An avian influenza exposure risk area identification method based on spatial and temporal distribution of water bird host species

Abstract

This invention discloses a method for identifying avian influenza exposure risk areas based on the spatiotemporal distribution of waterbird host species. The method for obtaining the waterbird host activity entropy includes: obtaining a list of waterbird hosts for avian influenza through multi-source data fusion; obtaining waterbird host distribution point data from public databases and filtering the distribution points; obtaining environmental data from the WorldClim database; removing highly collinear environmental data based on the Pearson correlation matrix; resampling to an appropriate spatial resolution; establishing a species distribution model to predict monthly species distribution; improving the Shannon index calculation method; and constructing a waterbird host activity entropy algorithm. The method for obtaining the high exposure risk threshold includes: obtaining avian influenza outbreak point data; plotting ROC curves; determining the high exposure risk threshold based on the Youden index; and evaluating the accuracy of risk prediction. The avian influenza exposure risk areas identified by this invention are accurate and efficient, have significant public health implications, and provide a scientific basis for the prevention and early warning of avian influenza viruses.

Description

Technical Field

[0001] This invention relates to the technical field of methods for identifying avian influenza exposure risk areas, specifically to a method for identifying avian influenza exposure risk areas based on the spatiotemporal distribution of waterbird host species. Background Technology

[0002] Avian influenza is an infectious disease of birds caused by influenza A viruses. It occurs and spreads in the environment, animals, and humans, exhibiting complete host diversity. As the natural host of avian influenza viruses, waterbirds play a crucial role in virus carriage, transmission, and mutation. Traditional methods for identifying avian influenza exposure risk areas mainly rely on long-term epidemiological surveys. The establishment of surveillance sites should prioritize the following high-risk areas: regions with a history of human or animal avian influenza outbreaks, areas with dense water bodies such as lakes and rivers, areas with concentrated waterfowl farming, and migratory bird habitats or migration routes. Surveillance sites should include live poultry markets, poultry farms, slaughterhouses, and wild bird habitats. Surveys should include collecting poultry contact history and fasting venous blood samples from occupationally exposed individuals, throat swabs from poultry and wild birds, feces and dead poultry tissue, as well as swabs from cages, slaughtering boards, and nearby water bodies in the external environment. The collected samples will be further used for virus isolation and subtype identification. Avian influenza virus (AIV) is a segmented, negative-sense, single-stranded RNA virus. Its replication and transcription rely on its own replicase and transcriptase, exhibiting a high error rate, which makes it prone to mutation during replication and transcription. Based on differences in its surface hemagglutinin protein (HA) and neuraminidase (NA), AIV can be classified into 18 HA subtypes and 11 NA subtypes. Virus isolation typically employs chicken embryo culture, followed by pathogen surveillance to further identify the HA and NA subtypes. This primarily includes antigen detection, serological detection, and molecular biological detection. With the rapid development of molecular biological detection, it plays an increasingly important role in AIV pathogen surveillance, such as quantitative real-time reverse transcription-polymerase chain reaction (RT-PCR), loop-mediated isothermal amplification, and gene chip technology. Through virus isolation and subtype identification, the carrier rate, seasonal fluctuations, and host diversity of AIV can be investigated, thereby enabling research on the variation, evolution, cross-species transmission, and pathogenicity of AIV. Furthermore, by employing reverse genetics site-directed mutagenesis and proteomics techniques, we revealed some key sites and mechanisms of action affecting the pathogenicity, transmissibility, and receptor binding ability of avian influenza viruses. While traditional methods for identifying avian influenza exposure risk areas play an important role in localized regions, they still have significant limitations. First, these methods require complex on-site sampling, sample processing, and experimental assays, demanding not only substantial human resources but also significant resource and financial support. This cost can place a considerable burden on public health and safety budgets, especially in long-term and large-scale surveillance projects. Second, pathogen surveillance relies not only on advanced instruments and complex analytical software but also on highly skilled operators. How to widely disseminate pathogen surveillance technology to primary healthcare units for its widespread application is also an urgent problem to be solved.Furthermore, the virus isolation and subtype identification process significantly increases the risk of avian influenza infection for monitoring personnel, and is time-consuming and lagging, failing to meet the needs of long-term and large-scale monitoring projects for real-time data and rapid response. Finally, traditional methods are limited to a finite geographical area, making it difficult to comprehensively identify a wider range of avian influenza exposure risk areas.

[0003] The transmission of avian influenza depends on exposure and contact between species. Against the backdrop of global change, the spatiotemporal distribution of species is undergoing significant shifts, further reshaping the spatiotemporal overlap between species and increasing the likelihood of viral emergence and sharing. To address this challenge, methods based on large-scale spatiotemporal prediction models of disease risk to simulate virus-sharing hotspots have emerged. Existing research uses bioclimatic envelope algorithms to predict mammalian species distribution, combines species habitat preferences to quantify habitat area, and constructs mammalian community composition in grid cells. Based on this, community-level reproduction rates are calculated to measure the invasion potential, persistence, and peak epidemic rate of pathogens in mammalian communities to assess global infectious disease risk patterns. Other studies simulate potential future virus-sharing hotspots based on mammalian species distribution under 2070 climate change and land-use scenarios. Bats, due to their unique transmission capabilities, dominate virus sharing and may share viruses with humans in the future. Although large-scale spatiotemporal prediction models of disease risk have made significant progress in simulating virus-sharing hotspots, current research focuses on mammals and pays less attention to the transmission patterns of avian influenza viruses in migratory birds. Due to their migratory nature, birds' distribution ranges cross national borders and even cover entire continents. Therefore, the risk of avian influenza exposure in birds must consider not only the host species richness resulting from spatial overlap but also the length and evenness of species residency time. This dynamic spatiotemporal overlap makes their virus-sharing patterns complex and unpredictable. However, existing large-scale spatiotemporal prediction models for disease risk are mostly based on static or local data and fail to fully consider the intra-annual variations in bird host migration behavior, leading to significant limitations in identifying avian influenza exposure risk areas.

[0004] With public health and safety issues becoming increasingly serious, there is an urgent need to accurately and efficiently identify areas of avian influenza exposure risk, as well as high-risk areas for occurrence and spread, on a global scale, so as to provide a scientific basis for public health and safety decision-making. Summary of the Invention

[0005] To address the above shortcomings, the present invention aims to propose a method for identifying avian influenza exposure risk areas by establishing a monthly species distribution model based on the spatiotemporal distribution of avian influenza waterbird host species, calculating the waterbird host activity entropy, and determining high exposure risk thresholds, thereby achieving a lower cost and more efficient and accurate approach. Specifically:

[0006] A method for identifying avian influenza exposure risk areas based on the spatiotemporal distribution of waterbird host species includes the following steps:

[0007] S1. Entropy acquisition of waterbird host activity;

[0008] S2. Determination of high exposure risk thresholds;

[0009] S3. Identify areas at risk of avian influenza exposure.

[0010] Furthermore, the acquisition of waterbird host activity entropy as described in S1 specifically involves:

[0011] (1) Obtain the list of waterbird hosts for avian influenza;

[0012] (2) Obtain and filter data on the distribution points of waterbird host species;

[0013] (3) Obtain environmental data and remove highly collinear environmental data based on the Pearson correlation matrix to obtain environmental covariates;

[0014] (4) Establish a monthly species distribution model: Randomly select a number of species non-occurrence points equal to the monthly species occurrence points of the waterbird host species to construct a point dataset. Utilize the geographical information of species occurrence points and non-occurrence points, combined with environmental covariates, and take the percentage of water body as the key driving factor to simulate the monthly species distribution of the waterbird host species on a global scale, and obtain the monthly species distribution model of the waterbird host species.

[0015] (5) By improving the Shannon index calculation method, a waterbird host activity entropy algorithm was constructed to quantify the exposure risk of avian influenza.

[0016] Furthermore, the criteria for obtaining the distribution points in step (2) are as follows:

[0017] ① Observational data with human and / or machine observations;

[0018] ② It can obtain species name, species classification, observation time and latitude and longitude coordinates;

[0019] The distribution point screening steps are as follows:

[0020] ①Remove observation records with empty or duplicate latitude and longitude coordinates;

[0021] ② Remove outliers based on biogeographical boundaries such as the Wallace Line and bird ranges to avoid including erroneous observation records;

[0022] ③ Remove redundant points according to the raster resolution, so that only one distribution point is retained in a raster cell;

[0023] ④ Remove species with fewer than 108 observation records and species with fewer than 9 monthly observation records.

[0024] Furthermore, the method for establishing the monthly species distribution model is as follows:

[0025] The optimal model was determined among the random forest model, the maximum entropy model, and the lightweight gradient booster model by using five-fold cross-validation to fine-tune the parameters.

[0026] Continuous predictions are converted to a binary presence-absence distribution using a 95% quantile threshold.

[0027] The accuracy of the monthly species distribution model is evaluated based on the area under the curve (AUC), thereby establishing the monthly species distribution model.

[0028] Furthermore, the waterbird host activity entropy is constructed using the following calculation model:

[0029]

[0030]

[0031] In the formula, The total number of months that the i-th type of bird stays in a given raster cell; This is 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.

[0032] Furthermore, the high exposure risk threshold is determined specifically as follows:

[0033] Data on avian influenza outbreak locations were obtained, ROC curves were plotted, high exposure risk thresholds were determined based on the Youden index, and the accuracy of risk prediction was assessed.

[0034] Furthermore, the avian influenza outbreak data includes virus type, country of occurrence, time of occurrence, and latitude and longitude information of the outbreak location; redundant points are removed according to the raster resolution so that only one avian influenza outbreak location is retained in a raster cell.

[0035] Furthermore, the ROC curve is a curve showing the change of the true positive rate and the false positive rate with a threshold; the Youden index is the difference between the true positive rate and the false positive rate.

[0036] Furthermore, the model's accuracy and sensitivity are determined using a confusion matrix to assess the accuracy of risk prediction; the formula is:

[0037]

[0038]

[0039] Among them, TP is a true positive, which is the number of samples correctly classified as avian influenza exposure risk areas; FP is a false positive, which is the number of samples that incorrectly classify non-avian influenza occurrence points as avian influenza exposure risk areas; TN is a true negative, which is the number of samples correctly classified as non-avian influenza exposure risk areas; and FN is a false negative, which is the number of samples that incorrectly classify avian influenza occurrence points as non-avian influenza exposure risk areas.

[0040] Furthermore, by calculating the activity entropy of waterbird hosts and determining the high exposure risk threshold, the spatial range where the activity entropy of waterbird hosts exceeds the threshold is identified as an avian influenza exposure risk area.

[0041] Beneficial Effects: This invention, based on the spatiotemporal distribution of avian influenza host species in waterbirds, effectively identifies avian influenza exposure risk areas by calculating the activity entropy of waterbird hosts and determining high exposure risk thresholds. Compared to traditional methods, this invention not only significantly reduces the required human and material resources and the risk of avian influenza infection for monitoring personnel, but also more accurately and efficiently identifies avian influenza exposure risk areas and potentially epidemic-prone areas on a large scale. In the context of increasingly serious public health and safety issues, this invention has significant public health implications, providing a scientific basis for the prevention and early warning of avian influenza viruses. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating the method for obtaining waterbird host activity entropy and high exposure risk threshold according to the present invention.

[0043] Figure 2 This is a schematic diagram of the monthly species distribution in the example;

[0044] Figure 3 The ROC curve and Yodden index points for Vietnam. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0046] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following drawings denote similar items; therefore, once an item is defined in one drawing, it need not be further discussed in subsequent drawings.

[0047] Unless otherwise specified, the technical terms in this specification have the same meaning as those generally understood by those skilled in the art; however, in case of any conflict, the definitions in this specification shall prevail.

[0048] The term "host" refers to an organism that provides a living environment for pathogens (such as bacteria, viruses, and parasites). Pathogens obtain the necessary nutrients and survival conditions by residing inside or on the surface of a host, which may cause the host to contract diseases, or even die in severe cases.

[0049] The term "host species" refers to the species of organism that provides a living and reproductive environment for pathogens (such as bacteria, viruses, and parasites), emphasizing the species identity of the host.

[0050] The term "waterbird" refers to birds that inhabit or frequently inhabit aquatic or wetland environments. According to taxonomy, waterbirds include all birds in the orders Loons, Grebes, Ciconiiformes, Anseriformes, and Charadriiformes, as well as some birds in the orders Pelecaniformes, Gruiformes, and Coraciiformes. Other orders also have a few birds adapted to aquatic or wetland environments.

[0051] The term "spatiotemporal overlap of host species" refers to the overlap in the time periods of two or more host species in the same location or nearby areas within a specific time and space, resulting in direct or indirect contact.

[0052] The term "avian influenza exposure risk" refers to the risk that an individual may become infected with the avian influenza virus through contact with a host carrying the virus or its secretions, excrement, etc., or through activities in an environment contaminated with the virus. Exposure risk is influenced by a variety of factors, including the species richness of the exposed host, the duration of exposure, and the evenness of exposure.

[0053] The term "avian influenza outbreak" refers to the widespread transmission of avian influenza virus in the environment, animals, and humans, causing disease and, in severe cases, resulting in a large number of infections and deaths, threatening public health and safety.

[0054] The term "species distribution model" (SDM), also known as an ecological niche model (ENM), is an important tool in ecology used to predict species distribution. Based on niche theory, a niche is the position a species occupies within an ecosystem and its functional relationships with the environment and other organisms, encompassing factors such as lifestyle, resource utilization, and behavioral habits. By analyzing location data of species presence and absence, and incorporating environmental covariates, species distribution models predict and simulate species distribution over a wider geographical area.

[0055] The term "entropy" is a parameter in thermodynamics that describes the state of matter; it is a measure of the disorder of a system. The higher the entropy of a system, the more uniform its energy distribution, meaning the more disordered the system.

[0056] The term "Shannon index," also known as the Shannon diversity index or Shannon-Weina index, is an ecological indicator used to measure the species diversity of a community. Its theoretical basis is Shannon information entropy. Proposed by Claude Shannon in 1948, Shannon information entropy is a concept in information theory used to quantify the uncertainty of information. The mathematical definition of Shannon information entropy is as follows:

[0057] Let X be a discrete random variable, whose possible values ​​are... , ,…, And the probability of each value is P(X= )= (where i = 1, 2, ..., n). The Shannon information entropy H(X) of a random variable X is defined as:

[0058]

[0059] in, It is a logarithmic function with base b. When b=2, the unit of information entropy is bits; when b=e, the unit is nats; and when b=1, the unit is Hartleys.

[0060] The higher the Shannon entropy value, the greater the uncertainty of the random variable, i.e., the more information it contains. Information entropy reaches its maximum when all possible events have equal probabilities, at which point the system is in a state of maximum uncertainty. Conversely, if the probability of one event is one and the probabilities of other events are zero, the information entropy is zero, indicating no uncertainty, i.e., zero information.

[0061] The Shannon index is a specific application of Shannon information entropy in the field of ecology. This index considers not only species richness (the number of species in a community) but also species evenness. The formula for calculating the Shannon index is:

[0062]

[0063] in, It is the proportion of the number of individuals of the i-th species out of the total number of individuals. It is a logarithmic function with the natural logarithm e as its base.

[0064] The Shannon index is positively correlated with community species diversity. A higher Shannon index indicates rich species diversity when species are evenly distributed in the community. Conversely, a lower Shannon index indicates low species diversity when a community is dominated by a single species with relatively few individuals of other species.

[0065] The term "ROC curve" is a graphical tool used to evaluate classification accuracy. It plots the true positive rate (TPR, i.e., sensitivity) versus the false positive rate (FPR, i.e., 1-specificity) as a function of a threshold, reflecting the model's ability to distinguish between positive and negative samples. The area under the curve (AUC) represents the model's overall performance; a higher AUC value indicates a better classification ability.

[0066] The term "Youden index" refers to the difference between the true positive rate (TPR) and the false positive rate (FPR), calculated as Sensitivity + Specificity – 1. The Youden index takes into account both sensitivity and specificity, and its maximum value corresponds to the optimal threshold point.

[0067] The term "confusion matrix," also known as the error matrix, is a commonly used evaluation method for assessing classification accuracy. It is represented by an n x n matrix. In this matrix, rows represent predicted classifications, and columns represent actual classifications. The confusion matrix consists of four parts: the number of samples correctly predicted as positive by the model (true positives, TP), the number of samples incorrectly predicted as positive by the model (false positives, FP), the number of samples correctly predicted as negative by the model (true negatives, TN), and the number of samples incorrectly predicted as negative by the model (false negatives, FN).

[0068] The embodiments of this application disclose a method for identifying avian influenza exposure risk areas based on the spatiotemporal distribution of waterbird host species, including the following steps:

[0069] S1. Entropy acquisition of waterbird host activity;

[0070] S2. Determination of high exposure risk thresholds;

[0071] S3. Identify areas at risk of avian influenza exposure.

[0072] In a specific implementation, the acquisition of the waterbird host activity entropy in S1 is specifically as follows:

[0073] (1) Obtain the list of waterbird hosts for avian influenza;

[0074] (2) Obtain and filter data on the distribution points of waterbird host species;

[0075] (3) Obtain environmental data and remove highly collinear environmental data based on the Pearson correlation matrix to obtain environmental covariates;

[0076] (4) Establish a monthly species distribution model: Randomly select a number of species non-occurrence points equal to the monthly species occurrence points of the waterbird host species to construct a point dataset. Utilize the geographical information of species occurrence points and non-occurrence points, combined with environmental covariates, and take the percentage of water body as the key driving factor to simulate the monthly species distribution of the waterbird host species within a certain range, and obtain the monthly species distribution model of the waterbird host species.

[0077] (5) By improving the Shannon index calculation method, a waterbird host activity entropy algorithm was constructed to quantify the exposure risk of avian influenza.

[0078] In one specific implementation, the list of waterbird hosts for avian influenza is obtained by integrating multi-source data, including on-site measurements, reviewing relevant scientific research reports, and accessing databases such as the Chinese Center for Disease Control and Prevention and the Global Initiative on Sharing All Influenza Data (GISAID), to comprehensively collect the list of avian influenza hosts. This list is then cross-matched with the list of waterbird hosts to obtain the list of waterbird hosts for avian influenza. It is understood that the list of waterbird hosts includes the Latin name, order, family, and genus classification of the waterbird hosts.

[0079] In one specific implementation, the distribution points of the waterbird host species are obtained through public databases such as GBIF and eBird. The selection of the waterbird host species distribution point data needs to meet the following conditions: (1) including observation records such as human observation and / or machine observation; (2) able to obtain species name, species classification, observation time and latitude and longitude coordinate information. In order to improve the accuracy of the model, after obtaining the above distribution point data, it is also necessary to clean (screen) the data. The distribution points are screened through the following steps: (1) remove observation records with empty latitude and longitude or duplicates; (2) remove outliers according to the Wallace line and biogeographical boundaries such as bird habitat range to avoid including erroneous observation records; (3) remove redundant points according to the raster resolution so that only one distribution point is retained in a raster cell; (4) remove species with fewer than 108 observation records and species months with fewer than 9 monthly observation records.

[0080] In one specific implementation, the environmental data is obtained from the WorldClim database, including climate variables and supplementary data such as altitude and water percentage; the climate variable data are taken as averages from 1970 to 2000, specifically including 19 climate variables, as shown in Table 1:

[0081] Table 1 Climate Variables

[0082]

[0083] Understandably, to avoid model overfitting, after selecting climate data, it is necessary to remove highly collinear environmental data based on the Pearson correlation matrix as environmental covariates for subsequent monthly species distribution models. Then, the data is resampled to an appropriate spatial resolution.

[0084] In one specific implementation, the method for establishing the monthly species distribution model is as follows:

[0085] The optimal model was determined among the random forest model, the maximum entropy model, and the lightweight gradient booster model by using five-fold cross-validation to fine-tune the parameters.

[0086] Continuous predictions are converted to a binary presence-absence distribution using a 95% quantile threshold.

[0087] The accuracy of the monthly species distribution model is evaluated based on the area under the curve (AUC), thereby establishing the monthly species distribution model.

[0088] In a further embodiment, the waterbird host activity entropy is constructed using the following calculation model:

[0089]

[0090]

[0091] In the formula, The total number of months that the i-th type of bird stays in a given raster cell; This is 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.

[0092] Waterbird host activity entropy reflects the host species richness formed by spatial overlap of species, the length of stay and evenness of each host species. The higher the waterbird host activity entropy, the greater the risk of avian influenza exposure.

[0093] In one specific implementation, the high exposure risk threshold is determined as follows:

[0094] Obtain data on avian influenza outbreak locations, plot ROC curves, determine high exposure risk thresholds based on the Youden index, and assess the accuracy of risk prediction.

[0095] It is understood that the avian influenza outbreak data should include virus type, country of occurrence, time of occurrence, and latitude and longitude information of the outbreak location. Redundant points are removed according to the raster resolution, so that only one avian influenza outbreak location is retained in each raster cell.

[0096] The ROC curve reflects the model's recognition ability by plotting the true positive rate (TPR, i.e., sensitivity) and false positive rate (FPR, i.e., 1-specificity) as a function of the threshold.

[0097] The Youden Index is the difference between the True Positive Rate (TPR) and the False Positive Rate (FPR), and is calculated as Sensitivity + Specificity – 1. The maximum value of the Youden Index corresponds to the high exposure risk threshold.

[0098] The accuracy of risk prediction is achieved by comparing the avian influenza exposure risk area with the avian influenza outbreak point, and verifying the accuracy through a confusion matrix.

[0099] The confusion matrix consists of four parts: the number of samples correctly classified as avian influenza exposure risk areas (True Positives, TP), the number of samples incorrectly classified as avian influenza exposure risk areas (False Positives, FP), the number of samples correctly classified as non-avian influenza exposure risk areas (True Negatives, TN), and the number of samples incorrectly classified as avian influenza exposure risk areas (False Negatives, FN). The accuracy of risk prediction is assessed based on the accuracy and sensitivity of the confusion matrix. The formula is as follows:

[0100]

[0101] Accuracy is the ratio of the number of correctly classified samples to the total number of samples, and it directly reflects the classification accuracy of all samples.

[0102]

[0103] Sensitivity refers to the proportion of samples that are correctly classified as avian influenza exposure risk areas out of all samples that are actually avian influenza outbreak sites. It can measure the accuracy of identifying avian influenza outbreak sites.

[0104] The identification of avian influenza exposure risk areas described in this application is based on the spatiotemporal distribution of avian influenza waterbird host species. A monthly species distribution model is established to predict the monthly species distribution. By calculating the waterbird host activity entropy and determining the high exposure risk threshold, the spatial range where the waterbird host activity entropy is higher than the threshold is identified as avian influenza exposure risk areas.

[0105] The present invention will now be described in detail with reference to embodiments and accompanying drawings. However, it should be understood that the embodiments and drawings are for illustrative purposes only and do not constitute any limitation on the scope of protection of the present invention. All reasonable modifications and combinations included within the inventive spirit of the present invention fall within the scope of protection of the present invention.

[0106] See attached document Figure 1 According to the technical solution of the present invention, in some specific embodiments, the method for obtaining the activity entropy of waterbird hosts may include:

[0107] S1 comprehensively collected a list of avian influenza hosts by integrating multi-source data, including on-site measurements, review of relevant scientific research reports, and access to databases such as the Chinese Center for Disease Control and Prevention and the Global Initiative on Sharing All Influenza Data (GISAID), and cross-matched this list with the list of waterbird hosts to obtain a list of waterbird hosts for avian influenza.

[0108] Specifically, this application uses the Indochina Peninsula as the region for which avian influenza exposure risk areas will be identified. The Indochina Peninsula is located in the heart of Southeast Asia, with Myanmar being the largest, accounting for 35% of the entire peninsula, followed by Thailand (27%), Vietnam (17%), Laos (12%), and Cambodia (9%). The aforementioned list of waterbird hosts includes: the Latin name of the waterbird host, and its order, family, and genus classification.

[0109] S2 obtains data on the distribution points of waterbird host species through public databases such as GBIF and eBird, and then filters the distribution points.

[0110] The data on the distribution points of waterbird host species must meet the following requirements: (1) include observation records such as human observation and machine observation; (2) be able to obtain species name, species classification, observation time and latitude and longitude coordinates.

[0111] The steps for screening distribution points include: (1) removing observation records with empty or duplicate latitude and longitude coordinates; (2) removing outliers based on biogeographical boundaries such as the Wallace Line and bird ranges to avoid including incorrect observation records; (3) removing redundant points based on raster resolution so that only one distribution point is retained in a raster cell; and (4) removing species with fewer than 108 observation records and species with fewer than 9 monthly observation records.

[0112] Specifically, data on the distribution points of waterbird host species were obtained, and the distribution points were filtered to obtain the final simulated list of waterbird hosts (Table 2).

[0113] Table 2 List of waterbird hosts in the Indochina Peninsula

[0114]

[0115] The results show that Charadriiformes make up the largest proportion of birds, approximately 52% of the total, and are highly diverse, with the European family (Europeans) leading the way, followed by sandpipers, plovers, quails, and rock plovers. Anseriformes are the second largest, accounting for about 14% of the total, but their diversity is limited, consisting only of ducks. Pelecaniformes are the same as Anseriformes, accounting for about 14% of the total, with herons leading the way, followed by ibises, pelicans, and snake pelicans. Boobies are the smallest, accounting for only 5% of the total, but their diversity is relatively rich, including cormorants, gannets, and frigatebirds. The remaining bird species are relatively few in number and relatively simple in species. Gruiformes accounts for about 5% of the total, including only rails and cranes; Strigiformes accounts for about 5% of the total, including only owls; Accipiformes accounts for about 2% of the total, including only eagles and ospreys; Grebes account for about 1% of the total, including only grebes; Ciconiiformes accounts for about 1% of the total, including only storks; and Procellariiformes accounts for about 1% of the total, including only somatic birds.

[0116] S3 acquires environmental data from the WorldClim database, including climate variables and supplementary data such as altitude and water percentage. Highly collinear environmental data are removed using the Pearson correlation matrix and used as environmental covariates in subsequent monthly species distribution models to avoid overfitting. The data is then resampled to an appropriate spatial resolution.

[0117] Specifically, Pearson correlation analysis was performed on 19 climate variables using the "ENMTools" tool in the r software. Four environmental data points were selected from these variables: temperature variation variance (BIO4), average temperature of the warmest season (BIO10), average annual precipitation (BIO12), and average precipitation of the coldest month (BIO19) to avoid overfitting of the model.

[0118] S4 establishes a monthly species distribution model, randomly selecting an equal number of species absence points to the monthly species presence points of the waterbird host species to construct a point dataset. Using the geographical information of species presence and absence points, combined with environmental covariates, and adding the percentage of water body as a key driving factor, the monthly species distribution of the waterbird host species within a certain range is simulated.

[0119] The monthly species distribution model employed three modeling approaches: Random Forest (RF), Maximum Entropy (MaxEnt), and Lightweight Gradient Boosting Machine (LightGBM). Parameters were fine-tuned using five-fold cross-validation to select the optimal model. Furthermore, a 95th percentile threshold was used to convert continuous predictions into a binary presence-absence distribution, and the accuracy of the species distribution model was evaluated based on the area under the curve (AUC).

[0120] For details, please refer to the appendix. Figure 2 Based on monthly grouping of waterbird host species distribution point data, a species distribution model is established to predict monthly species distribution, and then host residence time is obtained through spatiotemporal statistics.

[0121] S5 improves the Shannon index calculation method and constructs a waterbird host activity entropy algorithm to quantify the exposure risk of avian influenza.

[0122] The waterbird host activity entropy is obtained through the following calculation model:

[0123]

[0124]

[0125] In the formula:

[0126] The total number of months that the i-th type of bird stays in a given raster cell.

[0127] This is 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.

[0128] Specifically, the risk of avian influenza exposure is quantified using the activity entropy of waterbird hosts.

[0129] Furthermore, methods for determining the high exposure risk threshold may include:

[0130] S6 acquires data on avian influenza outbreak locations, plots ROC curves, determines high exposure risk thresholds based on the Youden index, and assesses the accuracy of risk prediction.

[0131] The data on avian influenza outbreak locations should include virus type, country of occurrence, time of occurrence, and latitude and longitude information of the location. Redundant points are removed based on raster resolution, ensuring that only one avian influenza outbreak location is retained per raster cell.

[0132] The ROC curve reflects the model's recognition ability by plotting the true positive rate (TPR, i.e., sensitivity) and false positive rate (FPR, i.e., 1 - specificity) as a function of a threshold. The Youden index is the difference between the true positive rate (TPR) and the false positive rate (FPR), and is calculated as sensitivity + specificity - 1. The maximum value of the Youden index corresponds to the high exposure risk threshold.

[0133] S7. Based on the identified high exposure risk threshold (entropy value) of waterbird host activity entropy, areas with entropy values ​​higher than the threshold are identified and designated as avian influenza exposure risk areas.

[0134] Figure 3 For the ROC curve and Youden index points of Vietnam, from Figure 3 It can be seen that the high exposure risk threshold corresponding to the Yoden Index point in Vietnam is 4.308, that is, areas where the waterbird host activity entropy is greater than 4.308 are avian influenza exposure risk areas.

[0135] The risk prediction accuracy is determined by comparing avian influenza exposure risk areas with avian influenza outbreak locations, and verifying the accuracy using a confusion matrix. The confusion matrix consists of four parts: the number of samples correctly classified as avian influenza exposure risk areas (True Positives, TP), the number of samples incorrectly classified as non-avian influenza outbreak locations (False Positives, FP), the number of samples correctly classified as non-avian influenza exposure risk areas (True Negatives, TN), and the number of samples incorrectly classified as avian influenza outbreak locations (False Negatives, FN). The risk prediction accuracy is then evaluated based on accuracy and sensitivity using the confusion matrix. The formula is as follows:

[0136]

[0137] Accuracy is the ratio of the number of correctly classified samples to the total number of samples, and it directly reflects the classification accuracy of all samples.

[0138]

[0139] Sensitivity refers to the proportion of samples that are correctly classified as avian influenza exposure risk areas out of all samples that are actually avian influenza outbreak sites. It can measure the accuracy of identifying avian influenza outbreak sites.

[0140] Specifically, the high exposure risk thresholds, accuracy, sensitivity, and exposure risk area parameters for various countries in the Indochina Peninsula are shown in Table 3.

[0141] Table 3. Parameters of various countries in the Indochina Peninsula

[0142]

[0143] The results above indicate that the high-exposure risk threshold for Vietnam is 4.308, with an accuracy of 86.28%, a sensitivity of 86.28%, and an exposure risk area of ​​126,525.96 km². 2 The high-exposure risk threshold for Cambodia is 4.237, with an accuracy of 85.12%, a sensitivity of 92.75%, and an exposure risk area of ​​101,612.90 km². 2 The high exposure risk threshold for Myanmar is 4.371, with an accuracy of 80.60%, a sensitivity of 93.28%, and an exposure risk area of ​​219,743.28 km². 2 The high exposure risk threshold for Laos is 3.820, with an accuracy of 43.21%, a sensitivity of 64.20%, and an exposure risk area of ​​168,720.00 km². 2 The high-exposure risk threshold for Thailand is 4.266, with an accuracy of 66.35%, a sensitivity of 66.65%, and an exposure risk area of ​​120,011.72 km². 2 .

[0144] Specifically, based on the spatiotemporal distribution of avian influenza host species in waterbirds, a monthly species distribution model is established to predict the monthly species distribution. By calculating the activity entropy of waterbird hosts and determining the high exposure risk threshold, avian influenza exposure risk areas are identified.

[0145] The results above show that in Vietnam, high-risk areas are mainly distributed along the eastern coastline, while low-risk areas are mainly located in northern Vietnam. Most of Laos is high-risk, with low-risk areas primarily in the north. Cambodia is high-risk in the west and low-risk in the east. In Thailand, high-risk areas are mainly distributed along the southern coastline, while most low-risk areas are in the north. In Myanmar, high-risk areas are mainly in the south, while low-risk areas are mainly in the north.

[0146] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for identifying avian influenza exposure risk areas based on the spatiotemporal distribution of waterbird host species, characterized in that, Includes the following steps: S1. Entropy acquisition of waterbird host activity; S2. Determination of high exposure risk threshold; S3. Identify areas at risk of avian influenza exposure; Specifically, the acquisition of the waterbird host activity entropy described in S1 is as follows: (1) Obtain the list of waterbird hosts for avian influenza; (2) Obtain and filter waterbird host species distribution point data: This is obtained by dividing the acquired data into 12 monthly distribution point datasets of waterbird host species according to the month label of the observation time, from 1 to 12; (3) Obtain environmental data and remove highly collinear environmental data based on the Pearson correlation matrix to obtain environmental covariates; (4) Establish a monthly species distribution model: Randomly select a number of species non-occurrence points equal to the monthly species occurrence points of the waterbird host species to construct a point dataset. Utilize the geographical information of species occurrence points and non-occurrence points, combined with environmental covariates, and take the percentage of water body as the key driving factor to simulate the monthly species distribution of the waterbird host species on a global scale, and obtain the monthly species distribution model of the waterbird host species. (5) By improving the Shannon index calculation method, a waterbird host activity entropy algorithm was constructed to quantify the exposure risk of avian influenza; The method for establishing the monthly species distribution model is as follows: The optimal model was determined among the random forest model, the maximum entropy model, and the lightweight gradient booster model by using five-fold cross-validation to fine-tune the parameters. Continuous predictions are converted to a binary presence-absence distribution using a 95% quantile threshold. The accuracy of the species distribution model is evaluated based on the area under the curve (AUC), thereby establishing a monthly species distribution model. The waterbird host activity entropy was constructed using the following calculation model: In the formula, dur i pi is the total number of months that the i-th type of bird stays in a certain raster cell; pi is the ratio of the total number of months that the i-th type of bird stays in a certain raster cell to the total number of months that all birds stay. By calculating the waterbird host activity entropy and determining the high exposure risk threshold, the spatial range where the waterbird host activity entropy is higher than the threshold is identified as an avian influenza exposure risk area.

2. The identification method according to claim 1, characterized in that, The criteria for obtaining the distribution points in step (2) are as follows: ① Observational data with human and / or machine observations; ② It can obtain species name, species classification, observation time and latitude and longitude coordinates; The distribution point screening steps are as follows: ①Remove observation records with empty or duplicate latitude and longitude coordinates; ② Remove outliers based on biogeographical boundaries such as the Wallace Line and bird ranges to avoid including erroneous observation records; ③ Remove redundant points according to the raster resolution, so that only one distribution point is retained in a raster cell; ④ Remove species with fewer than 108 observation records and species with fewer than 9 monthly observation records.

3. The identification method according to claim 1, characterized in that, The high exposure risk threshold is specifically determined as follows: Data on avian influenza outbreak locations were obtained, ROC curves were plotted, high exposure risk thresholds were determined based on the Youden index, and the accuracy of risk prediction was assessed.

4. The identification method according to claim 3, characterized in that, The avian influenza outbreak data includes virus type, country of occurrence, time of occurrence, and latitude and longitude information of the outbreak location; redundant points are removed according to the raster resolution so that only one avian influenza outbreak location is retained in a raster cell.

5. The identification method according to claim 3, characterized in that, The ROC curve represents the changes in the true positive rate and the false positive rate as a function of a threshold; the Youden index represents the difference between the true positive rate and the false positive rate.

6. The identification method according to claim 3, characterized in that, The accuracy and sensitivity of the model are determined using a confusion matrix to assess the accuracy of risk prediction; the formula is as follows: Among them, TP is a true positive, which is the number of samples correctly classified as avian influenza exposure risk areas; FP is a false positive, which is the number of samples that incorrectly classify non-avian influenza occurrence points as avian influenza exposure risk areas; TN is a true negative, which is the number of samples correctly classified as non-avian influenza exposure risk areas; and FN is a false negative, which is the number of samples that incorrectly classify avian influenza occurrence points as non-avian influenza exposure risk areas.

Images