Large-scale insect and bird migration monitoring system based on weather radar net

Abstract

The invention provides a large-scale insect and bird migratory flight monitoring system based on a weather radar net, and the system employs a data collection module to obtain S, C and X wave band weather radar monitoring data of the national weather radar net, and a quality control module carries out the quality control of the radar-based data in the monitoring data, and further provides the radar-based data to an echo recognition module to construct recognition features. A classification recognition result is obtained by adopting a recognition model corresponding to the wave band; the inversion quantification module inverts the insect and bird density, the insect and bird speed and the vertical distribution of the movement direction based on the classified and identified insect and bird echoes to obtain an inversion quantification result; and the classification identification result and the inversion quantification result of the single wave band form a local monitoring result. And the visualization module displays a single-band local monitoring result or a multi-band fusion monitoring result. According to the invention, multi-band data can be collected, monitoring results are fused, real-time and efficient monitoring of migration of insects and birds is realized, and the monitoring range is expanded.

Description

Technical Field

[0001] This invention belongs to the field of weather radar technology, specifically relating to a large-scale monitoring system for insect and bird migration based on a weather radar network. Background Technology

[0002] Every year, large numbers of birds and insects undertake long-distance migrations between their breeding grounds and wintering grounds. Birds serve as important indicators of ecosystem health, and their migration patterns can reflect changes in the ecological environment. Insects, on the other hand, play a crucial role in freshwater and terrestrial food webs, undertaking vital energy transfer functions from primary producers to advanced consumers, and providing important ecological services to humans.

[0003] In recent years, drastic changes in the global climate and environment have led to a significant decline in bird and insect populations, drawing considerable attention. Bird and insect migration is also closely intertwined with human life. For example, migratory birds may carry pathogens such as avian influenza viruses, spreading them to new areas and threatening public health and poultry farming. Furthermore, bird migration triggers numerous bird strike incidents. With the rapid development of the aviation industry, the frequency of bird strike accidents has been increasing year by year, becoming a major threat to aviation safety. Nearly 10,000 bird strike incidents occur globally each year, which the International Civil Aviation Organization (ICAO) defines as a "Category A aviation disaster," with 80% of these caused by bird collisions.

[0004] Meanwhile, insect migration is also a significant factor in the cross-regional spread of pests and diseases. According to the Food and Agriculture Organization of the United Nations (FAO), global food production losses due to pests and diseases exceed 20% annually, posing a major threat to agricultural security. Therefore, monitoring and studying the migration patterns of insects and birds is crucial for maintaining public health, aviation safety, and agricultural production.

[0005] Currently, dual-polarization weather radar has made significant progress in insect and bird identification. Based on dual-polarization parameters such as differential reflectivity, differential phase, and correlation coefficient, the types of biological echoes in radar echoes can be identified relatively accurately. These studies provide important tools for understanding the patterns of biological migration, especially in detecting large-scale bird migration and insect activity, where dual-polarization radar technology demonstrates unique advantages. However, despite the success of these technologies and methods in local insect and bird identification, there is still no truly established continental-scale insect and bird migration monitoring system based on a weather radar network. Existing research and applications are mostly concentrated in certain specific regions or short-term monitoring projects, while achieving continuous and automated migration monitoring across the country or larger remains a challenge. Weather radar networks possess the infrastructure for nationwide coverage and the potential for real-time monitoring, but how to utilize these resources to build a unified monitoring system still requires solving many problems. Summary of the Invention

[0006] In view of this, the present invention provides a large-scale insect and bird migration monitoring system based on a weather radar network. The system collects multi-band data, fuses monitoring results, realizes real-time and efficient insect and bird migration monitoring, and expands the monitoring range.

[0007] A large-scale monitoring system for insect and bird migration based on a weather radar network includes: a data acquisition module, a quality control module, an echo recognition module, an inversion quantization module, a visualization module, and a historical data analysis module; The data acquisition module is used to acquire real-time monitoring data from S-band, C-band, and X-band weather radars of the national weather radar network, including radar base data. The quality control module is used to perform quality control on the acquired radar base data; The echo recognition module is used to construct recognition features based on the radar base data after quality control, and to classify and recognize radar echoes using a band-corresponding recognition model to obtain classification and recognition results. The inversion quantization module is used to invert the vertical distribution of insect and bird density, speed and movement direction based on the insect and bird echoes identified by classification, and obtain the inversion quantization results; The visualization module is used to combine the classification and identification results and inversion quantization results of a single-band single-weather radar into local monitoring results. The visualization interface displays the local monitoring results of a single-band single-weather radar, the comprehensive monitoring results of a single-band multi-weather radar, or the fusion monitoring results of a multi-band multi-weather radar.

[0008] Preferably, the real-time monitoring data acquired by the data acquisition module further includes meteorological data; the visualization interface of the visualization module overlays the corresponding geographic map and meteorological data.

[0009] Preferably, the quality control module includes a ground clutter and superrefractive echo suppression unit and a radio frequency interference echo suppression unit; Ground clutter and super-refractive echo suppression unit, used to remove ground clutter and super-refractive echo by utilizing the time-invariant characteristics of ground clutter and super-refractive echo; The radio frequency interference echo suppression unit is used to remove radio frequency interference echoes by utilizing the elongated shape characteristic of radio frequency interference echoes.

[0010] Preferably, the echo identification module uses radar base data and its statistical data to form a feature space, and applies machine learning models or deep learning models to classify and identify radar echoes.

[0011] Preferably, when performing radar echo classification and recognition, the echo recognition module treats radar resolution units as sample points for S-band and C-band weather radars, and uses the radar base data and statistical data corresponding to the radar resolution units as point features. It then uses XGBoost or random forest classifiers to learn the point features to obtain the classification models corresponding to S-band and C-band weather radars. For X-band weather radar, the point features and surface features are extracted to form a hybrid feature. The hybrid feature is learned using a ResNet50 or VGG16 neural network to obtain the classification model corresponding to the X-band weather radar. The surface feature is the output of the neural network without the softmax layer, which represents the echo spatial texture.

[0012] Preferably, the inversion quantization module obtains the inversion quantization result in the following way: A quantitative relationship between the reflectivity factor, average radar cross section, and insect population of insect echoes was established using a swarm Rayleigh scattering model. A quantitative relationship between the reflectivity factor, average radar cross section, and insect population of bird echoes was established using a Gaussian distribution model based on body size and orientation. Based on the quantitative relationship, combined with the reflectivity factor and average radar cross section corresponding to the insect and bird echoes, the number of insects and birds was estimated. Using radar reflectivity data from multiple elevation angles, the vertical distribution of insect and bird numbers is inverted to obtain the vertical distribution of insect and bird density. Based on radial velocity and velocity orientation display technology, the vertical distribution of movement velocity and vertical distribution of movement direction of insect-bird groups are estimated.

[0013] Preferably, the visualization module is divided into a single-station display interface and a multi-station display interface; The single-station display interface shows the local monitoring results of the weather radar corresponding to each station; The multi-station display interface displays the comprehensive monitoring results of a single-band multi-weather radar obtained by spatial interpolation and spatial stitching of local monitoring results of some or all weather radars of the same band within the global spatial range, or displays the fused monitoring results of multi-band multi-weather radar obtained by spatial interpolation, spatial stitching and band fusion of local monitoring results of all weather radars of the global spatial range. The band fusion is as follows: when the coverage areas of S-band, C-band, and X-band weather radars overlap, the echo classification result and inversion quantization result of one band weather radar are selected as the data for the overlapping area, according to the priority of S-band > X-band > C-band.

[0014] Preferably, the system further includes a historical data analysis module for storing the inversion quantization results and analyzing them according to the time series.

[0015] Preferably, the historical data analysis module further establishes a long-term dynamic model of insect and bird migration by analyzing the migration time, path, and number changes of insects and birds.

[0016] Beneficial effects: (1) Large-scale coverage and real-time monitoring: This system relies on a widely deployed weather radar network to construct a multi-band weather radar collaborative sensing system. By integrating the advantages of multi-band monitoring, it achieves all-weather, high-precision, and real-time dynamic monitoring of insect and bird migration activities across the country. This invention breaks through the limitations of existing technologies that only monitor single frequency bands and local areas, extending the monitoring range from local scales to continental macroscales to meet my country's macroscopic monitoring needs for biological migration.

[0017] (2) Automation and Efficiency: This system constructs a fully automated processing system from radar data acquisition and intelligent echo identification to precise parameter quantification, and innovatively integrates the entire process seamlessly into an intelligent software platform, realizing real-time visualization of monitoring results and autonomous operation with zero human intervention. Compared with traditional manual analysis or partially automated systems, this invention solves the problem of the isolation between the three major modules of data processing, echo identification, and parameter inversion in the traditional process, adds real-time performance and integration, greatly improves monitoring efficiency, and can meet the needs of large-scale and long-term monitoring.

[0018] (3) Wide range of applications: This system can be used not only for ecological environment monitoring, but also for agricultural pest early warning, aviation safety, meteorological research and other fields. It has a wide range of application prospects and fills the gaps in existing technologies in multiple fields.

[0019] In summary, this invention significantly improves the accuracy, efficiency, and automation of large-scale insect and bird migration monitoring, and solves the shortcomings of existing technologies in large-scale, long-term monitoring. Attached Figure Description

[0020] Figure 1 This is a framework diagram of a large-scale insect and bird migration monitoring system.

[0021] Figure 2 This is the classification result of the 2.4° elevation angle measurement data from the dual-polarization weather radar at 00:22 on September 22, 2024. The echoes mainly originated from birds.

[0022] Figure 3 This is the classification result of the 2.4° elevation angle measurement data from the dual-polarization weather radar at 14:22 on September 22, 2024. The echoes mainly originated from insects.

[0023] Figure 4The data is derived from the bird density, speed, and vertical distribution of direction measured by the dual-polarization weather radar at 00:22 on September 22, 2024.

[0024] Figure 5 The data is derived from the vertical distribution of insect density, speed, size, and direction measured by the dual-polarization weather radar at 14:22 on September 22, 2024.

[0025] Figure 6 This is a time series of cumulative bird density measured by the dual-polarization weather radar at Station A on September 22, 2024.

[0026] Figure 7 This is a time series of cumulative insect density measured by the dual-polarization weather radar at Station A on September 22, 2024.

[0027] Figure 8 This is a time series of daily cumulative bird density measurements taken by Bilibili from March to October 2021.

[0028] Figure 9 This is a time series of daily cumulative insect density measurements taken by Bilibili from March to October 2021. Detailed Implementation

[0029] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0030] The overall architecture of the large-scale insect and bird migration monitoring system based on weather radar network provided in this embodiment of the invention is as follows: Figure 1 As shown, the system is divided into six modules: data acquisition, quality control, echo identification, inversion quantization, visualization, and historical data analysis. The following description of the system's workflow explains the function of each module: Step 1: Data Acquisition Module. This module acquires real-time monitoring data from S-band, C-band, and X-band weather radars across the national weather radar network, including radar baseline data and meteorological data. The radar baseline data includes six types of radar polarization parameters: reflectivity factor, radial velocity, differential reflectivity, differential phase, correlation coefficient, and velocity spectral width. Meteorological data includes temperature, humidity, and wind field.

[0031] Step 2, the quality control module, performs quality control on the weather radar base data collected in Step 1. Based on different types of clutter characteristics, corresponding methods are used to remove the corresponding clutter, including ground clutter suppression, superrefractive echo suppression, and radio frequency interference echo suppression. Ground clutter and superrefractive echoes can be removed based on their time-invariant characteristics, while radio frequency interference echoes can be removed based on their elongated shape characteristics. Specifically, ground clutter suppression can obtain a mask by statistically analyzing echo images from clear winter skies; subtracting the mask during use achieves the effect of removing ground clutter. Radio frequency interference echo suppression can use CFAR target detection technology to detect elongated shapes, thereby removing radio frequency interference.

[0032] Step 3: Echo recognition module. Based on the weather radar base data that has undergone quality control in Step 2, it constructs recognition features, inputs them into the recognition model, and realizes the classification and recognition of radar echoes.

[0033] In a preferred embodiment, radar base data is selected, and statistical data from the radar base data are obtained as derived features to jointly constitute the feature space; wherein the derived features refer to mean, standard deviation, variance, etc. Machine learning models or deep learning models (such as convolutional neural networks or XGBoost, etc.) are applied to classify and identify radar echoes into three categories: birds, insects, and meteorological data.

[0034] Considering that the national weather radar network includes radars of different bands, echo recognition algorithms need to be designed according to the characteristics of each band. For S-band radar, radar resolution units are treated as sample points. The radar base data corresponding to each resolution unit and its derived features constitute point features. XGBoost or random forest classifiers are used to learn these point features, resulting in a classification model for S-band radar. C-band radar is similar to S-band, using XGBoost or random forest classifiers for learning and classification. X-band radar has more complex insect and bird scattering characteristics; therefore, based on point features, neural networks are used to extract echo spatial texture as surface features. Point and surface features are combined to form a hybrid feature, which is learned using ResNet50 or VGG16 neural networks, resulting in a classification model for X-band radar. Here, echo spatial texture refers to the output of the neural network without the softmax layer.

[0035] Step 4: Inversion Quantization Module. Based on the bird and insect echoes output by the echo recognition module in Step 3, the vertical distribution of bird and insect numbers, movement speed, and movement direction is inverted to obtain the inversion quantization results.

[0036] The specific implementation steps of this process include: based on the bird and insect echoes output by the echo identification module in step three, estimating the number of insects and birds based on their corresponding reflectivity factors and average radar cross-sections (RCS). Specifically, for insects, a population Rayleigh scattering model is used to establish a quantitative relationship between the reflectivity factor, RCS, and insect number; for birds, a Gaussian distribution model based on body size and orientation is used to establish a quantitative relationship between the reflectivity factor, RCS, and insect number. Then, using radar reflectivity data at multiple elevation angles, the vertical distribution of insect and bird numbers, i.e., the vertical distribution of insect and bird density, is retrieved; based on radial velocity data and velocity-azimuth display technology, the vertical distribution of the movement speed and direction of the insect and bird population is estimated. The vertical distribution of insect and bird numbers, movement speed, and movement direction obtained by this quantification module is recorded as insect and bird monitoring data.

[0037] Step 5: Visualization Module. The radar echo classification results from the echo identification module in Step 3 and the inversion quantization results from the inversion quantization module in Step 4, including the number of insects and birds, their vertical distributions, and their motion characteristics, are output through a visualization interface.

[0038] If the classification and identification results and inversion quantization results of a single-band single-weather radar are combined to form a local monitoring result, then this visualization module can display the local monitoring result of a single-band single-weather radar, or the comprehensive monitoring result of a single-band multi-weather radar, or the fusion monitoring result of a multi-band multi-weather radar through a visualization interface.

[0039] Generally, weather radars of different bands cover different spatial areas, and there can be multiple stations for the same band, with each station corresponding to one weather radar. Therefore, the integrated monitoring results of multiple weather radars in a single band only involve spatial interpolation and spatial stitching. The local monitoring results of multiple radar stations in the same band are stitched together to form a single-band global mosaic.

[0040] The fusion of monitoring results from multiple bands and multiple weather radars includes not only fusion between different spatial regions but also fusion of band data in overlapping spatial regions. Local monitoring results from different spatial regions can be fused using spatial interpolation and spatial stitching. When the coverage areas of S-band, C-band, and X-band weather radars overlap, the local monitoring results of one band weather radar are selected as the data for the overlapping region, following the priority order: S-band > X-band > C-band.

[0041] Different weather radar bands have varying data quality and scattering characteristics, resulting in different ranges of radar parameter values. During data fusion, the radar parameters for each band can be standardized first, ensuring that monitoring results from different regions fall within the same range. Based on spatial interpolation and spatial stitching methods, nationwide monitoring results can then be obtained. In practice, the average RCS of insects and birds in the S, C, and X bands can also be summarized from historical data for inversion quantification.

[0042] The visualization module can be designed with both single-station and multi-station display interfaces. The single-station display interface shows the radar echo classification and inversion quantization results for each station's corresponding radar, including at least the vertical distribution of insect and bird density, vertical distribution of insect and bird velocity, and vertical distribution of horizontal movement direction. The multi-station display interface displays a mosaic of integrated monitoring results from single-band multi-weather radar, or a mosaic of fused monitoring results from multi-band multi-weather radar, including at least a mosaic of insect and bird density, insect and bird reflectivity, and insect and bird movement velocity. Switching between the single-station and multi-station display interfaces, and selecting which monitoring results are displayed in the multi-station display interface, can be done based on external selection.

[0043] Furthermore, both the single-station display interface and the multi-station display interface can overlay corresponding geographic maps, temperature, humidity, and wind field data.

[0044] The visualization results of this invention should be applicable to the needs of multiple fields, including ecological protection (such as bird habitat monitoring), agricultural protection (such as pest outbreak early warning), and aviation safety (such as bird strike risk early warning).

[0045] Step Six: Historical Data Analysis Module. This module stores the real-time insect and bird monitoring data obtained in Step Four (Inversion and Quantification Module) for long-term storage and analyzes it based on time series data. By analyzing data such as the migration time, path, and population changes of insects and birds, a long-term dynamic model of insect and bird migration is established to reveal their migration patterns at different time scales (such as daily, seasonal, and annual).

[0046] Therefore, this invention provides a large-scale insect and bird migration monitoring system based on a weather radar network. The implementation steps will be described below with specific embodiments: To verify the method described above, based on the measured data of multiple S-band dual-polarized weather radars in China's new generation weather radar network, the large-scale insect and bird migration monitoring system based on a weather radar network described in this invention is used to complete the real-time monitoring of large-scale insect and bird migration.

[0047] Step 1: Data Acquisition Module. This module collects measured data from the S-band dual-polarization weather radar of China's new generation weather radar network, including reflectivity factor, radial velocity, differential reflectivity, differential phase, correlation coefficient, and velocity spectral width.

[0048] Step 2: Quality control module performs quality control on the radar data collected in Step 1, removing ground clutter, super-reflection echoes, and radio frequency interference echoes.

[0049] Step 3: Echo identification module. The weather radar base data that has passed the quality control in step 2 is used as features. The corresponding band classification model is applied to classify the radar echoes into 3 categories: birds, insects, and meteorology. Figure 2 The data is a classification result of the 2.4° elevation angle measurement data of a dual-polarization weather radar at station A in a certain location in China at 00:22 on September 22, 2024. The echoes mainly originated from birds. Figure 3 The classification results are for the 2.4° elevation angle measurement data of the dual-polarization weather radar at 14:22 on September 22, 2024, at station A. The echoes mainly originated from insects.

[0050] Step 4: Inversion Quantization Module. Based on the insect and bird echo identification results in Step 3, combined with data such as reflectivity factor and radial velocity, average radar cross section data of birds, and average radar cross section data of insects, the vertical distribution of bird density and velocity and the vertical distribution of insect density and velocity can be calculated.

[0051] Step 5: Visualization module, which outputs the results of steps 3 and 4 through a visualization interface. Figure 4 The data is derived from the bird density, speed, and vertical distribution of direction measured by the dual-polarization weather radar at 00:22 on September 22, 2024. Figure 5 The data is derived from the vertical distribution of insect density, speed, size, and direction measured by the dual-polarization weather radar at 14:22 on September 22, 2024.

[0052] Step Six: Integrate the vertical distribution of insect and bird density at multiple times in Step Five according to altitude, and analyze it according to the time series to obtain the diurnal and seasonal patterns of insect and bird migration. Figure 6 This is a time series of bird vertical cumulative density data from the dual-polarization weather radar at Station A on September 22, 2024. Figure 7 This is a time series of insect vertical cumulative density data from the dual-polarization weather radar at Station A on September 22, 2024. Figure 8 This is a time series of daily vertical cumulative density of birds in a certain region of China from March to November 2021 on Bilibili. Figure 9 This is a time series of the daily vertical cumulative density of insects on Bilibili from March to November 2021.

[0053] The specific embodiments described above only illustrate the design principles of the present invention. The shapes and names of the components in this description may differ and are not limited. Therefore, those skilled in the art can modify or make equivalent substitutions to the technical solutions described in the foregoing embodiments; and these modifications and substitutions do not depart from the inventive spirit and technical solutions of the present invention, and should all fall within the protection scope of the present invention.

Claims

1. A large-scale bird migration monitoring system based on a weather radar network, characterized in that, The application relates to a weather radar insect and bird echo monitoring system. The system comprises a data acquisition module, a quality control module, an echo recognition module, an inversion quantization module, a visualization module and a historical data analysis module. The data acquisition module is used for acquiring real-time monitoring data of S-band, C-band and X-band weather radars of a national weather radar network, and the data comprises radar-based data. The quality control module is used for quality control of the collected radar-based data. The echo recognition module is used for constructing recognition features based on the quality-controlled radar-based data, classifying and recognizing radar echoes by using recognition models corresponding to different wave bands, and obtaining classification and recognition results. The inversion quantization module is used for inverting the vertical distribution of insect and bird density, insect and bird speed and movement direction based on the classified and recognized insect and bird echoes, and obtaining inversion quantization results. The visualization module is used for combining the classification and recognition results and the inversion quantization results of single-wave-band single weather radars into local monitoring results, and displaying the local monitoring results of single-wave-band single weather radars, comprehensive monitoring results of single-wave-band multiple weather radars or fusion monitoring results of multiple-wave-band multiple weather radars through a visualization interface.

2. The system of claim 1, wherein, The quality control module comprises a ground clutter and super-refraction echo suppression unit and a radio frequency interference echo suppression unit. The ground clutter and super-refraction echo suppression unit is used for removing ground clutter and super-refraction echoes by using the time-invariant characteristics of the ground clutter and super-refraction echoes. The radio frequency interference echo suppression unit is used for removing radio frequency interference echoes by using the long strip shape characteristics of the radio frequency interference echoes.

3. The system of claim 1, wherein, The echo recognition module uses radar-based data and statistical data thereof to construct a feature space, and applies a machine learning model or a deep learning model to classify and recognize radar echoes.

4. The system of claim 3, wherein, When classifying and recognizing radar echoes, the echo recognition module takes radar resolution units as sample points, and uses radar-based data and statistical data corresponding to the radar resolution units as point features, and uses an XGBoost or random forest classifier to learn the point features to obtain classification models corresponding to S-band weather radars and C-band weather radars. For X-band weather radars, the point features and surface features are extracted to form mixed features, and a ResNet50 or VGG16 neural network is used to learn the mixed features to obtain a classification model corresponding to the X-band weather radars; the surface features are output results of the neural network without a softmax layer, and represent echo spatial textures.

5. The system of claim 1, wherein, The inversion quantization module obtains inversion quantization results in the following manner: A quantitative relationship between reflectivity factors, average radar scattering cross sections and insect quantities of insect echoes is established by using a group Rayleigh scattering model; a quantitative relationship between reflectivity factors, average radar scattering cross sections and insect quantities of bird echoes is established by using a body shape and orientation Gaussian distribution model; based on the quantitative relationships, insect and bird quantities are estimated by combining reflectivity factors and average radar scattering cross sections corresponding to the insect and bird echoes; Radar reflectivity data of multiple elevation angles are used to invert the vertical distribution of insect and bird quantities to obtain insect and bird density vertical distribution; Based on radial velocity and velocity azimuth display technology, the vertical distribution of movement speed and the vertical distribution of movement direction of insect and bird groups are estimated.

6. The system of claim 1, wherein, The real-time monitoring data acquired by the data acquisition module further comprises meteorological data; the visualization interface of the visualization module is superimposed with corresponding geographical maps and meteorological data.

7. The system of claim 1 or 6, wherein, The visualization module is divided into a single-station display interface and a multi-station display interface; The single-station display interface displays the local monitoring results of each station corresponding to the weather radar; The multi-station display interface displays the comprehensive monitoring results of single-band multi-weather radars obtained by spatial interpolation and spatial splicing of the local monitoring results of part or all of the same band weather radars within the global spatial range, or displays the fusion monitoring results of multi-band multi-weather radars obtained by spatial interpolation, spatial splicing and band fusion of the local monitoring results of all-band weather radars within the global spatial range; The band fusion is: when the S-band, C-band and X-band weather radars cover the same space, according to the priority of S-band > X-band > C-band, the echo classification results and the inversion quantitative results of one band weather radar are selected as the data of the overlapping area.

8. The system of claim 1, wherein, The system further comprises a historical data analysis module for storing the inversion quantitative results and analyzing according to the time sequence.

9. The system of claim 8, wherein, The historical data analysis module further establishes a long-term dynamic model of insect-bird migration by analyzing the migration time, path and quantity change of insect-birds.

Images