Method for identifying bird flu overflow risk area based on waterfowl-sea bird interface

Abstract

The invention relates to an avian influenza overflow risk area identification method based on a waterfowl-sea bird interface. The method comprises the following steps: obtaining index data and environment data of an avian influenza waterfowl-sea bird host; wherein the index data comprises species distribution point data of bird flu waterfowl-sea bird hosts and sea bird flu occurrence point data; screening species distribution point data month by month; judging whether the species distribution point data is sufficient or not according to a preset condition; if all species distribution point data are sufficient, calculating a waterfowl-sea bird overflow risk index based on an activity intensity method; otherwise, calculating a waterfowl-sea bird overflow risk index based on a species richness method, further identifying an avian influenza overflow risk area between waterfowls and sea birds, and identifying an avian influenza overflow risk area between waterfowls and sea birds by integrating the avian influenza waterfowl-sea bird host index data and the environmental data according to different distribution conditions of species distribution points. And respectively calculating waterfowl-sea bird overflow risk indexes pixel by pixel by adopting different methods so as to accurately identify an avian influenza overflow risk area between waterfowls and sea birds.

Description

Technical Field

[0001] This application relates to the field of avian influenza spillover identification technology, and in particular to a method for identifying avian influenza spillover risk areas based on the waterfowl-seabird interface. Background Technology

[0002] Avian influenza is an infectious disease of birds caused by influenza A viruses. Its cross-species transmission poses a serious threat to wild birds, the poultry industry, and public health. Traditionally, seabirds were considered to have a low risk of virus exposure and play a secondary role in transmission networks due to their unique ecological niche. However, in recent years, there have been several large-scale seabird deaths globally, and the spread of highly pathogenic strains such as H5N1 among wild birds has intensified. In particular, the role of seabirds as a non-traditional host in transmission has become increasingly prominent. Waterfowl and seabirds form an "ecological bridge" by sharing coastal habitats, significantly increasing the risk of cross-species transmission of the virus.

[0003] Currently, there are various technical approaches used to assess the risk of cross-interface transmission of avian influenza viruses, but significant limitations still exist.

[0004] For example, traditional expert consultation-based assessment tools, such as the Pandemic Influenza Risk Assessment Tool (TIPRA), are systematic semi-quantitative assessment frameworks developed by international public health organizations (such as the WHO). They aim to assess the potential risk of zoonotic influenza viruses causing human pandemics by integrating viral characteristics, host infection manifestations, and animal epidemiological information. However, TIPRA relies on multidisciplinary experts to conduct targeted assessments of specific strains. While it can support risk rating and decision-making, it is highly subjective, cumbersome, and difficult to meet the needs of high-throughput, multi-strain automated assessments.

[0005] While prediction methods based on viral genome sequences and deep learning have high accuracy for known strains, they are highly dependent on viral HA / NA gene sequences, making them difficult to apply to unsequenced or new recombinant strains, and they only remain at the level of viral strains.

[0006] While methods based on waterbird host activity entropy can reveal the spatial pattern of avian influenza exposure risk, they only focus on a single group and cannot assess the risk of cross-host interface transmission. Furthermore, they are limited by the availability of dynamic species distribution data.

[0007] The novel avian influenza high-risk host species identification method based on viral sharing entropy, although the viral sharing entropy model introduces the concept of spatial overlap between hosts to assess potential viral sharing risk, only stays at the species level and is difficult to achieve accurate risk positioning at the geographical scale. Moreover, the assessment is based on static distribution and does not fully incorporate dynamic temporal factors such as migration behavior, which limits its comprehensive characterization of virus transmission risk.

[0008] While virus-sharing prediction methods that integrate species geographic distribution changes and phylogenetic similarities can identify potential first encounter events, they primarily focus on mammals and do not adequately consider the seasonal migration dynamics of avian hosts, thus limiting their applicability and accuracy in identifying high-risk areas for avian influenza.

[0009] Therefore, there is an urgent need to provide a method for identifying avian influenza spillover risk areas based on the waterfowl-seabird interface to solve the above-mentioned technical problems. Summary of the Invention

[0010] This application provides a method for identifying avian influenza spillover risk areas at the waterfowl-seabird interface. By integrating avian influenza waterfowl-seabird host index data and environmental data, and combining monthly or yearly dynamic modeling strategies, robust binary distribution prediction results are generated. Depending on whether the species distribution point data is sufficient or insufficient, the waterfowl-seabird spillover risk index (WSSR I) is calculated pixel by pixel using two methods. This accurately characterizes the potential risk of cross-interface transmission of avian influenza virus in high-risk interfaces such as coastal ecotones, thereby identifying avian influenza spillover risk areas between waterfowl and seabirds. This provides a scientific basis for cross-interface prevention and control under the "One Health" framework, and is particularly suitable for forward-looking risk warning in coastal areas of developing countries, with significant practical application value.

[0011] This application provides a method for identifying avian influenza spillover risk areas based on the waterfowl-seabird interface, comprising the following steps: acquiring indicator data and environmental data of avian influenza waterfowl-seabird hosts; wherein, the indicator data includes species distribution point data of avian influenza waterfowl-seabird hosts and avian influenza outbreak point data in seabirds, the species distribution point data includes species name, species classification, observation time and latitude and longitude coordinates, the seabird avian influenza outbreak point data includes virus type, outbreak country, outbreak time and latitude and longitude coordinates, and the environmental data includes two types of environmental variables: terrestrial and marine; filtering the species distribution point data monthly; determining whether the species distribution point data is sufficient according to preset conditions; if all species distribution point data is sufficient, calculating the waterfowl-seabird spillover risk index based on activity intensity; otherwise, calculating the waterfowl-seabird spillover risk index based on species richness; and identifying avian influenza spillover risk areas between waterfowl and seabirds based on the obtained spillover risk index.

[0012] Optionally, the species distribution point data is filtered monthly, including: removing species distribution point data with missing or duplicate latitude and longitude coordinates; excluding spatial outliers based on biogeographical boundaries; removing redundant points according to raster resolution, retaining only one species distribution point data in each raster cell; and statistically analyzing the filtered species distribution point data by month to construct a species classification hierarchy and obtain the number of species distribution points each month.

[0013] Optionally, the data on species distribution points can be determined based on preset conditions. These preset conditions include whether the number of species distribution points per month is less than a specific value. If the number is greater than the specific value, the data on species distribution points is considered sufficient; otherwise, the data on species distribution points is considered insufficient.

[0014] Optionally, if sufficient data on all species distribution points are available, a waterfowl-seabird spillover risk index can be calculated based on activity intensity. This includes: using monthly data on the species distribution points and background points (equal in number to the number of species distribution points) randomly generated monthly outside the species distribution range as pseudo-missing samples, and the corresponding environmental data as a training set; constructing monthly species distribution models for avian influenza waterfowl and seabird hosts based on multiple machine learning models, and then generating monthly binary distribution maps of avian influenza waterfowl and seabird hosts for each raster cell; wherein the binary distribution map indicates whether a certain species exists in a specific region in a specific month. The monthly binary distribution map represents the cumulative number of months in which each waterfowl and seabird host species appears within each pixel. Based on the monthly binary distribution map, activity intensity models for avian influenza waterfowl and seabird hosts are constructed pixel by pixel, yielding the avian influenza waterfowl host activity entropy (WFAE) and avian influenza seabird host activity entropy (SAE), respectively. The activity entropy represents a quantitative indicator of bird activity intensity; a higher activity entropy indicates a greater risk of avian influenza exposure for waterfowl and seabird hosts. Based on the avian influenza waterfowl host activity entropy (WFAE) and the avian influenza seabird host activity entropy (SAE), a waterfowl-seabird spillover risk index is calculated pixel by pixel, described by the following formula:

[0015] WSSRI = WFAE × SAE

[0016] In the formula, WSSRI represents the waterfowl-seabird spillover risk index, which reflects the degree of spatial co-occurrence and temporal overlap between waterfowl and seabird host species of avian influenza. The higher the WSSRI, the higher the risk of avian influenza spillover across the interface between waterfowl and seabirds.

[0017] Optionally, based on the monthly species distribution point data and the corresponding environmental data as training data, monthly species distribution models for avian influenza waterfowl hosts and seabird hosts are constructed according to multiple machine learning models, thereby generating monthly binary distribution maps of avian influenza waterfowl hosts and seabird hosts on each raster cell. This includes: using the monthly species distribution point data and the corresponding environmental data as training data, constructing monthly species distribution models for waterfowl hosts and seabird hosts according to three mainstream machine learning methods: generalized linear model, random forest, and maximum entropy model; randomly dividing the training data into five non-overlapping subsets, and using five-fold cross-validation to optimize the model parameters; averaging the continuous prediction maps output by each of the monthly species distribution models pixel by pixel to generate an average prediction map; wherein, the continuous prediction map represents the probability of a certain waterfowl or seabird host species existing in each raster cell in a certain month, with a value ranging from 0 to 1 continuously; the average prediction map is obtained by averaging the continuous prediction maps generated by multiple monthly species distribution models; based on TSS (True Skill Strategy)... (ic) Determine the optimal threshold and convert the continuous average prediction map into a monthly binary distribution map; wherein, the optimal threshold represents the critical value of the probability of a species existing in a certain month, that is, if the average prediction probability of a certain species in a certain grid cell in a certain month is greater than or equal to the threshold, then the species is determined to exist, otherwise it is determined to not exist.

[0018] The spatial prediction range of the model is limited by overlaying the monthly binary distribution map with the biogeographic boundary line.

[0019] Optionally, based on the TSS (True Skill Statistic), an optimal threshold is determined, and the average prediction map is converted into a monthly binary distribution map. This includes: setting thresholds according to preset intervals based on the range of the average prediction map, calculating the confusion matrix under each threshold, where true positives are the number of samples that actually exist and are predicted to exist, false positives are the number of samples that do not actually exist but are predicted to exist, true negatives are the number of samples that actually exist but are predicted not to exist, and false negatives are the number of samples that do not actually exist and are not predicted to exist; and based on the obtained confusion matrix, the TSS is calculated, described by the formula: TSS = Sensitivity + Specificity - 1.

[0020] Sensitivity = TP / (TP+FN)

[0021] Specifiability = TN / (TN+FP); where TP represents true positives, FP represents false positives, TN represents true negatives, and FN represents false negatives. The TSS range is [-1, 1]. A higher TSS value indicates a stronger model discrimination ability. Among all thresholds, the threshold that maximizes the TSS value is selected as the optimal existence threshold. According to the average prediction map, if the average prediction probability of a certain species in a certain raster cell in a certain month is greater than or equal to the optimal existence threshold, then the species is determined to exist and assigned a value of 1; otherwise, it is determined to not exist and assigned a value of 0, thus obtaining a monthly binary distribution map.

[0022] Optionally, based on the monthly binary distribution map, pixel-by-pixel activity intensity models for avian influenza waterfowl hosts and seabird hosts are constructed to obtain the avian influenza waterfowl host activity entropy (WFAE) and avian influenza seabird host activity entropy (SAE), respectively. This includes: according to the monthly binary distribution map, counting the cumulative number of months in which each waterfowl host species and seabird host species appear in each pixel; calculating the proportion of residence time of each host species in that pixel, i.e., the proportion of the number of months in which the species appears to the total number of months in which the species appears in the same group, i.e., all waterfowl host species or all seabird host species; and calculating the activity entropy of waterfowl hosts and seabird hosts based on the residence time proportion, wherein the calculation formula for the activity entropy is as follows:

[0023]

[0024] In the formula, AE represents the activity entropy, dur i p represents the total number of months that the i-th type of bird stays in a given raster cell. i This represents the ratio of the total number of months that the i-th type of bird stays in a given raster cell to the total number of months that all birds stay.

[0025] Optionally, if some of the species distribution point data is insufficient, a waterfowl-seabird spillover risk index is calculated based on species richness, including: merging the monthly species distribution point data to obtain annual species distribution point data; using the annual species distribution point data and background points randomly generated outside the species distribution range with an equal number of species distribution points as pseudo-missing samples, and the corresponding environmental variables as the training set, constructing annual species distribution models for avian influenza waterfowl hosts and seabird hosts according to multiple machine learning models, and then generating an annual binary distribution map of avian influenza waterfowl hosts and seabird hosts for each raster cell; wherein, the binary distribution map represents the distribution probability of whether a certain species exists in each raster cell, and the annual binary distribution map represents the statistical distribution of each cell. The cumulative probabilities of each waterfowl host species and seabird host species within the region are calculated. Based on the binary distribution map, the binary distribution maps of the waterfowl host species and the seabird host species are superimposed, and the statistics are accumulated for each raster cell. A species richness model for avian influenza waterfowl hosts and seabird hosts is constructed cell by cell to obtain avian influenza waterfowl host richness raster data and avian influenza seabird host richness raster data. The richness represents the species diversity of each waterfowl host species and seabird host carrying the avian influenza virus within each cell. Based on the avian influenza waterfowl host richness raster data and the avian influenza seabird host richness raster data, waterfowl-seabird species pairs are calculated cell by cell, and a waterfowl-seabird spillover risk index is calculated. The waterfowl-seabird spillover risk index is described by the following formula:

[0026]

[0027] In the formula, WSSRI represents the waterfowl-seabird spillover risk index, which reflects the ecological contact opportunities between waterfowl host species and seabird host species of avian influenza. The higher the WSSRI, the higher the risk of avian influenza spillover across the interface between waterfowl and seabirds.

[0028] Optionally, based on the obtained spillover risk index, identifying avian influenza spillover risk areas between waterfowl and seabirds includes: using the seabird avian influenza outbreak data as a validation dataset; validating the WSSRI results based on the validation dataset, calculating the true positive rate and false positive rate at each threshold of the WSSRI index, and plotting the false positive rate and true positive rate corresponding to each threshold as coordinate points in a coordinate system to obtain the ROC curve of the false positive rate and the true positive rate as a function of the threshold of the spillover risk index, and obtaining the area under the curve (AUC) value, which is used to verify whether the WSSRI index can effectively distinguish between the true positive samples and the true negative samples; selecting the threshold corresponding to the point that maximizes the true positive rate minus the false positive rate value as the optimal threshold of the waterfowl-seabird spillover risk index; wherein, areas where the waterfowl-seabird spillover risk index is lower than the optimal threshold are non-high-risk areas, and areas where it is higher than the optimal threshold are high-risk areas.

[0029] Optionally, using the seabird avian influenza outbreak data as a validation dataset further includes: calculating the Euclidean distance between each case point and the nearest natural water body based on the latitude and longitude of each case point and the latitude and longitude of the nearest natural water body; setting screening conditions based on the ecological dependence of the avian influenza seabird host on water bodies, and eliminating atypical habitat data according to the screening conditions; wherein, the screening conditions include retaining only case points whose distance from the coastline does not exceed a first preset value or whose distance from the nearest natural water body does not exceed a second preset value, thereby eliminating atypical habitat data, wherein atypical habitat data refers to case point data whose geographical location of the avian influenza outbreak point does not conform to the natural ecological habits of the host species.

[0030] This application has at least the following advantages:

[0031] The above steps mainly involve acquiring indicator data and environmental data, integrating avian influenza waterfowl-seabird host indicator data and environmental data, and combining monthly or year-round dynamic modeling strategies to generate robust binary distribution prediction results. Depending on whether the species distribution point data is sufficient or insufficient, the waterfowl-seabird spillover risk index (WSSRI) is calculated pixel by pixel based on either activity intensity or species richness methods. This accurately characterizes the potential risk of cross-interface transmission of avian influenza virus in high-risk interfaces such as coastal ecotones, in order to identify avian influenza spillover risk areas between waterfowl and seabirds. This provides a scientific basis for cross-interface prevention and control under the "One Health" framework, and is particularly suitable for forward-looking risk warning in coastal areas of developing countries, with significant practical application value. Attached Figure Description

[0032] Figure 1 This is a diagram illustrating the application environment of a method for identifying avian influenza spillover risk zones based on the waterfowl-seabird interface in one embodiment.

[0033] Figure 2 This is a flowchart illustrating the steps of a method for identifying avian influenza spillover risk zones based on the waterfowl-seabird interface in one embodiment.

[0034] Figure 3 Here is a flowchart illustrating a method for identifying avian influenza spillover risk areas based on the waterfowl-seabird interface in one embodiment;

[0035] Figure 4 This is a flowchart illustrating a method for calculating the waterfowl-seabird spillover risk index based on activity intensity in one embodiment.

[0036] Figure 5 This is a schematic diagram illustrating the activity entropy of avian influenza waterfowl hosts in one embodiment;

[0037] Figure 6 This is a schematic diagram illustrating the activity entropy of avian influenza seabird hosts in one embodiment;

[0038] Figure 7 This is a schematic diagram illustrating the calculation of the waterfowl-seabird spillover risk index based on activity entropy in one embodiment;

[0039] Figure 8 This is a schematic diagram of a waterfowl-seabird spillover risk index based on activity entropy in one embodiment;

[0040] Figure 9 This is a schematic diagram illustrating the species richness of waterfowl hosts for avian influenza in one embodiment;

[0041] Figure 10 This is a schematic diagram illustrating the species richness of seabird hosts of avian influenza in one embodiment;

[0042] Figure 11 This is a schematic diagram illustrating the calculation of the waterfowl-seabird spillover risk index based on abundance in one embodiment;

[0043] Figure 12 This is a schematic diagram of a waterfowl-seabird spillover risk index based on abundance in one embodiment;

[0044] Figure 13 This is a schematic diagram illustrating the ROC curve verification of the waterfowl-seabird spillover risk index based on activity entropy in one embodiment;

[0045] Figure 14 This is a schematic diagram illustrating the avian influenza spillover risk area at the waterfowl-seabird interface based on activity entropy in one embodiment.

[0046] Figure 15 This is a schematic diagram illustrating the ROC curve verification of the waterfowl-seabird spillover risk index based on abundance in one embodiment;

[0047] Figure 16 This is a schematic diagram of the avian influenza spillover risk area at the waterfowl-seabird interface based on abundance in one embodiment.

[0048] Figure 17 This is a block diagram illustrating the structure of a system for identifying avian influenza spillover risk zones based on the waterfowl-seabird interface, as shown in one embodiment.

[0049] Figure 18 This is a schematic structural diagram of a computer device in one embodiment. Detailed Implementation

[0050] The present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the scope of the present application.

[0051] For ease of understanding, the system to which this application applies will first be described. This application provides a method for identifying avian influenza spillover risk zones based on the waterfowl-seabird interface, which can be applied to, for example... Figure 1 The system architecture shown includes a user-space file server 103 and a terminal device 101. The terminal device 101 communicates with the user-space file server 103 via a network. The user-space file server 103 can be a file server based on the NFSv3 / v4 protocol, running in a Linux environment. NFS (Network File System) is a network abstraction on top of a file system, allowing remote clients running on the terminal device 101 to access the file system over the network in a manner similar to a local file system. The terminal device 101 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. The user-space file server 103 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0052] Figure 2 A flowchart illustrating a method for identifying avian influenza spillover risk areas based on the waterfowl-seabird interface, provided in this application embodiment, is included in the following steps:

[0053] S201. Obtain indicator data and environmental data for avian influenza waterfowl-seabird hosts; among which, indicator data includes species distribution point data for avian influenza waterfowl-seabird hosts and avian influenza outbreak point data for seabirds. Species distribution point data includes species name, species classification, observation time and latitude and longitude coordinates. Seabird avian influenza outbreak point data includes virus type, country of outbreak, outbreak time and latitude and longitude coordinates. Environmental data includes two types of environmental variables: terrestrial and marine.

[0054] S202. Monthly screening of species distribution point data;

[0055] S203. Based on the preset conditions, determine whether the species distribution point data is sufficient;

[0056] S204. If data on the distribution points of all species are sufficient, calculate the spillover risk index between waterfowl and seabirds based on the activity intensity method.

[0057] Otherwise, calculate the waterfowl-seabird spillover risk index based on species richness.

[0058] S205. Based on the obtained spillover risk index, identify the risk areas for avian influenza spillover between waterfowl and seabirds.

[0059] In this embodiment, it should be noted that by acquiring indicator data and environmental data, the acquired species distribution point data are processed and analyzed. If all species distribution point data are sufficient, the waterfowl-seabird spillover risk index is calculated based on activity intensity; otherwise, the waterfowl-seabird spillover risk index is calculated based on species richness. The waterfowl-seabird spillover risk index (WSSRI) is calculated pixel-by-pixel according to the corresponding method under different conditions, accurately characterizing the potential risk of cross-interface transmission of avian influenza virus in high-risk interfaces such as coastal ecotones, in order to identify avian influenza spillover risk areas between waterfowl and seabirds. This method not only compensates for the shortcomings of existing technologies in monitoring emerging hosts such as seabirds, but also provides a scientific basis for cross-interface prevention and control under the "One Health" framework. It is particularly suitable for forward-looking risk warning in coastal areas of developing countries and has significant practical application value.

[0060] The following is a detailed explanation of each step:

[0061] Please refer to Figure 2 , Figure 3 As shown, step S201 involves obtaining indicator data and environmental data of avian influenza waterfowl-seabird hosts;

[0062] In this embodiment, it should be noted that the indicator data includes species distribution point data for avian influenza waterfowl-seabird hosts and avian influenza outbreak point data for seabirds. Species distribution point data primarily comes from observation records of avian influenza waterfowl and seabird hosts worldwide, obtained from the Global Biodiversity Information Network (GBIF) and the eBird database, including both human and machine observation records. eBird is an open bird observation platform based on public participation. It collects standardized records of bird species, numbers, observation locations, and times submitted by birdwatchers worldwide through mobile applications and websites, constructing a dynamically updated database of bird distribution and migration to provide data support for scientific research, biodiversity conservation, and public education. Species distribution point data includes species name, species classification, observation time, and latitude and longitude coordinates. Avian influenza outbreak point data for seabirds includes virus type, outbreak country, outbreak time, and latitude and longitude coordinates, primarily obtained from the Food and Agriculture Organization of the United Nations (FAO), used to provide a data foundation for subsequent spillover risk index verification.

[0063] Environmental data includes two categories of environmental variables: terrestrial and marine. Terrestrial environmental variables were obtained from the WorldClim database, while marine environmental variables were obtained from the ERA5 and Bio-ORACLE databases. The screening and processing of environmental variables primarily involved retaining variables with importance values ​​greater than zero. For variable pairs with Pearson correlation coefficients greater than 0.7, only the variable with the higher importance value was retained. All environmental data were uniformly resampled to a spatial resolution of 0.5° to provide a data foundation for subsequent construction of species distribution models. Specific environmental variables were determined using Tables 1 and 2.

[0064] Table 1 Terrestrial Environmental Variables

[0065]

[0066]

[0067] Table 2 Marine environmental variables

[0068]

[0069]

[0070] Please continue to refer to Figure 2 , Figure 3 As shown, step S202, monthly screening of species distribution point data, includes: removing species distribution point data with missing or duplicate latitude and longitude, excluding spatial outliers based on biogeographical boundaries, removing redundant points based on raster resolution, retaining only one species distribution point data in each raster cell, statistically analyzing the screened species distribution point data by month, constructing a species classification hierarchy, and obtaining the number of species distribution points each month.

[0071] In this embodiment, it should be noted that biogeographical boundaries refer to the dividing areas between different biotas (zoota and phytota) on Earth, where there are significant differences in biological composition, species distribution, or ecosystem characteristics. These boundaries are usually formed by geographical, climatic, geological, or ecological factors, influencing the dispersal, evolution, and distribution of species. Spatial outliers refer to observations or objects in spatial data that exhibit significant differences from their neighboring locations. These points show obvious differences in spatial distribution from their surrounding environment. The filtered species distribution point data are statistically analyzed by month. Specifically, when processing the avian influenza waterfowl-seabird host species distribution point data in Excel, a "month" column is first generated based on the date field, a time label is added to each record, and then the species distribution is statistically analyzed by month using a pivot table.

[0072] Please continue to refer to Figure 2 , Figure 3 As shown, in step S203, the species distribution point data is determined to be sufficient according to preset conditions. The preset conditions include: whether the number of species distribution points per month is less than a specific value. If it is greater than the specific value, it means that the species distribution point data is sufficient; otherwise, it means that the species distribution point data is insufficient.

[0073] In this embodiment, it should be noted that, specifically, the sufficiency of avian influenza waterfowl and seabird host species distribution point data is determined based on whether the number of species distribution points per month is no less than 9. Based on whether the number of species distribution points is less than 9, an appropriate algorithm is selected for the subsequent calculation of the spillover risk index, in order to more accurately identify avian influenza spillover risk areas between waterfowl and seabirds.

[0074] Reference Figure 2 , Figure 4 As shown, in step S204, if data on the distribution points of all species are sufficient, the waterfowl-seabird spillover risk index is calculated based on the activity intensity method, including:

[0075] Step S2041: Based on the monthly species distribution point data and background points generated monthly outside the species distribution range with an equal number of species distribution points as pseudo-missing samples, and the corresponding environmental data as training set, monthly species distribution models of avian influenza waterfowl hosts and seabird hosts are constructed according to multiple machine learning models, and then monthly binary distribution maps of avian influenza waterfowl hosts and seabird hosts are generated on each raster cell.

[0076] In this embodiment, it should be noted that background points, which are randomly generated monthly outside the corresponding species distribution range and equal in number to the species distribution points, are used as pseudo-missing samples. By comparing the actual distribution environment of the species with an environment that could not exist, the subsequent monthly species distribution model can more robustly identify the niche boundaries of the species, significantly improving its ability and accuracy in predicting species distribution. It also conforms to the basic assumptions of niche theory, enhancing the scientific nature of the entire modeling process and the interpretability of the results.

[0077] Specifically, in one example, based on monthly species distribution point data and corresponding environmental data as training data, monthly species distribution models for waterfowl and seabird hosts are constructed using three mainstream machine learning methods: Generalized Linear Model (GLM), Random Forest (RF), and Maximum Entropy Model (MAXENT). The training data is randomly divided into five non-overlapping subsets, and five-fold cross-validation is used to optimize the model parameters. The continuous prediction maps output by each monthly species distribution model are averaged pixel by pixel to generate an average prediction map. The continuous prediction map represents the probability of a certain waterfowl or seabird host species existing in each raster cell in a certain month, with a value ranging from 0 to 1. The average prediction map is obtained by averaging the continuous prediction maps generated by multiple monthly species distribution models. Specifically, the results of three continuous prediction maps are arithmetically averaged to obtain the average prediction map for the species, thereby reducing the random error or overfitting risk of a single model and improving the robustness of the results.

[0078] The optimal threshold is determined based on TSS (True Skill Stability), and the continuous average prediction map is converted into a monthly binary distribution map. The optimal threshold represents the critical value of the probability of a species existing in a certain month. That is, if the average predicted probability of a species in a certain raster cell in a certain month is greater than or equal to the threshold, the species is determined to exist; otherwise, it is determined not to exist. The monthly binary distribution map is overlaid with the biogeographic boundary line to limit the spatial prediction range of the model.

[0079] In this embodiment, it should be noted that, specifically, based on the average prediction map range, thresholds are set according to preset intervals, and a confusion matrix is ​​calculated for each threshold, including True Positive (TP) – the number of samples that actually exist and are predicted to exist; False Positive (FP) – the number of samples that do not actually exist but are predicted to exist; False Negative (FN) – the number of samples that actually exist but are predicted not to exist; and False Negative (TN) – the number of samples that do not actually exist and are predicted not to exist. Based on the obtained confusion matrix, the TSS is calculated, and described by the formula:

[0080] TSS=Sensitivity+Specificity-1,

[0081] Sensitivity = TP / (TP+FN)

[0082] Specificity = TN / (TN+FP);

[0083] The TSS ranges from -1 to 1. A higher TSS value indicates a stronger model discrimination ability. Among all thresholds, the threshold that maximizes the TSS value is selected as the optimal existence threshold. According to the average prediction map, if the average prediction probability of a certain species in a certain raster cell in a certain month is greater than or equal to the optimal existence threshold, the species is judged to exist and assigned a value of 1. Otherwise, it is judged to not exist and assigned a value of 0, thus obtaining a monthly binary distribution map.

[0084] Specifically, in one example, the average predicted map ranges from (0-1). Thresholds are set at preset intervals of 0.01. All possible thresholds are iterated, such as 0.01, 0.02, 0.03, ..., 0.99. Then, the confusion matrix for each threshold is calculated: TP (True Positives), FP (False Positives), FN (False Negatives), and TN (False Positives): the number of samples that are neither actually present nor predicted to be present. Then, the TSS (Total Sensitivity Score) is calculated based on the confusion matrix, with the formula TSS = Sensitivity + Specificity - 1, where Sensitivity = TP / (TP + FN) and Specificity = TN / (TN + FP). The TSS range is [-1, 1], with higher values ​​indicating stronger model discrimination. Finally, the threshold that maximizes the TSS value is selected as the optimal threshold. The optimal threshold represents the critical value of the probability of a species existing in a certain month. If the average predicted probability of a certain species in a certain raster cell in a certain month is greater than or equal to the threshold, the species is considered to exist and is assigned a value of 1. Otherwise, it is determined to not exist and is assigned a value of 0. This converts the original continuous value (0-1) into a binary discrete value (0 or 1) to obtain a monthly binary distribution map.

[0085] Step S2042: Based on the monthly binary distribution map, construct the activity intensity models of avian influenza waterfowl hosts and seabird hosts pixel by pixel, and obtain the avian influenza waterfowl host activity entropy (WFAE) and avian influenza seabird host activity entropy (SAE) respectively.

[0086] In this embodiment, it should be noted that the activity entropy is mainly calculated based on the Shannon index. The Shannon index is a specific application of Shannon information entropy in the field of ecology. This index considers not only species richness but also species evenness. The Shannon index is positively correlated with community species diversity. When the species distribution in the community is even, the Shannon index is high, indicating that the community has rich species diversity. Conversely, when the community is dominated by a single species and the number of individuals of other species is relatively small, the Shannon index is low, indicating that the community has low species diversity.

[0087] Specifically, in one example, such as Figure 5 , Figure 6 As shown, based on the monthly binary distribution map, the cumulative number of months in which each waterfowl host species and seabird host species appear in each pixel is counted; the proportion of residence time of each host species in that pixel is calculated, that is, the proportion of the number of months in which the species appears to the total number of months in which the species appears in the same group, i.e., all waterfowl host species or all seabird host species, appear; based on the proportion of residence time, the activity entropy of waterfowl hosts and seabird hosts is calculated respectively. The formula for calculating the activity entropy is as follows:

[0088]

[0089] In the formula, AE represents the activity entropy, dur i p represents the total number of months that the i-th type of bird stays in a given raster cell. i This represents the ratio of the total number of months that the i-th type of bird stays in a given raster cell to the total number of months that all birds stay.

[0090] Step S2043: Based on the waterfowl host activity entropy (WFAE) and seabird host activity entropy (SAE) for avian influenza, calculate the waterfowl-seabird spillover risk index pixel by pixel, described by the formula:

[0091] WSSRI = WFAE × SAE

[0092] In the formula, WSSRI represents the waterfowl-seabird spillover risk index, which reflects the degree of spatial co-occurrence and temporal overlap between waterfowl and seabird host species of avian influenza. The higher the WSSRI, the higher the risk of avian influenza spillover across the interface between waterfowl and seabirds.

[0093] In this embodiment, it should be noted that, as Figure 7 , Figure 8 As shown, based on the obtained WFAE and SAE, the waterfowl-seabird spillover risk index WSSRI is calculated pixel by pixel to accurately characterize the potential risk of cross-interface transmission of avian influenza virus in high-risk interfaces such as coastal ecological ecotones, so as to accurately identify the avian influenza spillover risk area between waterfowl and seabirds.

[0094] Please continue to refer to Figure 2 , Figure 3Step S204: If some species distribution point data are insufficient, calculate the waterfowl-seabird spillover risk index based on species richness, including: merging monthly species distribution point data to obtain annual species distribution point data; using the annual species distribution point data and background points randomly generated outside the species distribution range with an equal number of species distribution points as pseudo-missing samples, and the corresponding environmental variables as the training set; constructing annual species distribution models for avian influenza waterfowl hosts and seabird hosts according to multiple machine learning models, thereby generating annual binary distribution maps of avian influenza waterfowl hosts and seabird hosts for each raster cell; based on the binary distribution maps, overlaying the binary distribution maps of waterfowl host species and seabird host species respectively, accumulating statistics for each raster cell, and constructing avian influenza waterfowl host and seabird host species richness models for each cell, obtaining avian influenza waterfowl host species richness (WFSR) raster data and avian influenza seabird host species richness (WFSR). Richness (SSR) raster data; based on WFSR and SSR raster data, waterfowl-seabird species pairs are calculated pixel by pixel, and the waterfowl-seabird spillover risk index is calculated. The waterfowl-seabird spillover risk index is described by the formula:

[0095]

[0096] In the formula, WSSRI represents the waterfowl-seabird spillover risk index, which reflects the ecological contact opportunities between waterfowl host species and seabird host species of avian influenza. The higher the WSSRI, the higher the risk of avian influenza spillover across the interface between waterfowl and seabirds.

[0097] In this embodiment, it should be noted that when data on the distribution points of some species is insufficient, the monthly analysis is changed to an annual analysis in the time dimension. The raster values ​​of the constructed species richness model are the sum of the number of species per pixel to form the species richness, which is used to calculate the spillover risk index in the subsequent calculation. The species activity intensity is calculated based on the Shannon-Wiener formula based on information entropy. The Shannon activity entropy of the host is obtained by using the activity months and their proportions calculated by the monthly distribution model, thereby calculating the spillover risk index.

[0098] Specifically, firstly, a year-round species distribution model is constructed using the acquired training set as input data. This then generates a year-round binary distribution map of avian influenza waterfowl and seabird hosts for each raster cell. The binary distribution map represents the probability of a species existing in each raster cell, and the year-round binary distribution map represents the cumulative probability of each waterfowl and seabird host species within each cell. Next, the binary distribution maps of all waterfowl and seabird host species are overlaid, and the statistics for each raster cell are accumulated to construct a species richness model for avian influenza waterfowl and seabird hosts. Finally, WFSR and SSR raster data are output. Figure 9 , Figure 10 Richness represents the species diversity of waterfowl and seabird host species carrying avian influenza virus within each pixel; such as Figure 11 , 12 As shown, based on WFSR and SSR raster data, waterfowl-seabird species pairs are calculated pixel-by-pixel to quantify the number of host species with potential cross-interface ecological contact in each raster. Specifically, in one example, if there are 3 waterfowl host species and 2 seabird host species in a certain raster cell, then each waterfowl species can pair with 2 seabird species, and the 3 waterfowl species can form a total of 6 different cross-interface transmission interspecies contact combinations with seabirds. Finally, the waterfowl-seabird spillover risk index is calculated based on the avian influenza waterfowl host species richness (WFSR) and avian influenza seabird host species richness (SSR).

[0099] Please continue to refer to Figure 2 , Figure 3 Step S205: Based on the obtained spillover risk index, identify the spillover risk area of ​​avian influenza between waterfowl and seabirds, including: using the data of avian influenza outbreaks in seabirds as a validation dataset; validating the WSSRI results based on the validation dataset, calculating the True Positive Rate (TPR) and False Positive Rate (FPR) at each threshold of the WSSRI index, and plotting the false positive rate and true positive rate corresponding to each threshold as coordinate points on a coordinate system to obtain the ROC curve of TPR and FPR changing with the threshold of the spillover risk index, and obtaining the area under the curve (AUC) value, which is used to test whether the WSSRI index can effectively distinguish between true positive samples and true negative samples; according to the ROC curve, selecting the threshold corresponding to the point that maximizes the TPR-FPR value as the optimal threshold of the waterfowl-seabird spillover risk index; where the waterfowl-seabird spillover risk index is lower than the optimal threshold, it is a non-high-risk area, and higher than the optimal threshold, it is a high-risk area.

[0100] In this embodiment, it should be noted that the data on avian influenza outbreaks in seabirds must be screened before being used as the validation dataset to more accurately validate the waterfowl-seabird spillover risk index results. Specifically, the Euclidean distance between each case site and the nearest natural water body is calculated based on the latitude and longitude of each case site and the latitude and longitude of the nearest natural water body. Based on the ecological dependence of avian influenza seabird hosts on water bodies, screening conditions are set, and atypical habitat data is eliminated according to these conditions. These screening conditions include retaining only case sites whose distance from the coastline does not exceed a first preset value or whose distance from the nearest natural water body does not exceed a second preset value. This could mean retaining case sites no more than 50 kilometers from the coastline or no more than 10 kilometers from the nearest natural water body, thereby eliminating atypical habitat data and ensuring the ecological rationality and accuracy of the validation dataset. Atypical habitat data refers to case site data where the geographical location of the avian influenza outbreak does not conform to the natural ecological habits of the host species.

[0101] Then, the waterfowl-seabird spillover risk index (WSSRI) results were validated using processed seabird avian influenza outbreak data. ROC curves were plotted based on the calculated True Positive Rate (TPR) and False Positive Rate (FPR) at each threshold of the WSSRI. The AUC value obtained from the ROC curves indicates that the WSSRI has good predictive ability. Furthermore, the optimal threshold for the WSSRI was determined, and areas above the optimal threshold were classified as high-risk areas, while areas below the optimal threshold were classified as non-high-risk areas, thus identifying avian influenza spillover risk areas at the waterfowl-seabird interface.

[0102] Specifically, in one example, such as Figure 13 As shown, the waterfowl-seabird spillover risk index (WSSRI) was calculated based on activity intensity. The results were validated using processed seabird avian influenza outbreak data. The AUC value obtained from the ROC curve was 0.886, indicating that the WSSRI has good predictive ability. The optimal threshold for the WSSRI was further determined to be 6. Figure 14 As shown, high-risk and non-high-risk areas are divided according to the optimal threshold.

[0103] In another example, such as Figure 15 As shown, the waterfowl-seabird spillover risk index (WSSRI) was calculated using a species richness-based method. The WSSRI results were validated using processed seabird avian influenza outbreak data. The AUC value obtained from the ROC curve was 0.662, indicating that the model has good discriminative ability. The optimal threshold for the WSSRI was further determined to be 32.5. Figure 16As shown, high-risk and non-high-risk areas are divided according to the optimal threshold, thereby identifying the risk areas of avian influenza spillover from the waterfowl-seabird interface.

[0104] The activity entropy-based method for identifying waterfowl-seabird spillover risk areas accurately captures the seasonal dynamics of waterfowl and seabirds, with an AUC of 0.886, demonstrating high predictive accuracy. In contrast, the species richness-based method simplifies the time dimension to a single integration layer, achieving an AUC of 0.662. While its predictive accuracy is slightly lower, it exhibits stronger universality in data-scarce regions. It should be noted that, to enable the parallel application of both activity intensity-based and species richness-based methods for calculating waterfowl-seabird spillover risk indices on a global scale, this application proposes the following assumption: the activity intensity-based method is not limited by the sufficiency of species distribution point data. Based on this assumption, the two methods can be independently implemented globally, supporting comparative analysis and complementary use of results. However, in practical applications, it is still necessary to determine whether the distribution point data for avian influenza waterfowl and seabird hosts in a specific region is sufficient. If data on the distribution points of all waterfowl and seabird host species are sufficient, the waterfowl-seabird spillover risk index is calculated using a method based on activity intensity; if data on the distribution points of some waterfowl or seabird host species are insufficient, the waterfowl-seabird spillover risk index is calculated using a method based on species richness.

[0105] The implementation principle of this embodiment is as follows: The above steps mainly involve acquiring indicator data and environmental data, processing and analyzing the acquired species distribution point data, and using integrated modeling of multiple machine learning algorithms. Based on a monthly or year-round dynamic modeling strategy, robust binary distribution prediction results can be generated even when waterfowl or seabird host species distribution data are insufficient, providing an accurate data foundation for subsequent calculations. When species distribution data is sufficient, the activity intensity-based method uses the Shannon-Wiener formula calculated based on information entropy, and utilizes the activity months and their proportions calculated by the monthly species distribution model to obtain the Shannon activity entropy of the host, thereby calculating the waterfowl-seabird spillover risk index. Otherwise, the monthly analysis is changed to a year-round analysis to obtain a year-round species distribution model and then a year-round binary distribution map. Based on the year-round binary distribution map, each raster cell is accumulated and statistically analyzed to obtain avian influenza waterfowl host richness (WFSR) raster data and avian influenza seabird host richness (SSR) raster data for subsequent calculation of the spillover risk index. Based on different species distribution points, two methods were used to calculate the Waterfowl-Seabird Spillover Risk Index (WSSRI) pixel-by-pixel. This method accurately characterizes the potential risk of cross-interface transmission of avian influenza virus in high-risk interfaces such as coastal ecotones, thereby identifying risk areas for avian influenza spillover between waterfowl and seabirds. This approach not only compensates for the shortcomings of existing technologies in monitoring emerging hosts such as seabirds, but also provides a scientific basis for cross-interface prevention and control under the "One Health" framework. It is particularly suitable for prospective risk warning in coastal areas of developing countries and has significant practical application value.

[0106] Reference Figure 17 As shown in the figure, this application embodiment also provides a system for identifying avian influenza spillover risk areas based on the waterfowl-seabird interface. This system may include: a data acquisition module 301, a data processing module 302, a condition analysis module 303, a risk calculation module 304, and a risk identification module 305; the main functions of each component module are as follows:

[0107] The data acquisition module 301 is used to acquire indicator data and environmental data of avian influenza waterfowl-seabird hosts. Among them, the indicator data includes species distribution point data of avian influenza waterfowl-seabird hosts and avian influenza occurrence point data of seabirds. The species distribution point data includes species name, species classification, observation time and latitude and longitude coordinate information. The avian influenza occurrence point data of seabirds includes virus type, country of outbreak, outbreak time and latitude and longitude coordinate information. The environmental data includes two types of environmental variables: terrestrial and marine.

[0108] Data processing module 302 is used to filter species distribution point data month by month;

[0109] The condition analysis module 303 is used to determine whether the species distribution point data is sufficient based on preset conditions.

[0110] The risk calculation module 304 is used to calculate the waterfowl-seabird spillover risk index based on the activity intensity method if all species distribution point data are sufficient; otherwise, it calculates the waterfowl-seabird spillover risk index based on the species richness method.

[0111] Risk identification module 305 is used to identify avian influenza spillover risk areas between waterfowl and seabirds based on the obtained spillover risk index.

[0112] like Figure 18 The diagram shown is a block diagram of a computer device according to an embodiment of this application. The term "computer device" is intended to represent various forms of digital computers or mobile devices. The digital computer may include a desktop computer, a portable computer, a workbench, a personal digital assistant, a server, a mainframe computer, and other suitable computers. The mobile device may include a tablet computer, a smartphone, a wearable device, etc.

[0113] like Figure 18 As shown, device 600 includes a computing unit 601, a ROM 602, a RAM 603, a bus 604, and an I / O interface 605. The computing unit 601, ROM 602, and RAM 603 are interconnected via the bus 604. The I / O interface 605 is also connected to the bus 604.

[0114] The computing unit 601 can execute various processes in the method embodiments of this application according to computer instructions stored in the read-only memory (ROM) 602 or computer instructions loaded from the storage unit 608 into the random access memory (RAM) 603. The computing unit 601 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. The computing unit 601 can include, but is not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. In some embodiments, the methods provided in the embodiments of this application can be implemented as computer software programs, which are tangibly contained in a computer-readable storage medium, such as the storage unit 608.

[0115] RAM 603 can also store various programs and data required for the operation of device 600. Part or all of the computer program can be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609.

[0116] The input unit 606, output unit 607, storage unit 608, and communication unit 609 in device 600 can be connected to I / O interface 605. The input unit 606 can be, for example, a keyboard, mouse, touchscreen, or microphone; the output unit 607 can be, for example, a display, speaker, or indicator light. Device 600 can exchange information and data with other devices through the communication unit 609.

[0117] It should be noted that the device may also include other components necessary for normal operation. It may also include only the components necessary for implementing the solution of this application, without necessarily including all the components shown in the figures.

[0118] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SOCs), payload programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof.

[0119] The computer instructions used to implement the methods of this application may be written in any combination of one or more programming languages. These computer instructions may be provided to the computing unit 601 such that when executed by the computing unit 601, such as a processor, the computer instructions cause the execution of the steps involved in the embodiments of the methods of this application.

[0120] The computer-readable storage medium provided in this application can be a tangible medium that can contain or store computer instructions for performing the steps involved in the method embodiments of this application. The computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, and other forms of storage media.

[0121] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for identifying an avian influenza spillover risk zone based on the waterfowl-seabird interface, characterized in that, The method comprises the following steps: acquiring index data and environmental data of an avian influenza waterfowl-seabird host; wherein the index data comprises species distribution point data and seabird avian influenza outbreak point data of the avian influenza waterfowl-seabird host, the species distribution point data comprises species name, species classification, observation time and longitude and latitude coordinate information, the seabird avian influenza outbreak point data comprises virus type, outbreak country, outbreak time and longitude and latitude coordinate information, and the environmental data comprises two types of environmental variables, i.e. land and sea; monthly screening of the species distribution point data; judging whether the species distribution point data is sufficient according to a preset condition; if all the species distribution point data is sufficient, calculating an overflow risk index of the waterfowl-seabird based on an activity intensity method; otherwise, calculating the overflow risk index of the waterfowl-seabird based on a species richness method; identifying an avian influenza overflow risk area between the waterfowl-seabird based on the obtained overflow risk index.

2. The method according to claim 1, wherein the waterfowl-seabird interface avian influenza spillover risk zone identification is based on, The monthly screening of the species distribution point data comprises: eliminating the species distribution point data with missing or repeated longitude and latitude; excluding spatial outliers according to biogeographical boundaries; removing redundant points according to a grid resolution, and retaining only one species distribution point data in each grid pixel; counting the screened species distribution point data by month, constructing a species classification hierarchy, and obtaining the number of monthly species distribution points.

3. The method according to claim 2, wherein the waterfowl-seabird interface is identified by a waterfowl-seabird interface line. The judging whether the species distribution point data is sufficient according to a preset condition comprises: judging whether the number of monthly species distribution points is less than a specific value, if yes, it indicates that the species distribution point data is sufficient; otherwise, it indicates that the species distribution point data is insufficient.

4. The method for identifying avian influenza spillover risk areas based on the waterfowl-seabird interface according to claim 3, characterized in that, The calculating the overflow risk index of the waterfowl-seabird based on the activity intensity method if all the species distribution point data is sufficient comprises: based on the monthly species distribution point data and background points equal in number to the species distribution point data randomly generated outside the species distribution range as pseudo-missing samples, and corresponding environmental data as a training set, constructing monthly species distribution models of the avian influenza waterfowl host and the seabird host according to multiple machine learning models, and then generating monthly binary distribution maps of the avian influenza waterfowl host and the seabird host in each grid pixel; wherein the binary distribution map represents a distribution map of whether a certain species exists in a specific area in a specific month, and the monthly binary distribution map represents the number of months in which each waterfowl host species and seabird host species appears in each pixel; based on the monthly binary distribution map, activity intensity models of the avian influenza waterfowl host and the seabird host are constructed pixel by pixel, respectively, to obtain avian influenza waterfowl host activity entropy WFAE and avian influenza seabird host activity entropy SAE; wherein the activity entropy represents a quantitative indicator of bird activity intensity, and the higher the activity entropy, the greater the exposure risk of the waterfowl host and the seabird host to avian influenza; based on the avian influenza waterfowl host activity entropy WFAE and the avian influenza seabird host activity entropy SAE, the waterfowl-seabird overflow risk index is calculated pixel by pixel, and the formula is as follows: WSSRI=WFAE×SAE In the formula, WSSRI represents a waterbird-seabird spillover risk index, reflecting the degree of spatial co-occurrence and temporal overlap of avian influenza waterbird host species and seabird host species, and the higher the WSSRI, the higher the risk of avian influenza cross-border spillover between waterbirds and seabirds.

5. The method according to claim 4, wherein the waterfowl-seabird interface avian influenza spillover risk zone identification is based on the following factors: The monthly species distribution point data and the corresponding environmental data are used as a training set to construct monthly species distribution models of avian influenza waterbird hosts and seabird hosts according to multiple machine learning models, and then generate monthly binary distribution maps of avian influenza waterbird hosts and seabird hosts on each grid pixel, including: ​ The monthly species distribution point data and the corresponding environmental data are used as training data to construct monthly species distribution models of waterbird hosts and seabird hosts according to three mainstream machine learning methods, namely, generalized linear models, random forests, and maximum entropy models; The training data is randomly divided into five non-overlapping subsets, and five-fold cross-validation is used to optimize model parameters; The continuous prediction maps output by each monthly species distribution model are averaged pixel by pixel to generate average prediction maps; wherein, The continuous prediction map represents the probability of a certain waterbird or seabird host species existing in each grid pixel in a certain month, and the value ranges from 0 to 1 continuously. The average prediction map is obtained by algorithmically averaging the continuous prediction maps generated by multiple monthly species distribution models for the species. The best threshold value is determined based on TSS (True Skill Statistic) to convert the continuous average prediction map into a monthly binary distribution map; wherein, the best threshold value represents the critical value of the existence probability of a certain species in a certain month, that is, if the average prediction probability of a certain species in a certain grid pixel in a certain month is greater than or equal to the threshold value, it is determined that the species exists, otherwise it is determined that the species does not exist. The monthly binary distribution map is superimposed with the biogeographical boundary to limit the spatial prediction range of the model.

6. The method according to claim 5, wherein the waterfowl-seabird interface avian influenza spillover risk zone identification is based on the following factors: The best threshold value is determined based on TSS (True Skill Statistic) to convert the average prediction map into a monthly binary distribution map, specifically including: ​ Based on the range of the average prediction map, set the threshold value at a predetermined interval, calculate the confusion matrix at each threshold value, including true positives (actual existence and predicted existence), false positives (actual non-existence but predicted existence), false negatives (actual existence but predicted non-existence), and true negatives (actual non-existence and predicted non-existence); Based on the obtained confusion matrix, calculate TSS, which is described by the formula: TSS = Sensitivity + Specificity - 1, Sensitivity = TP / (TP + FN), Specificity = TN / (TN + FP); Wherein, TP represents true positive, FP represents false positive, TN represents true negative, and FN represents false negative. The TSS ranges from -1 to 1, and the higher the value, the stronger the model's ability to distinguish. Among all threshold values, the threshold value that maximizes the TSS value is selected as the best existence threshold value. According to the average prediction map, if the average prediction probability of a certain species in a certain month in a certain grid pixel is greater than or equal to the optimal existence threshold, it is determined that the species exists, and is assigned a value of 1; otherwise, it is determined that the species does not exist, and is assigned a value of 0, to obtain a monthly binary distribution map.

7. The method according to any one of claims 4-6, wherein the method is based on the identification of the risk zones of spillover of avian influenza from waterfowl-seabird interface, characterized in that, The activity intensity models of the avian influenza waterfowl host and the seabird host are respectively constructed based on the monthly binary distribution map, and the avian influenza waterfowl host activity entropy WFAE and the avian influenza seabird host activity entropy SAE are obtained, including: According to the monthly binary distribution map, the number of months in which each waterfowl host species and seabird host species appears in each pixel is counted; The residence time proportion of each host species in the pixel is calculated, that is, the proportion of the number of months in which the species appears to the total number of months in which the same group, that is, all waterfowl host species or all seabird host species, appears; According to the residence time proportion, the activity entropy of the waterfowl host and the seabird host is calculated respectively, and the calculation formula of the activity entropy is as follows: where AE represents the active entropy, dur i where the total months of stay of the i-th bird species in a grid cell is denoted by and the ratio of the total months of stay of the i-th bird species to the total months of stay of all bird species in a grid cell is denoted by 8. The method according to claim 3, wherein the waterfowl-seabird interface is identified by a waterfowl-seabird interface identification model. If part of the species distribution point data is insufficient, the waterfowl-seabird overflow risk index is calculated based on the species richness, including: The monthly species distribution point data is combined to obtain annual species distribution point data; based on the annual species distribution point data and background points equal in number to the species distribution points randomly generated outside the species distribution range as pseudo-missing samples, and the corresponding environmental variables as a training set, the annual species distribution models of the avian influenza waterfowl host and the seabird host are respectively constructed according to multiple machine learning models, and then the annual binary distribution maps of the avian influenza waterfowl host and the seabird host in each grid pixel are generated; wherein, The binary distribution map represents the distribution probability of whether a certain species exists in each grid pixel, and the annual binary distribution map represents the cumulative probability of each waterfowl host species and seabird host species in each pixel; Based on the binary distribution map, the binary distribution maps of the waterfowl host species and the seabird host species are superimposed respectively, and each grid pixel is accumulated and counted, and the avian influenza waterfowl host species richness model and the seabird host species richness model are respectively constructed pixel by pixel to obtain the avian influenza waterfowl host richness grid data and the avian influenza seabird host richness grid data; wherein, The richness represents the species diversity of each waterfowl host species and seabird host carrying the avian influenza virus in each pixel; Based on the avian influenza waterfowl host richness grid data and the avian influenza seabird host richness grid data, waterfowl-seabird species pairs are calculated pixel by pixel, and the waterfowl-seabird overflow risk index is calculated, which is described by the formula: In the formula, WSSRI represents the waterfowl-seabird overflow risk index, reflecting the ecological contact opportunity of the avian influenza waterfowl host species and the seabird host species, and the higher the WSSRI, the higher the risk of avian influenza cross-border overflow between waterfowl and seabirds.

9. The waterfowl-seabird interface bird flu spill-over risk zone identification method according to claim 4 or 8, characterized in that, Based on the obtained overflow risk index, the avian influenza overflow risk area between waterfowl and seabirds is identified, including: The seabird avian influenza occurrence point data is used as a verification data set; According to the verification data set, the WSSRI result is verified, the true positive rate and false positive rate of the WSSRI index at each threshold value are calculated, and the false positive rate and true positive rate corresponding to each threshold value are taken as coordinate points and plotted in a coordinate system to obtain an ROC curve of the false positive rate and the true positive rate with the threshold value of the overflow risk index, and an area under curve (AUC) value is obtained, which is used to verify whether the WSSRI index can effectively distinguish the true positive samples and the true negative samples; According to the ROC curve, a threshold value corresponding to a point with the maximum true positive rate and false positive rate is selected as the optimal threshold value of the waterfowl-seabird overflow risk index; wherein the waterfowl-seabird overflow risk index lower than the optimal threshold value is a non-high-risk area, and the waterfowl-seabird overflow risk index higher than the optimal threshold value is a high-risk area.

10. The method according to claim 9, wherein the waterfowl-seabird interface avian influenza spillover risk zone identification is based on, The waterfowl-seabird overflow risk index is calculated by using the following steps: According to the longitude and latitude of each case point and the longitude and latitude of the nearest natural water body, the Euclidean distance between each case point and the nearest natural water body is calculated; Based on the ecological dependence of the avian influenza seabird host on water area, a screening condition is set, and atypical habitat data is removed according to the screening condition; wherein, The screening condition includes only keeping case points within a first preset value from the coastline or case points within a second preset value from the nearest natural water body, so as to remove atypical habitat data, and the atypical habitat data represents case point data whose geographical position of the avian influenza outbreak point does not conform to the natural ecological habit records of the host species.

Images