional earthquake monitoring and prediction models.

[0063] In some embodiments of this invention, the spatiotemporal correlation features were further analyzed for physical interpretability, and a physical coupling mechanism model between anomalous disturbances and earthquakes was established. Specifically, the physical mechanism correspondence between different types of anomalous disturbances and earthquake gestation was analyzed by combining the coupling theory of the lithosphere, atmosphere, and ionosphere. For example, for increasing anomalies, they may correspond to the accumulation of crustal stress in the earthquake gestation zone leading to enhanced radioactive gas and increased aerosols, which in turn causes an increase in ionospheric electron density; for decreasing anomalies, they may correspond to electromagnetic radiation anomalies in the precursory stage of earthquakes, causing depletion of ionospheric electron density. By establishing a correspondence table between anomaly types and physical mechanisms, the correlation features are combined with physical mechanisms, improving the scientific validity and interpretability of the model. Taking a 7.1 magnitude earthquake in 2019 as an example, an increasing anomaly observed 250 kilometers southeast of the epicenter 15 days before the earthquake coincided with the theoretically calculated lithosphere discharge area and corresponded temporally to the stage of accelerated release of crustal stress, verifying the scientific validity of the physical mechanism of the anomalous disturbance.

[0064] Step 104: Based on the spatiotemporal correlation characteristics, train a multi-dimensional earthquake monitoring and prediction model based on plasma observation data.

[0065] Among them, the multi-dimensional earthquake monitoring and prediction model refers to a comprehensive model that uses plasma observation data to fuse temporal distribution characteristics, spatial distribution characteristics, and disturbance amplitude characteristics to predict the time, location, and magnitude of potentially hazardous earthquakes.

[0066] In a specific implementation of this invention, firstly, the spatiotemporal correlation characteristics between the anomalous disturbance event and the mainshock event are obtained from step 103, including the anomalous disturbance events associated with each mainshock event and their occurrence time, center location, and disturbance amplitude. Then, for different magnitude ranges, three-dimensional sub-models are constructed: a temporal distribution model to learn the pattern of earthquake probability change over time after the occurrence of an anomalous disturbance; a spatial distribution model to learn the spatial correspondence between the location of the anomalous disturbance and the epicenter location; and a disturbance amplitude correlation model to learn the quantitative relationship between the intensity of the anomalous disturbance and the earthquake magnitude. Finally, these three sub-models are jointly trained, enabling the model to comprehensively predict the occurrence time, orientation, and magnitude of potentially hazardous earthquakes from the three dimensions of the occurrence time, center location, and disturbance amplitude, resulting in a trained multi-dimensional earthquake monitoring and prediction model.

[0067] In some embodiments of the present invention, a clustering analysis algorithm is introduced to dynamically divide the optimal magnitude intervals based on the distribution characteristics of historical earthquake data in a specific region. For example, for the Mediterranean-Himalayan seismic belt, historical earthquake statistics show that M5.8-6.4 and M6.5-7.3 are two obvious earthquake clusters. Based on the clustering analysis results, the magnitude intervals are divided into four intervals: M < 5.8, 5.8 ≤ M < 6.5, 6.5 ≤ M < 7.3, and M ≥ 7.3, making the earthquake samples in each interval more evenly distributed and improving the model training effect.

[0068] In some embodiments of the present invention, step 104, "training a multi-dimensional earthquake monitoring and prediction model based on plasma observation data," specifically includes the following sub-steps: Sub-step 1041: For different magnitude ranges, construct a time distribution model, a spatial distribution model, and a disturbance amplitude correlation model; the time distribution model represents the temporal correlation between anomalous disturbance events and the main shock event, the spatial distribution model represents the spatial correlation between anomalous disturbance events and the main shock event, and the disturbance amplitude correlation model represents the correlation between disturbance amplitude and magnitude; Sub-step 1042: Jointly train the time distribution model, spatial distribution model, and disturbance amplitude correlation model to obtain a multi-dimensional earthquake monitoring and prediction model.

[0069] Joint training refers to optimizing the parameters of the three sub-models—the time distribution model, the spatial distribution model, and the disturbance amplitude correlation model—as a whole, enabling the model to comprehensively output earthquake prediction information from three dimensions: the occurrence time, the center location, and the disturbance amplitude of the anomalous disturbance. The multi-dimensional earthquake monitoring and prediction model refers to the comprehensive model that can be used for actual prediction after training. Its inputs are the occurrence time, the center location, and the disturbance amplitude of the anomalous disturbance, and its outputs are the predicted occurrence time, the azimuth and distance, and the predicted magnitude range.

[0070] In some embodiments of the present invention, the "constructing a time distribution model" in sub-step 1041 may specifically include the following sub-steps: Sub-step 10411: Using the occurrence time of the abnormal disturbance event as the baseline zero point, define the preset time range before and after the occurrence of the abnormal disturbance event, divide the preset time range into multiple continuous time units according to the preset time interval, count the cumulative number of main shock events in different magnitude intervals within each time unit, and fit the distribution function of the probability of the occurrence of the main shock event after the occurrence of the abnormal disturbance event as a function of time.

[0071] In a specific implementation of this invention, the time distribution model is constructed using a statistical fitting method. Specifically, taking the occurrence time of the anomalous disturbance event as the baseline zero point, a preset time range of 5 days before and 20 days after the occurrence of the anomalous disturbance event is defined. This range is then divided into 25 consecutive time units with a 1-day time interval. For each magnitude interval, the cumulative number of mainshock events occurring within each time unit is counted, resulting in a histogram showing the change in the number of mainshock events over time. Curve fitting is then performed on this histogram to obtain the distribution function of the probability of the mainshock event occurring after the occurrence of the anomalous disturbance event over time. Taking the magnitude interval of 5.0 ≤ M < 6.0 as an example, if statistics show that the 3rd to 8th day after the occurrence of the anomalous disturbance is a high-incidence period for earthquakes in this interval, the time distribution model will assign a higher probability of occurrence to this period.

[0072] In some embodiments of the present invention, the "constructing a spatial distribution model" in sub-step 1041 may specifically include the following sub-steps: Sub-step 10412: For each magnitude interval, calculate the azimuth and spatial distance between the earthquake center of the mainshock event and the center of the anomalous disturbance event within the magnitude interval. The center of the anomalous disturbance event and the mainshock event have a spatiotemporal correlation. Sub-step 10413: Determine the theoretical influence radius based on the correspondence between earthquake magnitude and spatial influence radius, use the theoretical influence radius as a constraint condition for spatial attenuation, and construct the probability density distribution function between the center of the abnormal disturbance event and the earthquake center.

[0073] In the specific implementation of this invention, the spatial distribution model is constructed using a probability density estimation method. Specifically, for each magnitude interval, all anomalous disturbance events and mainshock events with spatiotemporal correlation within that interval are statistically analyzed, and the azimuth and spatial distance from the center of the anomalous disturbance event to the epicenter of the mainshock event are calculated for each event pair. Based on the correspondence between earthquake magnitude and spatial influence radius, the theoretical influence radius is determined. This theoretical influence radius is used as a constraint condition for spatial attenuation, and a two-dimensional probability density distribution function between the center of the anomalous disturbance event and the epicenter of the mainshock event is constructed using a kernel density estimation method. Taking the 6.0 ≤ M < 7.0 magnitude interval as an example, if statistics show a high probability that the center of the anomalous disturbance event is located within 200-300 kilometers southeast of the epicenter, the spatial distribution model will assign a higher probability density to this azimuth and distance range.

[0074] In some embodiments of the present invention, the "constructing a disturbance amplitude correlation model" in sub-step 1041 may specifically include the following sub-steps: Sub-step 10414: For each magnitude interval, statistically analyze the distribution of the effective anomalous disturbance amplitude corresponding to the mainshock event within the magnitude interval; Sub-step 10415: Establish a regression model between the disturbance amplitude and the earthquake magnitude. The regression model uses the law represented by the correspondence between earthquake magnitude and spatial influence radius as the constraint condition.

[0075] In the specific implementation of this invention, the disturbance amplitude correlation model is constructed using regression analysis. Specifically, for each magnitude interval, the disturbance amplitudes of all anomalous disturbance events with spatiotemporal correlations within that interval are statistically analyzed, and a regression model between the disturbance amplitude and the earthquake magnitude is established. This regression model uses the physical laws characterized by the correspondence between earthquake magnitude and spatial influence radius as constraints to ensure that the model conforms to the basic physical principles of earthquake ionospheric coupling. Taking the M≥7.0 magnitude interval as an example, if statistics show that anomalous disturbances with disturbance amplitudes between 20% and 30% mostly correspond to earthquakes of magnitude 7.0-7.5, and anomalous disturbances with disturbance amplitudes between 30% and 40% mostly correspond to earthquakes of magnitude 7.5-8.0, then the regression model will establish a quantitative mapping relationship between the disturbance amplitude and the magnitude.

[0076] In some embodiments of the present invention, joint training employs a multi-task learning framework. Specifically, the temporal distribution model, spatial distribution model, and disturbance amplitude correlation model are treated as three correlation tasks, sharing the underlying spatiotemporal correlation feature representation. During training, the loss functions of the three tasks are optimized simultaneously, enabling the model to learn the intrinsic correlation between the three dimensions. For example, for the same anomalous disturbance event, its occurrence time, center location, and disturbance amplitude jointly determine the occurrence time, orientation, and magnitude of the associated earthquake. Through multi-task joint training, the model can capture the synergistic change patterns among these three dimensions, improving overall prediction performance.

[0077] Step 105: Obtain real-time plasma observation data from electromagnetic satellites, input the real-time plasma observation data into the trained multi-dimensional earthquake monitoring and prediction model, and obtain potential hazardous earthquake prediction information.

[0078] Among them, real-time plasma observation data refers to orbital observation data containing physical quantities such as electron density and electron temperature, which are collected in real time or recently acquired by electromagnetic satellites; potential hazardous earthquake prediction information refers to the result information obtained after predicting the occurrence time, location, and magnitude of potential hazardous earthquakes.

[0079] In a specific implementation of this invention, real-time plasma observation data from an electromagnetic satellite is first acquired. Following the same method as in step 102, the real-time plasma observation data is compared with a pre-constructed plasma background value to identify any anomalous disturbances in the current data. If an anomalous disturbance is identified, its occurrence time, center location, and amplitude are extracted. These three characteristic parameters are then input into the multi-dimensional earthquake monitoring and prediction model trained in step 104. The temporal distribution model outputs the predicted occurrence time based on the anomalous occurrence time, the spatial distribution model outputs the predicted occurrence azimuth and distance based on the anomalous center location, and the disturbance amplitude correlation model outputs the predicted magnitude range based on the disturbance amplitude. Finally, the prediction results from the three dimensions are integrated into earthquake prediction information for monitoring and early warning of potentially hazardous earthquakes.

[0080] In some embodiments of the present invention, step 105, "inputting real-time plasma observation data into the trained multi-dimensional earthquake monitoring and prediction model to obtain potential hazardous earthquake prediction information," specifically includes the following sub-steps: Sub-step 1051: Compare the real-time plasma observation data with the plasma background value, identify abnormal disturbances in the real-time observation data, and extract the occurrence time, center location, and disturbance amplitude of the abnormal disturbances in the real-time observation data; Sub-step 1052: Input the occurrence time, center location, and amplitude of the real-time observation data anomalies into the multi-dimensional earthquake monitoring and prediction model; Sub-step 1053: Obtain the predicted occurrence time output by the time distribution model, the predicted occurrence location and distance output by the spatial distribution model, and the predicted magnitude range output by the disturbance amplitude correlation model.

[0081] Among them, the predicted occurrence time refers to the time interval inferred by the time distribution model based on the occurrence time of the current anomalous disturbance; the predicted occurrence azimuth and distance refer to the range of azimuth and distance ranges of the epicenter of the potential hazardous earthquake inferred by the spatial distribution model based on the center location of the current anomalous disturbance; and the predicted magnitude range refers to the range of the magnitudes of the potential hazardous earthquake inferred by the disturbance amplitude correlation model based on the disturbance amplitude of the current anomalous disturbance.

[0082] In some embodiments of the present invention, the identification of the current anomalous disturbance adopts the same screening criteria as in step 102. Taking the observation data of Zhang Heng-1 satellite as an example, assuming that a certain current orbital data is obtained, it is compared with the pre-constructed plasma background value. It is found that there is a continuous 45-second increase in electron density near 35.2 degrees north latitude and 118.5 degrees east longitude, exceeding the upper limit of the background by 25%. It is determined that the center of this anomaly is located within the preset seismic zone and the duration is between 15 seconds and 2 minutes, so it is identified as the current anomalous disturbance. The occurrence time of this anomaly is extracted as 14:25 on March 17, 2026, the center location is 35.2 degrees north latitude and 118.5 degrees east longitude, and the disturbance amplitude is an increasing anomaly of 25%.

[0083] In some embodiments of the present invention, the extracted anomaly occurrence time is input into a time distribution model. Based on the trained distribution function, the model outputs a predicted occurrence time of 3 to 8 days after the anomaly occurs. The anomaly center location is input into a spatial distribution model. Based on the trained probability density distribution function, the model outputs a predicted occurrence direction of southeast, at a distance of 200-300 kilometers, meaning the potential epicenter is located approximately 200-300 kilometers southeast of the anomaly center. A 25% disturbance amplitude is input into a disturbance amplitude correlation model. Based on the trained regression relationship, the model outputs a predicted magnitude range of 6.0 ≤ M < 7.0. Combining the prediction results from these three dimensions, the earthquake prediction information is obtained: an earthquake of magnitude 6.0-6.9 is expected to occur within the next 3-8 days within a range of 200-300 kilometers southeast of the anomaly center.

[0084] In some embodiments of the present invention, the output of earthquake prediction information adopts a probabilistic expression. Specifically, the temporal distribution model not only outputs the most likely time interval of occurrence but also the probability of occurrence for each time unit; the spatial distribution model not only outputs the most likely location and distance but also outputs a spatial probability distribution heatmap; the disturbance amplitude correlation model not only outputs the most likely magnitude range but also outputs the confidence level for each magnitude interval. For example, for the above-mentioned anomalous disturbance, the output of the temporal distribution model can be: 20% probability on day 3, 35% probability on day 4, 30% probability on day 5, 10% probability on day 6, and 5% probability on day 7; the spatial distribution model can output a probability ellipse centered on the predicted point; the disturbance amplitude correlation model can output: 60% probability of magnitude 6.0-6.5, 30% probability of magnitude 6.5-7.0, and 10% probability of magnitude 7.0 and above. The probabilistic expression adopted in the embodiments of the present invention can more comprehensively reflect the uncertainty of prediction and provide richer decision-making information for earthquake monitoring and early warning.

[0085] In some embodiments of the present invention, if multiple current anomalous disturbances are identified within the same time period, each anomalous disturbance is input into the model to obtain its own prediction result, and then spatiotemporal clustering analysis is performed on all prediction results. For example, if the predicted occurrence time and spatial region of multiple anomalous disturbances overlap, the prediction confidence of the overlapping region is enhanced; if the prediction results are scattered and non-overlapping, they are monitored separately as multiple independent seismic hazard zones. This multi-anomaly comprehensive judgment strategy can more comprehensively capture earthquake precursor information and improve the reliability of prediction.

[0086] In summary, this invention establishes the spatiotemporal correlation characteristics between earthquake events and anomalous disturbance events by constructing a standardized mainshock dataset and a plasma observation anomaly dataset, and trains a multi-dimensional earthquake monitoring and prediction model. This enables accurate prediction of the time, location, and magnitude of potentially hazardous earthquakes based on electromagnetic satellite plasma observation data, significantly improving the practicality and accuracy of earthquake monitoring and prediction.

[0087] Reference Figure 2 The diagram shows the structure of an earthquake monitoring and prediction system based on anomalies in electromagnetic satellite plasma observation data. The system includes: The mainshock dataset construction module 201 is used to acquire historical earthquake data and construct a mainshock dataset based on the historical earthquake data. Anomaly dataset construction module 202 is used to acquire plasma observation data accumulated by electromagnetic satellites and construct anomaly datasets based on the accumulated plasma observation data. The association construction module 203 is used to establish the spatiotemporal correlation features between earthquake events and anomalous disturbance events based on the mainshock dataset and the plasma observation anomaly dataset. Training module 204 is used to train a multi-dimensional earthquake monitoring and prediction model based on plasma observation data based on the spatiotemporal correlation features. The prediction module 205 is used to acquire real-time plasma observation data from the electromagnetic satellite, input the real-time plasma observation data into the trained multi-dimensional earthquake monitoring and prediction model, and obtain potential hazardous earthquake prediction information.

[0088] As the system implementation is basically similar to the method implementation, it is described in a relatively simple way. For relevant details, please refer to the description of the method implementation.

[0089] This invention also provides an electronic device that may include a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements the method described above.

[0090] This invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it implements the method described above.

[0091] The specific embodiments described above are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A method for earthquake monitoring and prediction based on anomalies in electromagnetic satellite plasma observation data, characterized in that, The method includes the following steps: Acquire historical earthquake data and construct a mainshock dataset based on the historical earthquake data; Acquire plasma observation data accumulated by electromagnetic satellites, and construct a plasma observation anomaly dataset based on the accumulated plasma observation data; Based on the mainshock dataset and the plasma observation anomaly dataset, the spatiotemporal correlation characteristics between seismic events and anomalous disturbance events are established; Based on the aforementioned spatiotemporal correlation characteristics, a multi-dimensional earthquake monitoring and prediction model based on plasma observation data is trained. Real-time plasma observation data from the electromagnetic satellite is acquired, and the real-time plasma observation data is input into the trained multi-dimensional earthquake monitoring and prediction model to obtain potential hazardous earthquake prediction information.

2. The method according to claim 1, characterized in that, The process of constructing the mainshock dataset based on the historical earthquake data includes: Obtain the earthquake center of each historical earthquake event in the historical earthquake data, and delineate the target spatial region based on the earthquake center and a first spatial range. The earthquake event with the largest magnitude within the first time range in the target spatial region is identified as the mainshock event. The mainshock events are classified and stored according to their magnitude ranges to obtain the mainshock dataset.

3. The method according to claim 1, characterized in that, The construction of the plasma observation anomaly dataset based on the accumulated plasma observation data includes: For each trajectory data to be judged, several revisited trajectory data are selected as background trajectory data. Based on the background trajectory data, a plasma background value is constructed and the upper and lower limits of the background are determined. The trajectory data to be judged is compared with the upper and lower boundaries of the background, and data segments exceeding the upper and lower boundaries of the background are identified as candidate abnormal disturbances. The candidate anomalous perturbations are screened to obtain valid anomalous perturbations, which are then stored as the plasma observation anomalous dataset. The plasma observation anomalous dataset is continuously updated based on the accumulation of observation data.

4. The method according to claim 3, characterized in that, The process of filtering the candidate abnormal perturbations to obtain valid abnormal perturbations specifically includes: If the center of the candidate abnormal disturbance is located within a preset seismic zone and the duration of the candidate abnormal disturbance is within a preset time threshold, then the candidate abnormal disturbance is determined as a valid abnormal disturbance. Calculate the disturbance amplitude of the effective abnormal disturbance and determine whether the disturbance type of the effective abnormal disturbance is an increasing anomaly or a decreasing anomaly.

5. The method according to claim 1, characterized in that, The establishment of spatiotemporal correlation features between seismic events and anomalous disturbance events specifically includes: The spatial impact range of each mainshock event is determined based on the correspondence between earthquake magnitude and spatial influence radius, and a preset time window is defined based on the occurrence time of each mainshock event. Effective abnormal disturbances located within the spatial influence range and occurring within the preset time window are identified as abnormal disturbance events that have a spatiotemporal correlation with the main shock event. The number of anomalous disturbance events associated with each mainshock event and the number of mainshock events associated with each anomalous disturbance event are counted to obtain the correlation distribution between anomalous disturbance events and mainshock events, forming the spatiotemporal correlation characteristics.

6. The method according to claim 1, characterized in that, The training of the multi-dimensional earthquake monitoring and prediction model based on plasma observation data specifically includes: For different magnitude ranges, a temporal distribution model, a spatial distribution model, and a disturbance amplitude correlation model are constructed. The temporal distribution model represents the temporal correlation between anomalous disturbance events and the main shock event, the spatial distribution model represents the spatial correlation between anomalous disturbance events and the main shock event, and the disturbance amplitude correlation model represents the correlation between disturbance amplitude and magnitude. The multi-dimensional earthquake monitoring and prediction model is obtained by jointly training the time distribution model, the spatial distribution model, and the disturbance amplitude correlation model.

7. The method according to claim 6, characterized in that, The construction of the time distribution model includes: Using the occurrence time of the abnormal disturbance event as the baseline zero point, a preset time range before and after the occurrence of the abnormal disturbance event is defined. The preset time range is divided into multiple consecutive time units according to a preset time interval. The cumulative number of main shock events in different magnitude intervals within each time unit is counted. The distribution function of the probability of the main shock event occurring after the occurrence of the abnormal disturbance event is obtained by fitting the data. Constructing the spatial distribution model includes: For each magnitude interval, the azimuth and spatial distance of the earthquake center of the main shock event relative to the center of the anomalous disturbance event within the magnitude interval are statistically analyzed. The center of the anomalous disturbance event has a spatiotemporal correlation with the main shock event. The theoretical influence radius is determined based on the correspondence between earthquake magnitude and spatial influence radius. The theoretical influence radius is used as a constraint condition for spatial attenuation to construct the probability density distribution function between the center of the anomalous disturbance event and the earthquake center. Constructing the disturbance amplitude correlation model includes: For each magnitude interval, the amplitude distribution of the effective anomalous disturbances corresponding to the mainshock event within the magnitude interval is statistically analyzed; A regression model is established between the disturbance amplitude and the earthquake magnitude. The regression model uses the law represented by the correspondence between the earthquake magnitude and the radius of influence as a constraint condition.

8. The method according to claim 3, characterized in that, The step of inputting the real-time plasma observation data into the trained multi-dimensional earthquake monitoring and prediction model to obtain earthquake prediction information specifically includes: The real-time plasma observation data is compared with the plasma background value to identify abnormal disturbances in the real-time observation data, and the occurrence time, center location, and disturbance amplitude of the abnormal disturbances in the real-time observation data are extracted. The occurrence time, center location, and amplitude of the abnormal disturbances in the real-time observation data are input into the multi-dimensional earthquake monitoring and prediction model; The predicted occurrence time output by the time distribution model, the predicted occurrence location and distance output by the spatial distribution model, and the predicted magnitude range output by the disturbance amplitude correlation model are obtained.

9. An earthquake monitoring and prediction system based on electromagnetic satellite plasma observation anomalies, characterized in that, For implementing the method according to any one of claims 1-8, the system comprises: The mainshock dataset construction module is used to acquire historical earthquake data and construct the mainshock dataset based on the historical earthquake data. An anomaly dataset construction module is used to acquire plasma observation data accumulated by electromagnetic satellites and construct an anomaly dataset of plasma observations based on the accumulated plasma observation data. The association construction module is used to establish the spatiotemporal correlation characteristics between earthquake events and anomalous disturbance events based on the mainshock dataset and the plasma observation anomaly dataset. The training module is used to train a multi-dimensional earthquake monitoring and prediction model based on plasma observation data based on the spatiotemporal correlation features. The prediction module is used to acquire real-time plasma observation data from the electromagnetic satellite, input the real-time plasma observation data into the trained multi-dimensional earthquake monitoring and prediction model, and obtain potential hazardous earthquake prediction information.

10. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the method as described in any one of claims 1-8.

Images