A method and device for detecting water ecological stress based on machine learning

Abstract

The present invention provides a method and device for detecting water ecological pressure based on machine learning. The method includes obtaining an aquatic organism monitoring data set, processing the aquatic organism monitoring data set to obtain a sample-species matrix; obtaining functional traits and pedigree information of aquatic organisms involved in the sample-species matrix, generating a species information matrix corresponding to the sample-species matrix, and generating a sample-environmental factor matrix corresponding to the sample-species matrix; determining a diversity index corresponding to a study area, the diversity index including a taxonomic diversity index, a functional diversity index, and a pedigree diversity index, and generating a sample-diversity index matrix corresponding to the sample-species matrix; constructing a structural equation model, determining the action path of each first influencing factor to be analyzed on the water ecology according to the ranking; and determining the water ecological pressure of the study area based on the action path of each first influencing factor to be analyzed on the water ecology.

Description

Technical Field

[0001] The present invention relates to the field of water resources technology, and in particular to a water ecological pressure detection method and device based on machine learning. Background Art

[0002] The health of aquatic ecosystems is a key indicator of water environment quality and is influenced by both natural conditions and human activities. With the rapid development of industrialization and urbanization, aquatic ecosystems are facing increasing pressures, including increased pollution loads, habitat destruction, and biodiversity loss. Therefore, accurately diagnosing aquatic ecosystem stress is crucial for developing effective water environment protection and restoration measures. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes a water ecological stress detection method and device based on machine learning.

[0004] According to a first aspect of the present invention, a method for detecting water ecological stress based on machine learning is provided, the method comprising the following steps:

[0005] Step S1: Acquire an aquatic organism monitoring dataset, and process the aquatic organism monitoring dataset to obtain a sample point-species matrix, wherein the sample point-species matrix represents the observed quantity information or existence information of all aquatic organism species monitored in the study area at each sampling point;

[0006] Step S2: Acquire functional traits and pedigree information of aquatic organisms involved in the sample point-species matrix, and generate a species information matrix corresponding to the sample point-species matrix, wherein each aquatic organism in the species information matrix corresponds to the numerical value of its functional traits in morphology, ecology, and habits, as well as Linnaean biological classification information; Based on the sample point-species matrix, determine the pressure factor of each sampling point, wherein the pressure factor is a plurality of influencing factors related to human influence, natural environment, and spatial distance, and generate a sample point-environment factor matrix corresponding to the sample point-species matrix;

[0007] Step S3: Based on the sample site-species matrix, the species information matrix and the aquatic organism monitoring dataset, determining the diversity index corresponding to the study area, wherein the diversity index includes a taxonomic diversity index, a functional diversity index and a phylogenetic diversity index, and generating a sample site-diversity index matrix corresponding to the sample site-species matrix;

[0008] Step S4: Calculate the Euclidean distance matrix of each diversity index vector in the sample point-diversity index matrix and each pressure factor vector in the sample point-environmental factor matrix, perform Mantel analysis between the two types of distance matrices, and determine several influencing factors with an impact greater than a preset threshold as the first influencing factors to be analyzed; use the random forest algorithm to rank the importance of the first influencing factors to be analyzed; construct a structural equation model, and determine the action path of each first influencing factor to be analyzed on the water ecology according to the ranking;

[0009] Step S5: Determine the water ecological pressure of the study area based on the action paths of each first influencing factor to be analyzed on the water ecology.

[0010] Preferably, in step S2, the functional traits and pedigree information of the aquatic organisms involved in the sample-species matrix are obtained, that is, the sample-species matrix D is obtained. m,n The functional traits and pedigree information of n aquatic organisms involved in the project include: obtaining Linnaean taxonomic information of n aquatic organisms, including their phylum, class, order, family, genus, species information and corresponding unique Latin name; obtaining 9 types of species functional traits of n aquatic organisms, including morphological characteristics, ecological characteristics, and habits. Morphological characteristics include exoskeleton status, breathing mode, body size, and body shape. Ecological characteristics include substrate selection, lifestyle, and functional feeding groups. Habits include refuge and chemotaxis. Corresponding values are assigned to different biological traits.

[0011] Preferably, the taxonomic diversity index includes species richness index, Shannon diversity index, Simpson diversity index, and Pielou evenness index.

[0012] Preferably, the calculation formula of the species richness index is:

[0013] q i =S i

[0014] Among them, q i is the species richness of the i-th sampling point, S i is the number of aquatic species observed at sampling point i, corresponding to the sample point-species matrix D m,n The number of aquatic species with a species abundance greater than 0 in the sampling point corresponding to the sampling point;

[0015] The calculation formula of the Shannon diversity index is:

[0016] p ij =n ij / N i

[0017]

[0018] N 1i =exp(H i )

[0019] Among them, N 1i is the Shannon diversity index of sampling point i, H i is the Shannon entropy index of sampling point i, p ij is the relative abundance of aquatic species j at sampling point i, q i is the species richness of sampling point i, n ij is the abundance of aquatic species j at sampling point i, N i is the total abundance of all aquatic species in sampling site i;

[0020] The calculation formula of Simpson's diversity index is:

[0021]

[0022] Among them, D' i is the Simpson diversity index of sampling point i, S i is the number of aquatic species in sampling point i, p ij is the relative abundance of aquatic species j at sampling site i;

[0023] The calculation formula of Pielou evenness index is:

[0024]

[0025] J i =H i / H i max

[0026] Among them, q i is the species richness of sampling point i, H i is the Shannon entropy index of sampling point i, H i max is the maximum value of the Shannon entropy index when the abundance of all aquatic species in sampling point i is consistent, J i is the Pielou uniformity index of sampling point i.

[0027] Preferably, the functional diversity index includes functional richness FRic, functional evenness FEve, functional differentiation FDiv and functional dispersion FDis, wherein:

[0028] The functional richness FRic corresponding to each sampling point was calculated using the Quickhull algorithm;

[0029] The calculation formula of the functional uniformity FEve is:

[0030]

[0031] Among them, EW li is the weighted uniformity of the aquatic organism set at sampling point i, l is a branch of the minimum spanning tree consisting of all functional sites of aquatic species in the multidimensional functional convex hull corresponding to the sampling point, a and b are the aquatic species involved in branch l, dist(a,b) is the Euclidean distance between the functional sites of aquatic species a and aquatic species b, w a 、w b is the relative abundance of aquatic species a and aquatic species b, PEW li is the partial weighted uniformity of sampling point i, S i is the number of aquatic species observed at sampling point i, FEve i is the functional uniformity of sampling point i;

[0032] The calculation formula of the functional differentiation FDiv is:

[0033]

[0034]

[0035] Among them, g ik is the coordinate of the center of gravity of the multidimensional functional convex hull of the aquatic organism set corresponding to sampling point i on functional trait k, V i is the number of aquatic species that constitute the vertices of the multidimensional functional convex hull at sampling point i, x ijk is the coordinate of aquatic species j on functional trait k at sampling point i, dG ij is the Euclidean distance from the functional space position of aquatic species j in sampling point i to the center of gravity of the multidimensional functional convex hull, is the average distance from all aquatic species in sampling point i to the centroid of the multidimensional functional convex hull, Δd i is the abundance-weighted deviation of all aquatic species at sampling point i from the center of gravity of the multidimensional functional convex hull, Δ|d i | is the absolute abundance weighted deviation of all aquatic species at sampling point i from the centroid, p ij is the relative abundance of aquatic species j at sampling site i, T i is the number of functional trait categories in sampling point i; S i is the number of aquatic species in sampling point i, FDiv i is the functional differentiation FDiv index corresponding to aquatic organisms in sampling point i;

[0036] The calculation formula of the functional dispersion FDis is:

[0037]

[0038] Among them, ci is the weighted centroid of the multidimensional space composed of the functional traits of all aquatic species at sampling point i, c ik is the position of the weighted centroid of the multidimensional space of the functional trait in sampling point i on the functional trait k, a ij is the abundance of aquatic species j at sampling site i, x ijt is the trait attribute of functional trait k of aquatic species j at sampling site i, z ij is the distance from aquatic species j at sampling point i to the centroid of the multidimensional space, FDis i is the functional discreteness of sampling point i.

[0039] Preferably, the calculation formula of the pedigree diversity index is:

[0040] PD i =(N′ i -1)+P i

[0041] Among them, N' i is the number of biological taxa participating in the phylogenetic tree at sampling point i, P i is the number of internal nodes of the minimum spanning path of the phylogenetic tree of all aquatic species spanning sampling point i, PD i is the phylogenetic diversity index of sampling point i, and the phylogenetic tree refers to a tree-like branch diagram reflecting the phylogenetic relationship determined by the Linnaean classification method for all aquatic organisms in the basin.

[0042] Preferably, in step S4, the Mantel analysis is performed using the Pearson correlation coefficient method, the significance of the correlation coefficient is obtained by t-test, and the calculation formula of the correlation coefficient is as follows:

[0043]

[0044] Where r xy is the Pearson correlation coefficient between the pressure factor x and the diversity index y, COV xy is the covariance between pressure factors and diversity index, SD x is the standard deviation of the pressure factor x, SD y is the standard deviation of the diversity index y; the value of the correlation coefficient is used as the degree of influence.

[0045] According to a second aspect of the present invention, a device for detecting water ecological stress based on machine learning is provided, the device comprising:

[0046] Initialization module: configured to obtain an aquatic organism monitoring dataset, process the aquatic organism monitoring dataset to obtain a sample point-species matrix, wherein the sample point-species matrix represents the observed quantity information or existence information of all aquatic organism species monitored in the study area at each sampling point;

[0047] The first matrix module is configured to obtain the functional traits and pedigree information of the aquatic organisms involved in the sample point-species matrix, generate a species information matrix corresponding to the sample point-species matrix, wherein each aquatic organism in the species information matrix corresponds to the numerical value of its functional traits in morphology, ecology, and habits, as well as Linnaean biological classification information; based on the sample point-species matrix, determine the pressure factor of each sampling point, wherein the pressure factor is a plurality of influencing factors related to human influence, natural environment, and spatial distance, and generate a sample point-environment factor matrix corresponding to the sample point-species matrix;

[0048] A second matrix module is configured to determine the diversity index corresponding to the study area based on the sample-species matrix, the species information matrix and the aquatic organism monitoring dataset, wherein the diversity index includes a taxonomic diversity index, a functional diversity index and a phylogenetic diversity index, and generate a sample-diversity index matrix corresponding to the sample-species matrix;

[0049] The first analysis module is configured to calculate the Euclidean distance matrix of each diversity index vector in the sample point-diversity index matrix and each pressure factor vector in the sample point-environmental factor matrix, and perform Mantel analysis between the two types of distance matrices to determine a number of influencing factors with an impact greater than a preset threshold as first influencing factors to be analyzed; use a random forest algorithm to rank the importance of the first influencing factors to be analyzed; construct a structural equation model to determine the action path of each first influencing factor to be analyzed on the water ecology according to the ranking;

[0050] The second analysis module is configured to determine the water ecological pressure of the study area based on the action path of each first influencing factor to be analyzed on the water ecology.

[0051] According to a third aspect of the present invention, there is provided an electronic device, comprising:

[0052] A processor, which is used to execute multiple instructions;

[0053] A memory for storing a plurality of instructions;

[0054] The plurality of instructions are used to be stored by the memory and loaded and executed by the processor to implement the method as described above.

[0055] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein a plurality of instructions are stored in the storage medium; the plurality of instructions are used for a processor to load and execute the aforementioned method.

[0056] The present invention has the following technical effects:

[0057] The diagnostic method established in this application can achieve unified and reasonable pressure diagnosis for all river basins; at the same time, the diagnostic method proposed in this application takes into account the monitoring data of all biological groups and considers the impact of different species, making the pressure diagnosis results of aquatic organisms more accurate.

[0058] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The accompanying drawings, which constitute part of the present invention, are used to provide a further understanding of the present invention. The present invention is described with the following accompanying drawings. In the accompanying drawings:

[0060] Figure 1 This is a flow chart of the water ecological stress detection method based on machine learning of the present invention;

[0061] Figure 2 This is a schematic diagram of the architecture of the water ecological stress detection method based on machine learning of the present invention;

[0062] Figure 3 This is a structural block diagram of the water ecological pressure detection device based on machine learning in the present invention. DETAILED DESCRIPTION

[0063] First combine Figure 1-Figure 2 The present invention describes a method for detecting water ecological stress based on machine learning in accordance with an embodiment of the present invention. Figure 1-Figure 2 As shown, the method includes the following steps:

[0064] Step S1: Acquire an aquatic organism monitoring dataset, and process the aquatic organism monitoring dataset to obtain a sample point-species matrix, wherein the sample point-species matrix represents the observed quantity information or existence information of all aquatic organism species monitored in the study area at each sampling point;

[0065] Step S2: Acquire functional traits and pedigree information of aquatic organisms involved in the sample point-species matrix, and generate a species information matrix corresponding to the sample point-species matrix, wherein each aquatic organism in the species information matrix corresponds to the numerical value of its functional traits in morphology, ecology, and habits, as well as Linnaean biological classification information; Based on the sample point-species matrix, determine the pressure factor of each sampling point, wherein the pressure factor is a plurality of influencing factors related to human influence, natural environment, and spatial distance, and generate a sample point-environment factor matrix corresponding to the sample point-species matrix;

[0066] Step S3: Based on the sample site-species matrix, the species information matrix and the aquatic organism monitoring dataset, determining the diversity index corresponding to the study area, wherein the diversity index includes a taxonomic diversity index, a functional diversity index and a phylogenetic diversity index, and generating a sample site-diversity index matrix corresponding to the sample site-species matrix;

[0067] Step S4: Calculate the Euclidean distance matrix of each diversity index vector in the sample point-diversity index matrix and each pressure factor vector in the sample point-environmental factor matrix, perform Mantel analysis between the two types of distance matrices, and determine several influencing factors with an impact greater than a preset threshold as the first influencing factors to be analyzed; use the random forest algorithm to rank the importance of the first influencing factors to be analyzed; construct a structural equation model, and determine the action path of each first influencing factor to be analyzed on the water ecology according to the ranking;

[0068] Step S5: Determine the water ecological pressure of the study area based on the action paths of each first influencing factor to be analyzed on the water ecology.

[0069] In the present invention, the sample-species matrix provides information for collecting biological traits and pedigree information, collecting sample pressure factor data, and calculating biodiversity index.

[0070] Obtain aquatic biological monitoring data sets and organize them into sample-species matrix D m,n , where m is the number of water bodies (sampling points) involved in the study area, n is the number of aquatic species monitored, and each value d in the sample point-species matrix i,j Represents every m in the i-th sampling point 3 There is j-th aquatic organism d i,j indivual.

[0071] The step S2 is to obtain the functional traits and pedigree information of the aquatic organisms involved in the sample point-species matrix, that is, to obtain the sample point-species matrix D m,nThe functional traits and pedigree information of n aquatic organisms involved in the project include: obtaining Linnaean taxonomic information of n aquatic organisms, including their phylum, class, order, family, genus, species information and corresponding unique Latin name; obtaining 9 types of species functional traits of n aquatic organisms, including morphological characteristics, ecological characteristics, and habits. Morphological characteristics include exoskeleton status, breathing mode, body size, and body shape. Ecological characteristics include substrate selection, lifestyle, and functional feeding groups. Habits include refuge and chemotaxis. Corresponding values are assigned to different biological traits.

[0072] In the present invention, the assignment rules are shown in Table 1. The acquired species information is calculated according to the matrix D m,n The n species are sorted in order to obtain the species information dataset T n,16 , where T n,16 A dataset consisting of 7 columns of taxonomic information and 9 types of traits for n species is constructed to construct a species information matrix.

[0073] Table 1 Functional trait assignment

[0074]

[0075]

[0076] In step S2, based on the sample point-species matrix, the pressure factor of each sampling point is determined, where the pressure factor is a plurality of influencing factors related to human influence, natural environment, and spatial distance, and a sample point-environment factor matrix corresponding to the sample point-species matrix is generated, wherein:

[0077] The influencing factors related to human impact include land use factors and water quality factors; the influencing factors related to the natural environment include hydrological factors and climate factors; and the influencing factors related to spatial distance include latitude and longitude factors and elevation data factors.

[0078] In the present invention, the sample-species matrix D is obtained m,n The environmental information of the m water bodies (sampling points) in the evaluation includes: obtaining information on the degree of human impact of the m sampling points, including land use data and water quality data; obtaining natural environmental data of the sampling points, including hydrological data and climate data; obtaining geographic spatial data of the sampling points, including latitude, longitude and elevation data. The environmental information obtained is calculated according to the matrix D m,n The m sample points are sorted in sequence to obtain the sample-environmental factor matrix E m,e , E m,e Contains e environmental data of m sample points.

[0079] Step S3: Based on the sample-species matrix, the species information matrix, and the aquatic organism monitoring dataset, determining the diversity index corresponding to the study area, wherein the diversity index includes a taxonomic diversity index, a functional diversity index, and a phylogenetic diversity index, and generating a sample-diversity index matrix corresponding to the sample-species matrix, wherein:

[0080] The taxonomic diversity index includes species richness index (Richness), Shannon diversity index (Shannon), Simpson diversity index (Simpson), and Pielou evenness index;

[0081] The calculation formula of the species richness index is:

[0082] q i =S i

[0083] Among them, q i is the species richness of the i-th sampling point, S i is the number of aquatic species observed at sampling point i, corresponding to the sample point-species matrix D m,n The number of aquatic species whose species abundance (number of species individuals) is greater than 0 in the sampling point corresponding to the sampling point;

[0084] The calculation formula of the Shannon diversity index is:

[0085] p ij =n ij / N i

[0086]

[0087] N 1i =exp(H i )

[0088] Among them, N 1i is the Shannon diversity index of sampling point i, H i is the Shannon entropy index of sampling point i, p ij is the relative abundance of aquatic species j at sampling point i, q i is the species richness of sampling point i, n ij is the abundance of aquatic species j at sampling point i, N i is the total abundance of all aquatic species in sampling site i;

[0089] The calculation formula of Simpson's diversity index is:

[0090]

[0091] Among them, D'i is the Simpson diversity index of sampling point i, S i is the number of aquatic species in sampling point i, p ij is the relative abundance of aquatic species j at sampling site i;

[0092] The calculation formula of Pielou evenness index is:

[0093]

[0094] J i =H i / H i max

[0095] Among them, q i is the species richness of sampling point i, H i is the Shannon entropy index of sampling point i, H i max is the maximum value of the Shannon entropy index when the abundance of all aquatic species in sampling point i is consistent, J i is the Pielou uniformity index of sampling point i.

[0096] In the present invention, species richness is a quantitative index that describes the species richness in a community; the Shannon diversity index measures the number of species in a community and the uniformity of species frequency distribution. The Shannon diversity index increases with the increase in the number of species in the community. When the number of species and species abundance are the same, the index reaches a maximum value; the Simpson diversity index is used to measure species richness and is represented by the probability that two randomly selected samples from the community belong to the same species (in order to make the index size positively correlated with species richness, the present invention uses a modified form of the inverse Simpson diversity index as a representative of the Simpson diversity index); the Pielou evenness index reflects the uniformity of species abundance in a community and is calculated from the Shannon entropy index and species richness.

[0097] The functional diversity index includes functional richness FRic, functional evenness FEve, functional differentiation FDiv and functional dispersion FDis, where:

[0098] The functional richness FRic corresponding to each sampling point was calculated using the Quickhull algorithm;

[0099] The calculation formula of the functional uniformity FEve is:

[0100]

[0101] Among them, EW liis the weighted uniformity of the aquatic organism set at sampling point i, l is a branch of the minimum spanning tree consisting of all functional sites of aquatic species in the multidimensional functional convex hull corresponding to the sampling point, a and b are the aquatic species involved in branch l, dist(a,b) is the Euclidean distance between the functional sites of aquatic species a and aquatic species b, w a 、w b is the relative abundance of aquatic species a and aquatic species b, PEW li is the partial weighted uniformity of sampling point i, S i is the number of aquatic species observed at sampling point i, FEve i is the functional uniformity of sampling point i;

[0102] The calculation formula of the functional differentiation FDiv is:

[0103]

[0104]

[0105] Among them, g ik is the coordinate of the center of gravity of the multidimensional functional convex hull of the aquatic organism set corresponding to sampling point i on functional trait k, V i is the number of aquatic species that constitute the vertices of the multidimensional functional convex hull at sampling point i, x ijk is the coordinate of aquatic species j on functional trait k at sampling point i, dG ij is the Euclidean distance from the functional space position of aquatic species j in sampling point i to the center of gravity of the multidimensional functional convex hull, is the average distance from all aquatic species in sampling point i to the centroid of the multidimensional functional convex hull, Δd i is the abundance-weighted deviation of all aquatic species at sampling point i from the center of gravity of the multidimensional functional convex hull, Δ|d i | is the absolute abundance weighted deviation of all aquatic species at sampling point i from the centroid, p ij is the relative abundance of aquatic species j at sampling site i, T i is the number of functional trait categories in sampling point i; S i is the number of aquatic species in sampling point i, FDiv i is the functional differentiation FDiv index corresponding to aquatic organisms in sampling point i;

[0106] The calculation formula of the functional dispersion FDis is:

[0107]

[0108] Among them, c iis the weighted centroid of the multidimensional space composed of the functional traits of all aquatic species at sampling point i, c ik is the position of the weighted centroid of the multidimensional space of the functional trait in sampling point i on the functional trait k, a ij is the abundance of aquatic species j at sampling site i, x ijt is the trait attribute of functional trait k of aquatic species j at sampling site i, z ij is the distance from aquatic species j at sampling point i to the centroid of the multidimensional space, FDis i is the functional discreteness of sampling point i.

[0109] In the present invention, species functional traits and species information matrix T n,16 Corresponding to the biological function information, FRic is characterized by the volume of the convex hull formed by the trait set of the sample point biological community in the multidimensional space, reflecting the degree of utilization of the ecological functional space by the community organisms; the functional evenness FEve represents the uniformity or regularity of the abundance distribution of each functional characteristic in the multidimensional functional convex hull of the sampling point community; the functional differentiation FDiv represents the degree of differentiation of the abundance distribution of each functional characteristic in the multidimensional functional convex hull of the sampling point community; FDis calculates the average distance from a single species to the centroid of the multidimensional functional space of all species, measuring the similarity or dispersion of species functions in the community.

[0110] The calculation formula of the pedigree diversity index is:

[0111] PD i =(N′ i -1)+P i

[0112] Among them, N' i is the number of biological taxa participating in the phylogenetic tree at sampling point i, P i is the number of internal nodes of the minimum spanning path of the phylogenetic tree of all aquatic species spanning sampling point i, PD i is the phylogenetic diversity index of sampling point i, and the phylogenetic tree refers to a tree-like branch diagram reflecting the phylogenetic relationship determined by the Linnaean classification method for all aquatic organisms in the basin.

[0113] In the present invention, the phylogenetic diversity index PD represents the minimum total length of all phylogenetic branches experienced by a given set of taxa on the phylogenetic tree. The required phylogenetic tree information and species information matrix T n,16 Corresponding to the biological taxonomic information, the branch length information can be simplified to unit length.

[0114] Furthermore, the taxonomic diversity index, functional diversity index and lineage diversity index are summarized to construct a sample-diversity index matrix I m,9 .

[0115] In the present invention, a sample-diversity index matrix I corresponding to the sample-species matrix is generated. m,9 , where I m,9 is a matrix consisting of 9 biodiversity indices for m sampling points.

[0116] Said step S4, wherein:

[0117] Mantel analysis was performed using the Pearson correlation coefficient method, and the significance of the correlation coefficient was obtained using the t-test. The calculation formula for the correlation coefficient is as follows:

[0118]

[0119] Where r xy is the Pearson correlation coefficient between the pressure factor x and the diversity index y, COV xy is the covariance between pressure factors and diversity index, SD x is the standard deviation of the pressure factor x, SD y is the standard deviation of the diversity index y; the value of the correlation coefficient is used as the degree of influence.

[0120] In the present invention, the results are further organized and plotted to obtain the correlation between aquatic biodiversity and stress factors. The random forest algorithm and structural equation model involved in the present invention are both well-known random forest algorithms and structural equation models in the art.

[0121] The present invention uses a random forest algorithm to rank the importance of the first influencing factor to be analyzed for explaining the biodiversity index.

[0122] In the random forest algorithm, the parameter ntree, the number of optimal decision trees included, was specified as a fixed value of 500. The importance of stress factors was ranked according to the increase in mean squared error. The parameter num.rep, the number of permutations of the response variable (biodiversity index), was specified as a fixed value of 100. A new random forest model was constructed for each newly permuted data set to form a null distribution and obtain the significance of the importance of each stress factor.

[0123] Arrange the results and plot them to obtain the importance ranking of the stress factors.

[0124] Based on the data obtained above, a piecewise structural equation model was constructed to identify the pathways by which important stress factors affect water ecology.

[0125] Based on the nature of the first influencing factors to be analyzed, all factors were categorized into five groups: land use, water quality, hydrology, climate, and geography. General linear regression was used to construct five corresponding composite pressure factors. The composite factors were combined with the diversity index to construct a structural equation model. The piecewise structural equation model consisted of interconnected piecewise linear mixed models, each with a random intercept assigned to the location of m water bodies. Model adjustment and optimization were performed using the Fisher's C value, chi-squared test, and AIC values. Ultimately, the P value was greater than 0.05, and the structural equation set with the lowest AIC value within the model set was selected as the optimal model.

[0126] Organize the results and draw a map to identify the pathways by which each of the first influencing factors to be analyzed affects water ecology.

[0127] Step S5: determining the water ecological pressure of the study area based on the action paths of the first influencing factors to be analyzed on the water ecology, wherein:

[0128] Based on the structural equation, the explained rate (R2) of various diversity indices was obtained to identify the overall driving force of stress factors on biodiversity variation in the basin; by drawing the structural equation path diagram and calculating the direct effect, indirect effect and total effect value of each type of stress factor on various biodiversity indices, the magnitude and mode of the effect of different stress factors on diversity variation were determined.

[0129] In the present invention, based on the results obtained above, the correlation between biodiversity and stress factors is obtained, the importance ranking of stress factors is obtained, the action path of the first influencing factor to be analyzed affecting water ecology is identified, and the water ecological stress diagnosis based on the machine learning algorithm is completed.

[0130] like Figure 3 As shown, the present invention provides a water ecological pressure detection device based on machine learning, the device comprising:

[0131] Initialization module: configured to obtain an aquatic organism monitoring dataset, process the aquatic organism monitoring dataset to obtain a sample point-species matrix, wherein the sample point-species matrix represents the observed quantity information or existence information of all aquatic organism species monitored in the study area at each sampling point;

[0132] The first matrix module is configured to obtain the functional traits and pedigree information of the aquatic organisms involved in the sample point-species matrix, generate a species information matrix corresponding to the sample point-species matrix, wherein each aquatic organism in the species information matrix corresponds to the numerical value of its functional traits in morphology, ecology, and habits, as well as Linnaean biological classification information; based on the sample point-species matrix, determine the pressure factor of each sampling point, wherein the pressure factor is a plurality of influencing factors related to human influence, natural environment, and spatial distance, and generate a sample point-environment factor matrix corresponding to the sample point-species matrix;

[0133] A second matrix module is configured to determine the diversity index corresponding to the study area based on the sample-species matrix, the species information matrix and the aquatic organism monitoring dataset, wherein the diversity index includes a taxonomic diversity index, a functional diversity index and a phylogenetic diversity index, and generate a sample-diversity index matrix corresponding to the sample-species matrix;

[0134] The first analysis module is configured to calculate the Euclidean distance matrix of each diversity index vector in the sample point-diversity index matrix and each pressure factor vector in the sample point-environmental factor matrix, and perform Mantel analysis between the two types of distance matrices to determine a number of influencing factors with an impact greater than a preset threshold as first influencing factors to be analyzed; use a random forest algorithm to rank the importance of the first influencing factors to be analyzed; construct a structural equation model to determine the action path of each first influencing factor to be analyzed on the water ecology according to the ranking;

[0135] The second analysis module is configured to determine the water ecological pressure of the study area based on the action path of each first influencing factor to be analyzed on the water ecology.

[0136] An embodiment of the present invention further provides an electronic device, including:

[0137] A processor, which is used to execute multiple instructions;

[0138] A memory for storing a plurality of instructions;

[0139] The plurality of instructions are used to be stored by the memory and loaded and executed by the processor to implement the method as described above.

[0140] An embodiment of the present invention further provides a computer-readable storage medium, wherein a plurality of instructions are stored in the storage medium; the plurality of instructions are used for a processor to load and execute the method described above.

[0141] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0142] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, which may be electrical, mechanical or other forms.

[0143] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0144] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or hardware plus software functional units.

[0145] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a physical server, or a network cloud server, etc., and requires the Ubuntu operating system to be installed) to perform some of the steps of the method described in various embodiments of the present invention. The aforementioned storage medium includes: a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0146] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Any simple modifications, equivalent changes and modifications made to the above embodiment based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for detecting water ecological stress based on machine learning, characterized in that: The method comprises the following steps: Step S1: Acquire an aquatic organism monitoring dataset, and process the aquatic organism monitoring dataset to obtain a sample point-species matrix, wherein the sample point-species matrix represents the observed quantity information or existence information of all aquatic organism species monitored in the study area at each sampling point; Step S2: Acquire functional traits and pedigree information of aquatic organisms involved in the sample point-species matrix, and generate a species information matrix corresponding to the sample point-species matrix, wherein each aquatic organism in the species information matrix corresponds to the numerical value of its functional traits in morphology, ecology, and habits, as well as Linnaean biological classification information; Based on the sample point-species matrix, determine the pressure factor of each sampling point, wherein the pressure factor is a plurality of influencing factors related to human influence, natural environment, and spatial distance, and generate a sample point-environment factor matrix corresponding to the sample point-species matrix; Step S3: Based on the sample site-species matrix, the species information matrix and the aquatic organism monitoring dataset, determining the diversity index corresponding to the study area, wherein the diversity index includes a taxonomic diversity index, a functional diversity index and a phylogenetic diversity index, and generating a sample site-diversity index matrix corresponding to the sample site-species matrix; Step S4: Calculate the Euclidean distance matrix of each diversity index vector in the sample point-diversity index matrix and each pressure factor vector in the sample point-environmental factor matrix, perform Mantel analysis between the two types of distance matrices, and determine several influencing factors with an impact greater than a preset threshold as the first influencing factors to be analyzed; use the random forest algorithm to rank the importance of the first influencing factors to be analyzed; construct a structural equation model, and determine the action path of each first influencing factor to be analyzed on the water ecology according to the ranking; Step S5: Determine the water ecological pressure of the study area based on the action paths of each first influencing factor to be analyzed on the water ecology.

2. The method according to claim 1, wherein The step S2 is to obtain the functional traits and pedigree information of the aquatic organisms involved in the sample point-species matrix, that is, to obtain the sample point-species matrix D m,n The functional traits and pedigree information of n aquatic organisms involved in the project include: obtaining Linnaean taxonomic information of n aquatic organisms, including their phylum, class, order, family, genus, species information and corresponding unique Latin name; obtaining 9 types of species functional traits of n aquatic organisms, including morphological characteristics, ecological characteristics, and habits. Morphological characteristics include exoskeleton status, breathing mode, body size, and body shape. Ecological characteristics include substrate selection, lifestyle, and functional feeding groups. Habits include refuge and chemotaxis. Corresponding values are assigned to different biological traits.

3. The method according to claim 1, wherein The taxonomic diversity index includes species richness index, Shannon diversity index, Simpson diversity index, and Pielou evenness index.

4. The method according to claim 3, wherein The calculation formula of the species richness index is: q i =S i Among them, q i is the species richness of the i-th sampling point, S i is the number of aquatic species observed at sampling point i, corresponding to the sample point-species matrix D m,n The number of aquatic species with a species abundance greater than 0 in the sampling point corresponding to the sampling point; The calculation formula of the Shannon diversity index is: p ij =n ij / N i N 1i =exp(H i ) Among them, N 1i is the Shannon diversity index of sampling point i, H i is the Shannon entropy index of sampling point i, p ij is the relative abundance of aquatic species j at sampling point i, q i is the species richness of sampling point i, n ij is the abundance of aquatic species j at sampling point i, N i is the total abundance of all aquatic species in sampling site i; The calculation formula of Simpson's diversity index is: Among them, D' i is the Simpson diversity index of sampling point i, S i is the number of aquatic species in sampling point i, p ij is the relative abundance of aquatic species j at sampling site i; The calculation formula of Pielou evenness index is: J i =H i / H i max Among them, q i is the species richness of sampling point i, H i is the Shannon entropy index of sampling point i, H i max is the maximum value of the Shannon entropy index when the abundance of all aquatic species in sampling point i is consistent, J i is the Pielou uniformity index of sampling point i.

5. The method according to claim 1, wherein The functional diversity index includes functional richness FRic, functional evenness FEve, functional differentiation FDiv and functional dispersion FDis, where: The functional richness FRic corresponding to each sampling point was calculated using the Quickhull algorithm; The calculation formula of the functional uniformity FEve is: Among them, EW li is the weighted uniformity of the aquatic organism set at sampling point i, l is a branch of the minimum spanning tree consisting of all functional sites of aquatic species in the multidimensional functional convex hull corresponding to the sampling point, a and b are the aquatic species involved in branch l, dist(a,b) is the Euclidean distance between the functional sites of aquatic species a and aquatic species b, w a 、w b is the relative abundance of aquatic species a and aquatic species b, PEW li is the partial weighted uniformity of sampling point i, S i is the number of aquatic species observed at sampling point i, FEve i is the functional uniformity of sampling point i; The calculation formula of the functional differentiation FDiv is: Among them, g ik is the coordinate of the center of gravity of the multidimensional functional convex hull of the aquatic organism set corresponding to sampling point i on functional trait k, V i is the number of aquatic species that constitute the vertices of the multidimensional functional convex hull at sampling point i, x ijk is the coordinate of aquatic species j on functional trait k at sampling point i, dG ij is the Euclidean distance from the functional space position of aquatic species j in sampling point i to the center of gravity of the multidimensional functional convex hull, is the average distance from all aquatic species in sampling point i to the centroid of the multidimensional functional convex hull, Δd i is the abundance-weighted deviation of all aquatic species at sampling point i from the center of gravity of the multidimensional functional convex hull, Δ|d i | is the absolute abundance weighted deviation of all aquatic species at sampling point i from the centroid, p ij is the relative abundance of aquatic species j at sampling site i, T i is the number of functional trait categories in sampling point i; S i is the number of aquatic species in sampling point i, FDiv i is the functional differentiation FDiv index corresponding to aquatic organisms in sampling point i; The calculation formula of the functional dispersion FDis is: Among them, c i is the weighted centroid of the multidimensional space composed of the functional traits of all aquatic species at sampling point i, c ik is the position of the weighted centroid of the multidimensional space of functional traits in sampling point i on functional trait k, a ij is the abundance of aquatic species j at sampling site i, x ijt is the trait attribute of functional trait k of aquatic species j at sampling site i, z ij is the distance from aquatic species j at sampling point i to the centroid of the multidimensional space, FDis i is the functional discreteness of sampling point i.

6. The method according to claim 1, wherein The calculation formula of the pedigree diversity index is: PD i =(N′ i -1)+P i Among them, N' i is the number of biological taxa participating in the phylogenetic tree at sampling point i, P i is the number of internal nodes of the minimum spanning path of the phylogenetic tree of all aquatic species spanning sampling point i, PD i is the phylogenetic diversity index of sampling point i, and the phylogenetic tree refers to a tree-like branch diagram reflecting the phylogenetic relationship determined by the Linnaean classification method for all aquatic organisms in the basin.

7. The method according to claim 3, wherein In step S4, the Mantel analysis is performed using the Pearson correlation coefficient method, and the significance of the correlation coefficient is obtained by t-test. The calculation formula of the correlation coefficient is as follows: Where r xy is the Pearson correlation coefficient between the pressure factor x and the diversity index y, COV xy is the covariance between pressure factors and diversity index, SD x is the standard deviation of the pressure factor x, SD y is the standard deviation of the diversity index y; the value of the correlation coefficient is used as the degree of influence.

8. A water ecological pressure detection device based on machine learning, characterized in that: The device comprises: Initialization module: configured to obtain an aquatic organism monitoring dataset, process the aquatic organism monitoring dataset to obtain a sample point-species matrix, wherein the sample point-species matrix represents the observed quantity information or existence information of all aquatic organism species monitored in the study area at each sampling point; The first matrix module is configured to obtain the functional traits and pedigree information of the aquatic organisms involved in the sample point-species matrix, generate a species information matrix corresponding to the sample point-species matrix, wherein each aquatic organism in the species information matrix corresponds to the numerical value of its functional traits in morphology, ecology, and habits, as well as Linnaean biological classification information; based on the sample point-species matrix, determine the pressure factor of each sampling point, wherein the pressure factor is a plurality of influencing factors related to human influence, natural environment, and spatial distance, and generate a sample point-environment factor matrix corresponding to the sample point-species matrix; A second matrix module is configured to determine the diversity index corresponding to the study area based on the sample-species matrix, the species information matrix and the aquatic organism monitoring dataset, wherein the diversity index includes a taxonomic diversity index, a functional diversity index and a phylogenetic diversity index, and generate a sample-diversity index matrix corresponding to the sample-species matrix; The first analysis module is configured to calculate the Euclidean distance matrix of each diversity index vector in the sample point-diversity index matrix and each pressure factor vector in the sample point-environmental factor matrix, and perform Mantel analysis between the two types of distance matrices to determine a number of influencing factors with an impact greater than a preset threshold as first influencing factors to be analyzed; use a random forest algorithm to rank the importance of the first influencing factors to be analyzed; construct a structural equation model to determine the action path of each first influencing factor to be analyzed on the water ecology according to the ranking; The second analysis module is configured to determine the water ecological pressure of the study area based on the action path of each first influencing factor to be analyzed on the water ecology.

9. An electronic device comprising: A processor, which is used to execute multiple instructions; A memory for storing a plurality of instructions; The plurality of instructions are used to be stored in the memory and loaded and executed by the processor according to any one of claims 1 to 7.

10. A computer-readable storage medium, wherein a plurality of instructions are stored in the storage medium; the plurality of instructions are used for a processor to load and execute the method according to any one of claims 1 to 7.

Images