Water quality treatment unit identification method based on forest ecological integrity

Abstract

The invention relates to the field of geographic information, in particular to a forest ecological integrity-based water quality treatment unit identification method, which comprises the following steps of: constructing a feature matrix according to forest ecological integrity data and water quality data of grid units of each forest region; obtaining a forest ecological integrity characteristic matrix and a water quality characteristic matrix of each grid unit; performing feature fusion on the forest ecological integrity feature matrix and the water quality feature matrix of each grid unit to obtain a dynamic coupling matrix of each grid unit; according to the dynamic coupling matrix of each grid unit, performing dynamic influence evaluation on water quality by forest ecological integrity to obtain a dynamic influence evaluation value of each grid unit; and water quality treatment unit identification is carried out according to the dynamic influence evaluation value, a water quality treatment unit topological graph is constructed, the dynamic influence of forest ecological integrity on water quality is quantified, and the accuracy of water quality treatment unit identification is improved.

Description

Technical Field

[0001] This invention relates to the field of geographic information, and in particular to a method, apparatus, computer equipment, and storage medium for identifying water quality management units based on forest ecological integrity. Background Technology

[0002] Forest disturbance and management have a significant impact on water quality; deforestation and forest landscape fragmentation lead to severe water degradation. Therefore, quantitatively revealing the relationship between forest change and water quality is crucial for ensuring sustainable water environment protection through the regulation of forest management.

[0003] Current analyses of forest ecological integrity widely employ a species composition-stand structure-function framework based on ecosystem characteristics. However, this framework primarily focuses on two-dimensional vegetation indices, making it difficult to quantify the three-dimensional structure of forests and their correlation with functions such as water conservation, soil and water conservation, and water purification. Furthermore, watershed water quality assessments typically rely on physicochemical parameter monitoring and hydrological model simulations, resulting in insufficient quantification of forest ecological structure. This often overlooks the regulatory role of forests in water quality and makes it difficult to accurately assess the impact of forest degradation on water quality. Summary of the Invention

[0004] Based on this, the purpose of this invention is to provide a method, apparatus, computer equipment, and storage medium for identifying water quality management units based on forest ecological integrity, which quantifies the dynamic impact of forest ecological integrity on water quality and improves the accuracy of water quality management unit identification.

[0005] In a first aspect, embodiments of this application provide a method for identifying water quality management units based on forest ecological integrity, comprising the following steps: The system obtains multi-source data and a dynamic coupling model for forest areas. The multi-source data includes forest ecological integrity data and water quality data for several raster units. The dynamic coupling model includes a feature matrix construction module, a feature fusion module, and an impact assessment module. The forest ecological integrity data and water quality data of each grid cell are input into the feature matrix construction module to construct the feature matrix, thereby obtaining the forest ecological integrity feature matrix and water quality feature matrix of each grid cell. The forest ecological integrity feature matrix and the water quality feature matrix of each grid cell are input into the feature fusion module, and the spatiotemporal attention fusion mechanism is used to perform feature fusion to obtain the dynamic coupling matrix of each grid cell. The dynamic coupling matrix of each grid cell is input into the impact assessment module to assess the dynamic impact of forest ecological integrity on water quality, and the dynamic impact assessment value of each grid cell is obtained. Water quality treatment units are identified based on the dynamic impact assessment values ​​of each grid cell, and a topology map of water quality treatment units is constructed.

[0006] Secondly, embodiments of this application provide a water quality management unit identification device based on forest ecological integrity, comprising: The data acquisition module is used to acquire multi-source data and a dynamic coupling model of the forest area. The multi-source data includes forest ecological integrity data and water quality data of several raster units. The dynamic coupling model includes a feature matrix construction module, a feature fusion module, and an impact assessment module. The matrix construction module is used to input the forest ecological integrity data and the water quality data of each grid cell into the feature matrix construction module to construct the feature matrix, thereby obtaining the forest ecological integrity feature matrix and the water quality feature matrix of each grid cell. The matrix fusion module is used to input the forest ecological integrity feature matrix and the water quality feature matrix of each grid cell into the feature fusion module, and perform feature fusion using a spatiotemporal attention fusion mechanism to obtain the dynamic coupling matrix of each grid cell. The grid impact assessment module is used to input the dynamic coupling matrix of each grid cell into the impact assessment module to conduct a dynamic impact assessment of the forest ecological integrity on water quality and obtain the dynamic impact assessment value of each grid cell. The water quality treatment unit identification module is used to identify water quality treatment units based on the dynamic impact assessment values ​​of each grid unit and to construct a water quality treatment unit topology map.

[0007] Thirdly, embodiments of this application provide a computer device, including: a processor, a memory, and a computer program stored in the memory and executable on the processor; when the computer program is executed by the processor, it implements the steps of the water quality management unit identification method based on forest ecological integrity as described in the first aspect.

[0008] Fourthly, embodiments of this application provide a storage medium storing a computer program that, when executed by a processor, implements the steps of the water quality management unit identification method based on forest ecological integrity as described in the first aspect.

[0009] In this application embodiment, a method, apparatus, computer equipment, and storage medium for identifying water quality management units based on forest ecological integrity are provided, which quantifies the dynamic impact of forest ecological integrity on water quality and improves the accuracy of water quality management unit identification.

[0010] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0011] Figure 1 A flowchart illustrating the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application; Figure 2 This is a flowchart illustrating step S2 of the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application. Figure 3 This is a flowchart illustrating step S2 in the water quality management unit identification method based on forest ecological integrity provided in the second embodiment of this application. Figure 4 This is a flowchart illustrating step S24 of the water quality management unit identification method based on forest ecological integrity provided in the second embodiment of this application. Figure 5 A flowchart illustrating step S24 of the water quality management unit identification method based on forest ecological integrity provided in the third embodiment of this application; Figure 6 This is a flowchart illustrating step S24 of the water quality management unit identification method based on forest ecological integrity provided in the fourth embodiment of this application. Figure 7 This is a flowchart illustrating step S3 of the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application. Figure 8 This is a flowchart illustrating step S4 of the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application. Figure 9 This is a flowchart illustrating step S5 of the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application. Figure 10 A schematic diagram of the structure of the water quality treatment unit identification device based on forest ecological integrity provided in the fifth embodiment of this application; Figure 11 This is a schematic diagram of the structure of a computer device provided in the sixth embodiment of this application. Detailed Implementation

[0012] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application.

[0013] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0014] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0015] Please see Figure 1 , Figure 1 The flowchart of the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application is shown. The method includes the following steps: S1: Obtain multi-source data and dynamic coupling models of the forest area.

[0016] The implementing entity of the water quality management unit identification method based on forest ecological integrity in this application is the identification device (hereinafter referred to as the identification device) of the water quality management unit identification method based on forest ecological integrity. In an optional embodiment, the identification device may be a computer device, a server, or a server cluster composed of multiple computer devices.

[0017] In this embodiment, the identification device obtains multi-source data of the forest area, wherein the multi-source data includes forest ecological integrity data and water quality data of several grid units, and the forest ecological integrity data includes species diversity parameters, forest stand structure parameters, ecological process parameters and soil property parameters.

[0018] Specifically, the species diversity parameters include species richness and the Shannon-Wiener diversity index; the stand structure parameters include canopy closure, leaf area index, vertical stratification, and tree density; the ecological process parameters include mortality rate, litter storage, litter thickness, litter nitrogen and phosphorus content, and leaf surface water holding capacity; the soil property parameters include soil bulk density, soil moisture content, soil erosion, total soil nitrogen content, and total soil phosphorus content. The water quality data is acquired using a water quality sensor network and includes pH, water temperature, five-day biochemical oxygen demand (BOD), dissolved oxygen, conductivity, permanganate index, nitrogen form data, and phosphorus form data. The nitrogen form data includes total nitrogen, total dissolved nitrogen, and nitrate nitrogen; the phosphorus form data includes total phosphorus, total dissolved phosphorus, and particulate phosphorus.

[0019] The identification device obtains a dynamic coupling model, wherein the dynamic coupling model includes a feature matrix construction module, a feature fusion module, and an impact assessment module.

[0020] S2: Input the forest ecological integrity data and water quality data of each grid cell into the feature matrix construction module to construct the feature matrix, and obtain the forest ecological integrity feature matrix and water quality feature matrix of each grid cell.

[0021] In this embodiment, the identification device inputs the forest ecological integrity data and water quality data of each grid cell into the feature matrix construction module to construct the feature matrix, thereby obtaining the forest ecological integrity feature matrix and water quality feature matrix of each grid cell.

[0022] Please see Figure 2 , Figure 2 The flowchart of step S2 in the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application includes steps S21 to S23, as follows: S21: Extract local voxel features from the forest ecological integrity data to obtain local voxel feature representations.

[0023] In this embodiment, the recognition device extracts local voxel features from the forest ecological integrity data based on a preset convolution kernel, specifically a 7×7×7 three-dimensional convolution kernel. It captures the nonlinear response of tree density and soil erosion in the local voxel domain to obtain a local voxel feature representation.

[0024] S22: Construct a forest ecological integrity heterogeneous graph based on the local voxel feature representation to obtain a forest ecological integrity heterogeneous graph; use the meta-path mapping method to construct a matrix based on the forest ecological integrity heterogeneous graph to obtain a forest ecological integrity heterogeneous graph matrix.

[0025] In this embodiment, the identification device constructs a forest ecological integrity heterogeneous graph based on the local voxel feature representation to obtain the forest ecological integrity heterogeneous graph. The forest ecological integrity heterogeneous graph includes several types of nodes and edges between these nodes, as described below:

[0026] In the formula, These are nodes in the heterogeneity map of forest ecological integrity. Generate functions for node features. This is a local voxel feature. For K-means clustering function, This refers to the channel sub-feature within the local voxel features. To represent the fast point feature histogram calculation function, For spatial sub-features in local voxel features, For terrain gradient, For feature splicing operators, The first in the forest ecological integrity heterogeneity diagram i The type of node and the first j Edges between nodes of different types Construct functions for edges, The first in the forest ecological integrity heterogeneity diagram i The type of node and the first j Spatial distance between nodes of different types For the first i Connection radius of each type For the first i Node feature vectors of each type of node For the first i The type of node and the first j Similarity thresholds between nodes of different types For the first i Each type of node type identifier.

[0027] The identification device performs meta-path mapping between nodes corresponding to various types of forest ecological integrity data and nodes corresponding to different types of forest ecological integrity data in the forest ecological integrity heterogeneity map, thereby obtaining a forest ecological integrity heterogeneity map matrix. The forest ecological integrity heterogeneity map matrix includes label vectors for several meta-paths, as described below:

[0028] In the formula, This is an example of heterogeneous primitive paths for forest ecological integrity. For specific meta-paths, For relation types, For node sequence, For meta-path embedding, For type encoding, It is a meta-path type. Embedded for path instances, M This represents the total number of meta-paths. Embedded dimension for meta-path, This is a meta-path embedding algorithm. This is path data for heterogeneous primitives representing forest ecological integrity.

[0029] S23: Construct a gated aggregation feature matrix based on the label vectors of several metapaths in the heterogeneous graph matrix of forest ecological integrity, and obtain a gated aggregation feature matrix; perform spatial three-dimensional compression on the gated aggregation feature matrix to obtain a compressed feature matrix; construct a matrix based on the compressed feature matrix and the metapath data of the heterogeneous graph of forest ecological integrity to obtain a forest ecological integrity feature matrix.

[0030] In this embodiment, the identification device constructs a gated aggregation feature matrix based on the label vectors of several metapaths in the forest ecological integrity heterogeneous graph matrix and a preset gated aggregation feature algorithm. The gated aggregation feature matrix includes aggregation feature vectors of several nodes. The gated aggregation feature algorithm is as follows:

[0031] In the formula, For the first m The first metapath l Layer node embedding vector, It is a non-linear activation function. For max pooling operators, For the meta-path transformation weights, For the first m Neighboring nodes connected by each meta-path For meta-path gating weights, For the Sigmoid function, For the gating parameter vector, T It is the transpose symbol. L The number of convolutional layers in the graph. For gated aggregation feature matrix, It is a tensor product.

[0032] The identification device performs three-dimensional spatial compression on the gated aggregated feature matrix according to a preset spatial compression algorithm to obtain a compressed feature matrix, wherein the spatial compression algorithm is:

[0033] In the formula, For dense spatial feature tensors, It is the nearest neighbor interpolation function. For deformable offset, It is a three-dimensional deformable convolution function. For average pooling, For deep attention parameters, To compress the feature matrix, This is the multiplication symbol for channel broadcasting.

[0034] The identification device constructs a matrix based on the compressed feature matrix and the heterogeneous primitive path data of forest ecological integrity, thereby obtaining a forest ecological integrity feature matrix, which is as follows:

[0035] In the formula, This is a matrix representing the characteristics of forest ecological integrity. For the graph gating weight matrix, This is the adjacency feature matrix constructed from the heterogeneous primitive path data of the forest ecological integrity. The diagonal weight of the metapath. For the sigmoid function, This is the meta-path feature transformation function. For feature splicing operators, The symbol for element-wise product. This is a local pooling function.

[0036] Please see Figure 3 , Figure 3 The flowchart of step S2 in the water quality management unit identification method based on forest ecological integrity provided in the second embodiment of this application includes steps S24 to S27, as follows: S24: Perform water quality characteristic analysis on the water quality data to obtain water quality characteristic analysis data.

[0037] In this embodiment, the identification device performs water quality characteristic analysis on the water quality data to obtain water quality characteristic analysis data. This data includes a precipitation event dissolved oxygen sudden drop pattern analysis sequence, a water quality parameter lag response topology map, and a multimodal water quality reliability tensor. The precipitation event dissolved oxygen sudden drop pattern analysis sequence is used to quantify the probability or intensity of a sudden drop in dissolved oxygen concentration caused by precipitation impact; the water quality parameter lag response topology map is used to indicate the spatial migration paths and temporal lag patterns of pollutants; and the multimodal water quality reliability tensor reflects the reliability of the water quality monitoring data.

[0038] Please see Figure 4 , Figure 4The flowchart of step S24 in the water quality management unit identification method based on forest ecological integrity provided in the second embodiment of this application includes steps S2401 to S2405, as follows: S2401: Nonlinear fusion and three-dimensional reconstruction of the pH, water temperature parameters and conductivity at each time point in the water quality data are performed to obtain the three-dimensional hidden state feature matrix at each time point.

[0039] In this embodiment, the identification device performs nonlinear fusion and three-dimensional reconstruction of the pH value, water temperature parameters and conductivity at various times in the water quality data to obtain the three-dimensional latent feature matrix at each time.

[0040] Specifically, the identification device constructs a data matrix based on the pH, water temperature, and conductivity parameters at various times in the water quality data. The identification device utilizes weight transformation to perform nonlinear fusion based on the data matrix and a preset nonlinear fusion algorithm, obtaining a nonlinear fusion matrix. The nonlinear fusion matrix is ​​then three-dimensionally reconstructed to obtain a three-dimensional latent feature matrix, wherein the three-dimensional latent feature matrix contains several spatial response patterns at several times and several physical channel features. The nonlinear fusion algorithm is as follows:

[0041] In the formula, It is a non-linear fusion matrix. It is the tangent curve function. The non-linear fusion weight matrix is... This is the feature bias vector.

[0042] S2402: Calculate the event association parameters based on the three-dimensional hidden state feature matrix at each time point and the preset rain intensity mask matrix to obtain the event association parameter set at each time point.

[0043] In this embodiment, the identification device calculates event association parameters based on the three-dimensional latent feature matrix at each time point and the preset rain intensity mask matrix to obtain the event association parameter set at each time point. The event association parameter set includes association parameters for several events, including acid dissolution event association parameters, temperature-density stratification event association parameters, and salt oxygen transport barrier event association parameters.

[0044] Specifically, the identification device performs vector calculations based on the three-dimensional latent feature matrix, a preset rainfall intensity mask matrix, and a corresponding nonlinear activation function to obtain a pH value vector, a water temperature parameter abrupt change vector, and a conductivity gradient vector. The rainfall intensity mask matrix is ​​constructed from rainfall intensity sequences acquired by weather radar, and the rainfall intensity mask matrix is ​​as follows:

[0045] In the formula, For the rain intensity mask matrix, for The sudden drop in rating data at any moment for Rainfall intensity at any given moment This represents the time difference between the current moment and the start of the rainstorm. This refers to the intensity of rainfall.

[0046] The identification device obtains acid dissolution event correlation parameters, temperature-density stratification event correlation parameters, and salt-impedance oxygen transport event correlation parameters based on the sudden drop score data, pH value vector, water temperature parameter abrupt change vector, conductivity gradient vector, and event correlation parameter algorithm in the rain intensity mask matrix. The event correlation parameter algorithm is as follows:

[0047] In the formula, For parameters related to acidic dissolution events, This is a vector of acid and base values. For temperature-density hierarchical event correlation parameters, This is the vector of abrupt changes in water temperature parameters. For parameters related to oxygen transport events in salt-barrier environments, This is the conductivity gradient vector.

[0048] S2403: Perform label vector transformation and matrix construction based on the event association parameter set at each time point to obtain the event label matrix at each time point; perform spatial weight distribution calculation based on the three-dimensional latent feature matrix, the event label matrix, and the preset spatial feature algorithm at each time point to obtain the spatial feature vector at each time point.

[0049] In this embodiment, the identification device performs tag vector conversion and matrix construction based on the event association parameter set at each time moment to obtain the event tag matrix at each time moment. The event tag matrix includes tag vectors of several events. The event tag vectors include acid dissolution event tag vectors, temperature-density stratification event tag vectors, and salt oxygen transport barrier event tag vectors.

[0050] Specifically, the identification device compares the associated parameters of the acid dissolution event, the temperature-density stratification event, and the salt-oxygen transport barrier event with their respective thresholds. Based on the comparison results, it constructs corresponding label vectors to obtain label vectors for the acid dissolution event, the temperature-density stratification event, and the salt-oxygen transport barrier event. The identification device combines the label vectors of the acid dissolution event, the temperature-density stratification event, and the salt-oxygen transport barrier event at the same time to obtain the event label matrix for each time point.

[0051] The recognition device calculates the spatial weight distribution based on the three-dimensional latent feature matrix, the event label matrix, and a preset spatial feature algorithm at each time step to obtain the spatial feature vector at each time step. The spatial feature algorithm is as follows:

[0052] In the formula, For the first k Attention weight parameters at each time step. For normalized exponential functions, For the event projection matrix, For the first k The event label matrix at each moment, For the first k Spatial feature vector at each time step For the first k The moment of the first j Attention weight parameters for each spatial pattern, The third hidden state feature matrix is ​​the... k The moment of the first j The latent feature vectors of a spatial pattern.

[0053] S2404: Perform spatiotemporal convolutional co-mapping on the event label matrix at each time step to obtain the channel gating signal at each time step; perform feature aggregation based on the three-dimensional latent feature matrix, the channel gating signal, and the preset spatiotemporal feature aggregation algorithm at each time step to obtain the spatiotemporal aggregated features at each time step.

[0054] In this embodiment, the identification device performs spatiotemporal convolutional co-mapping based on the event label matrix to obtain a channel gating signal, wherein the channel gating signal is:

[0055] In the formula, For the first k The channel gating signal at each moment, This is the event gating matrix.

[0056] The recognition device obtains spatiotemporal aggregated features at each time step based on the three-dimensional latent feature matrix, channel gating signals, and a preset spatiotemporal feature aggregation algorithm. The spatiotemporal feature aggregation algorithm is as follows:

[0057] In the formula, For the first k The spatiotemporal aggregation characteristics at each moment, For spatiotemporal convolution kernels, The third hidden state feature matrix is ​​the...k + τ At the [time]th moment j Latent feature vectors of spatial patterns τ The time step is the lag step.

[0058] S2405: Based on the spatial feature vectors, spatiotemporal aggregation features, event label vectors, and the preset warning vector algorithm at each time point, the warning vectors at each time point are obtained; based on the warning vectors at each time point, a sequence is constructed to obtain the dissolved oxygen drop pattern analysis sequence for precipitation events.

[0059] In this embodiment, the identification device obtains the warning vector for each moment based on the spatial feature vector, spatiotemporal aggregation feature, event tag vector, and a preset warning vector algorithm. A sequence is constructed based on the warning vectors for each moment to obtain the dissolved oxygen drop pattern analysis sequence for precipitation events. The warning vector algorithm is as follows:

[0060] In the formula, For the first k The warning vector at each moment, To fuse the projection matrix, For feature bias, This is the layer normalization function.

[0061] Please see Figure 5 , Figure 5 The flowchart of step S24 in the water quality management unit identification method based on forest ecological integrity provided in the third embodiment of this application includes steps S2411 to S2413, as follows: S2411: Encode the nitrogen and phosphorus speciation data at each time point in the water quality data to obtain nitrogen speciation coding sequences and phosphorus speciation coding sequences.

[0062] In this embodiment, the identification device encodes the nitrogen and phosphorus speciation data at various times in the water quality data to obtain nitrogen speciation encoding sequences and phosphorus speciation encoding sequences, wherein the nitrogen speciation encoding sequences and phosphorus speciation encoding sequences reflect the migration forms of particulate and dissolved states.

[0063] S2412: Input the nitrogen form coding sequence and phosphorus form coding sequence into the bidirectional gated loop unit to calculate the effect state feature and obtain the effect state feature sequence; perform pollutant effect classification analysis based on the effect state feature sequence and the preset pollutant effect classification algorithm to obtain the pollutant effect category sequence.

[0064] In this embodiment, the identification device inputs the nitrogen form encoding sequence and the phosphorus form encoding sequence into a bidirectional gated recurrent unit for effect-state feature calculation to obtain the effect-state feature sequence. Specifically, the identification device concatenates the nitrogen form encoding sequence and the phosphorus form encoding sequence, performs forward hidden state calculation and backward hidden state calculation based on the concatenated sequence and the bidirectional gated recurrent unit to obtain the forward hidden state sequence and the backward hidden state sequence, and performs effect-state feature calculation based on the forward hidden state sequence and the backward hidden state sequence to obtain the effect-state feature sequence.

[0065] The identification device performs pollutant effect classification analysis according to a preset pollutant effect classification algorithm to obtain a pollutant effect category sequence. This sequence includes effect category vectors at various time points, and each vector includes a synchronous effect vector, a lag effect vector, and an antagonistic effect vector. The pollutant effect classification algorithm is as follows:

[0066] In the formula, The first of the pollutant effect category sequence t The effect category vector at each time step. For classification weights, The first of the effect state characteristic sequences t The effect state eigenvector at time 1.

[0067] S2413: Perform multi-temporal coupling analysis of pollutants based on the pollutant effect category sequence to construct a water quality lag linkage response topology map.

[0068] In this embodiment, the identification device performs multi-temporal coupling analysis of pollutants based on the pollutant effect category sequence to construct a water quality lag-linkage response topology map. The water quality lag-linkage response topology map includes a node set, an edge set, and topological weights. The node set includes node vectors representing the various pollutant morphologies corresponding to each node at each time point. The edge set includes edge vectors between the various pollutant morphologies of each node, as described below:

[0069] In the formula, For a set of nodes, For the first i The node of the first p The node vector of each pollutant form This is a nitrogen-encoding sequence. The sequence is a phosphorus speciation encoding sequence. For the first iThe morphological component of each node is a numerical value that directly separates the phase ratio of a specific pollutant by multiplying the total nitrogen concentration obtained based on nitrogen speciation data by a preset organic nitrogen ratio coefficient, or by multiplying the total phosphorus concentration obtained based on phosphorus speciation data by a preset solubility equilibrium coefficient. For the first i Hydrodynamic regulation parameters of each node For the first i The effect state feature vector of each node, For node fusion function, It is the hyperbolic tangent function. Let be the set of edges. For the first i The node of the first p The form of the pollutant and the first j The node of the first l Boundary vectors between pollutant morphologies Indexed by time nodes, For the maximum time delay window, For the first i The node of the first p The form of the pollutant and the first j The node of the first l The strength of the correlation between different pollutant forms For the correlation threshold, The source projection matrix, For the target projection matrix, For the first i The node of the first p The node vector of each pollutant form For the first j The node of the first l The node vector of each pollutant form The attenuation coefficient is... For the first i The node and the first j The difference in hysteresis characteristics between nodes For topological weights, This is the weighted sharpening coefficient. For morphological correlation similarity function, It is a time-delay decay function. For the first i A set of neighboring nodes for a given node, wherein the set of neighboring nodes includes several other neighboring nodes, which are all other nodes that may migrate pollutants within a preset time window based on the various pollutant forms of the node. m This is the index of a neighbor node in the set of neighbor nodes.

[0070] Please see Figure 6 , Figure 6The flowchart of step S24 in the water quality management unit identification method based on forest ecological integrity provided in the fourth embodiment of this application includes steps S2421 to S2424, as follows: S2421: Sequences are constructed for the biochemical oxygen demand and initial permanganate index at each time point in the water quality data to obtain the initial biochemical oxygen demand sequence and the initial permanganate index sequence.

[0071] In this embodiment, the identification device constructs sequences of the biochemical oxygen demand (BOD) and the initial permanganate index at various times in the water quality data to obtain the initial BOD sequence and the initial permanganate index sequence.

[0072] S2422: Perform composite noise modeling and signal separation on the initial biochemical oxygen demand sequence to obtain the real biochemical oxygen demand sequence; perform asymmetric dilated convolution processing on the real biochemical oxygen demand sequence to obtain a reliable biochemical oxygen demand sequence.

[0073] In this embodiment, the identification device performs composite noise modeling and signal separation on the initial biochemical oxygen demand (BOD) sequence to obtain the true BOD sequence; the true BOD sequence is then subjected to asymmetric dilated convolution processing to obtain a reliable BOD sequence, wherein the true BOD sequence is:

[0074] In the formula, The first in the true biochemical oxygen demand sequence t The true biochemical oxygen demand vector at any given moment. The first in the initial biochemical oxygen demand sequence t The initial biochemical oxygen demand vector at each time point. The aging factor is... For the first t Electrode aging index at a given time point Historical decay factor This is the temperature fluctuation coefficient. It serves as the periodic base.

[0075] The reliable biochemical oxygen demand sequence is as follows:

[0076] In the formula, For the first t A reliable biochemical oxygen demand vector at each moment. It is a first-order difference operator. The mutation threshold, For convolution kernel weights, b It is an expansion factor.

[0077] S2423: Anomaly mask generation is performed on the initial permanganate index sequence to obtain a corrected permanganate index sequence; a credibility index matrix is ​​constructed based on the credible biochemical oxygen demand sequence and the corrected permanganate index sequence to obtain a credibility index matrix.

[0078] In this embodiment, the identification device generates an anomaly mask on the initial permanganate index sequence to obtain a corrected permanganate index sequence; the identification device constructs a credibility index matrix based on the reliable biochemical oxygen demand sequence and the corrected permanganate index sequence to obtain a credibility index matrix, wherein the credibility index matrix is:

[0079] In the formula, This is a credibility index matrix. For the first N A reliable biochemical oxygen demand vector at each moment. For the first N The corrected permanganate index vector at each time step.

[0080] S2424: Perform forward attention calculation and backward attention calculation based on the credibility index matrix and the preset bidirectional attention gating algorithm to obtain the forward attention vector and backward attention vector at each time step; perform deep fusion and tensor generation based on the forward attention vector and backward attention vector at each time step to obtain the multimodal water quality credibility tensor.

[0081] In this embodiment, the recognition device performs forward attention calculation and backward attention calculation based on the credibility index matrix and a preset bidirectional attention gating algorithm to obtain the forward attention vector and backward attention vector at each time step. The bidirectional attention gating algorithm is as follows:

[0082] In the formula, For the first t The forward attention vector at each time step For the first t Credibility index at a given moment For historical time window (1: t A credibility cumulative matrix is ​​constructed by accumulating credibility indicators. For future time windows ( t : N A credibility cumulative matrix is ​​constructed by accumulating credibility indicators. For dimension parameters, For the first t The forward hidden state at time 1 For the first tThe backward attention vector at time step 1. For similarity scale, For the first t The backward hidden state at time 1 For sparse activation functions, For max pooling function, It is the Euclidean norm.

[0083] The recognition device performs deep fusion and tensor generation based on the forward and backward attention vectors at each time step to obtain a multimodal water quality confidence tensor, wherein the multimodal water quality confidence tensor is:

[0084] In the formula, For deep fusion vectors, For the fusion matrix, Forward gate vector, This is the backward gating vector. For the multimodal water quality confidence tensor, This is the time mapping matrix. The spectral mapping matrix, Modulo 1 product, It is a product of two modulo products.

[0085] S25: Obtain the hydrological topology node matrix; calculate spatial features based on the hydrological topology node matrix and the dissolved oxygen drop pattern analysis sequence of precipitation events to obtain spatial feature vectors; perform spatiotemporal dynamic coding based on the water quality parameter hysteresis response topology map and the multimodal water quality confidence tensor to obtain a spatiotemporal dynamic coding matrix.

[0086] In this embodiment, the identification device obtains a hydrological topology node matrix, wherein the hydrological topology node matrix is ​​an adjacency matrix generated based on the DEM confluence path.

[0087] The identification device performs spatial feature calculations based on the hydrological topology node matrix, the peak warning vector in the dissolved oxygen drop pattern analysis sequence of precipitation events, and a preset spatial feature algorithm to obtain spatial feature vectors for convolution alignment. The spatial feature algorithm is as follows:

[0088] In the formula, F For spatial feature vectors, for RELU Activation function For peak warning vector, Specifically, the river network density feature vector is obtained by feature extraction based on the hydrological topological node matrix.

[0089] The identification device performs spatiotemporal dynamic encoding based on the water quality parameter hysteresis response topology map and the multimodal water quality confidence tensor to obtain a spatiotemporal dynamic encoding matrix. The spatiotemporal dynamic encoding algorithm is as follows:

[0090] In the formula, It is a spatiotemporal dynamic encoding matrix. For dynamic encoding functions, For topology encoding functions, This is the symbol for topological weight fusion. This is the credibility transformation function.

[0091] S26: Based on the spatial feature vector, spatiotemporal dynamic coding matrix, and analysis sequence of dissolved oxygen drop pattern in precipitation events, feature fusion is performed to obtain a fused node feature matrix; based on the fused node feature matrix and a preset graph attention algorithm, adjacent node features are aggregated to obtain an aggregated node feature matrix.

[0092] In this embodiment, the identification device performs feature fusion based on the spatial feature vector, the spatiotemporal dynamic coding matrix, the peak warning vector in the dissolved oxygen sudden drop pattern analysis sequence of the precipitation event, and a preset feature fusion algorithm to obtain a fused node feature matrix. The fused node feature matrix includes node features of several nodes, and the feature fusion algorithm is as follows:

[0093] In the formula, To fuse the node feature matrix, For tensor vectorization operators, This is the peak warning vector.

[0094] The recognition device performs adjacent node feature aggregation on the fused node feature matrix based on the fused node feature matrix and a preset graph attention algorithm to obtain an aggregated node feature matrix. The graph attention algorithm is as follows:

[0095] In the formula, For the first i The node and the first j Attention weight coefficients among neighboring nodes This is the transpose of the attention mechanism weight vector. For the first i The node characteristics of each node. For the first j The node characteristics of each neighboring node, To query the transformation matrix, The key-value transformation matrix, For the first i The set of adjacent nodes of a node. For the first i The node aggregation characteristics of each node. For multi-head splicing operators, M For the number of heads to focus on, For the first m The first i The node and the first j Attention weight coefficients among neighboring nodes For the first m The value transformation matrix of the head.

[0096] S27: Using the hyperspherical projection method, the feature matrix of the aggregated nodes is mapped to precipitation intensity to obtain a precipitation intensity matrix; based on the precipitation intensity matrix, the nitrogen and phosphorus content data of litter in the forest ecological integrity data, and the preset litter nitrogen and phosphorus release kinetic model, a water quality feature matrix is ​​constructed to obtain a water quality feature matrix.

[0097] In this embodiment, the identification device employs a hyperspherical projection method to map the aggregated node feature matrix to precipitation intensity according to a preset precipitation intensity mapping algorithm, thereby obtaining a precipitation intensity matrix. The precipitation intensity matrix includes the precipitation intensity vectors of each node at each time point. The precipitation intensity mapping algorithm is as follows:

[0098] In the formula, For the first t The first moment i The precipitation intensity vector of the node. This is a time-dependent intensity scaling factor. The time-dynamic sphere center vector, No. j Node aggregation characteristics For time dynamic radius scalar, N This represents the total number of nodes.

[0099] The identification device calculates nutrient pulse flux and node cumulative flux based on the precipitation intensity matrix, litter nitrogen and phosphorus content data in the forest ecological integrity data, and a preset litter nitrogen and phosphorus release kinetic model, using a preset enzyme kinetics-fluid dynamics coupling algorithm. Based on the obtained nutrient pulse flux and node cumulative flux, a water quality feature matrix is ​​constructed to obtain an initial water quality feature matrix. The enzyme kinetics-fluid dynamics coupling algorithm is as follows:

[0100] In the formula, For the first t The first moment i Nutrient pulse flux at nodes The nitrogen release index, This is the terrain damping coefficient. The phosphorus release index, The first in the litter nitrogen data matrix t The first moment i Nitrogen content of litter at nodes Power of 1 For the first i The topographic gradient modulus of the node, For the maximum enzymatic hydrolysis rate, It is the Michaelis constant. The first in the litter nitrogen data matrix t The first moment i The phosphorus content of the litter at the nodes Power of 1 For the first i The cumulative throughput of a node. For time step, This is the surface runoff vector. This is the soil normal vector.

[0101] The identification device performs a pooling weighted pooling on the initial water quality feature matrix to obtain a water quality feature matrix, wherein the water quality feature matrix is:

[0102] In the formula, This is the feature matrix after pooling. D For degree matrix, A This is the hydrological topological adjacency matrix. This is the confluence weight matrix. This is the initial water quality feature matrix. The characteristic transformation matrix, For multilayer perceptron functions, The first feature fusion matrix, This is the second feature fusion matrix. For max pooling function, For average pooling function, This is the water quality characteristic matrix. Q This is the water quality enhancement feature matrix obtained by enhancing the initial water quality feature matrix.

[0103] S3: Input the forest ecological integrity feature matrix and the water quality feature matrix of each grid cell into the feature fusion module, and perform feature fusion using a spatiotemporal attention fusion mechanism to obtain the dynamic coupling matrix of each grid cell.

[0104] In this embodiment, the identification device inputs the forest ecological integrity feature matrix and the water quality feature matrix of each grid cell into the feature fusion module, and performs feature fusion using a spatiotemporal attention fusion mechanism to obtain the dynamic coupling matrix of each grid cell.

[0105] Please see Figure 7 , Figure 7 The flowchart of step S3 in the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application includes steps S31 to S33, as follows: S31: Obtain the terrain-weighted mask; perform Hadamard product operation based on the forest ecological integrity feature matrix and the terrain-weighted mask to construct a terrain-enhanced forest ecological field.

[0106] In this embodiment, the identification device obtains a terrain-weighted mask, which is dynamically generated by using a soil erosion gradient matrix constructed based on the terrain gradient tensor and the land cover type, through three layers of calculation: coupled erosion-sensitive area calibration, root anchoring weight quantification, and canopy interception correction.

[0107] The identification device performs Hadamard product calculations based on the forest ecological integrity feature matrix and the terrain-weighted mask to construct a terrain-enhanced forest ecological field, wherein the terrain-enhanced forest ecological field is:

[0108] In the formula, To enhance the forest ecosystem by adapting to the terrain, For terrain-weighted masking, This is the terrain gradient enhancement coefficient. This is the Sobel spatial gradient operator.

[0109] S32: Perform hydrological response modeling on the water quality feature matrix to obtain precipitation time delay sequence; obtain time-varying water quality decay feature matrix based on the water quality feature matrix, precipitation time delay sequence and preset time-varying water quality decay feature algorithm.

[0110] In this embodiment, the identification device performs hydrological response modeling on the water quality feature matrix to obtain a precipitation time-lag sequence; based on the water quality feature matrix, the precipitation time-lag sequence, and a preset time-varying water quality attenuation feature algorithm, a time-varying water quality attenuation feature matrix is ​​obtained, wherein the time-varying water quality attenuation feature algorithm is:

[0111] In the formula, This is the time-varying water quality degradation characteristic matrix. The pollutant degradation coefficient, It is a precipitation time lag sequence. For time-varying gradient weights, This is the time gradient matrix.

[0112] S33: Based on the terrain-enhanced forest ecological field, the time-varying water quality decay feature matrix, and the preset spatiotemporal attention fusion algorithm, feature fusion is performed to obtain a fused feature matrix, which serves as the dynamic coupling matrix.

[0113] In this embodiment, the identification device performs feature fusion based on the terrain-enhanced forest ecological field, the time-varying water quality decay feature matrix, and a preset spatiotemporal attention fusion algorithm to obtain a fused feature matrix, which serves as the dynamic coupling matrix. The spatiotemporal attention fusion algorithm is as follows:

[0114] In the formula, To fuse the feature matrix, For the Laplace operator, For curvature mapping function, For graph attention networks, For temporal convolutional networks, It is a cross-domain gating function, which is constructed using the spatial dominant factor of the forest ecological field construction based on the terrain.

[0115] S4: Input the dynamic coupling matrix of each grid cell into the impact assessment module to conduct a dynamic impact assessment of the forest ecological integrity on water quality, and obtain the dynamic impact assessment value of each grid cell.

[0116] In this embodiment, the identification device inputs the dynamic coupling matrix of each grid cell into the impact assessment module to conduct a dynamic impact assessment of the forest ecological integrity on water quality, and obtains the dynamic impact assessment value of each grid cell, wherein the dynamic impact assessment value reflects the dynamic impact of the forest ecological integrity on water quality.

[0117] The impact assessment module includes a feature coupling module and a fully connected module. Please refer to [link / reference]. Figure 8 , Figure 8 The flowchart of step S4 in the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application includes steps S41 to S42, as follows: S41: Input the dynamic coupling matrix into the feature coupling module for parameter extraction to obtain the scale parameter and translation parameter of the dynamic coupling matrix; perform feature decoupling on the dynamic coupling matrix according to the scale parameter and translation parameter to obtain the feature decoupling matrix; perform feature compression on the feature decoupling matrix to obtain the compressed feature matrix.

[0118] The feature coupling module adopts a reversible normalized flow model. In this embodiment, the recognition device inputs the dynamic coupling matrix into the feature coupling module, divides the dynamic coupling matrix into two components, and uses a three-layer convolutional network to generate scale parameters and translation parameters to obtain the scale parameters and translation parameters of the dynamic coupling matrix.

[0119] The recognition device uses the Householder orthogonal transformation method to decouple the dynamic coupling matrix based on the scale and translation parameters, obtaining a feature decoupling matrix. The recognition device then uses a positive definite constrained linear layer to compress the feature decoupling matrix, obtaining a compressed feature matrix.

[0120] S42: Input the compressed feature matrix into the fully connected module for nonlinear dimensionality reduction to obtain a nonlinear dimensionality reduction vector; based on the nonlinear dimensionality reduction vector and the preset dynamic impact assessment algorithm of forest ecological integrity on water quality, obtain the dynamic impact assessment value.

[0121] In this embodiment, the identification device inputs the compressed feature matrix into the fully connected module for nonlinear dimensionality reduction to obtain a nonlinear dimensionality-reduced vector. Specifically, the identification device inputs the compressed feature matrix into the residual block layer of the fully connected module, extracts higher-order ecological associations using the GeLU activation function, and outputs a hidden state vector. The identification device then inputs the hidden state vector into the affine transformation layer of the fully connected module, compresses the hidden state to a single scalar using a trainable weight matrix, and obtains the nonlinear dimensionality-reduced vector.

[0122] The identification device obtains a dynamic impact assessment value based on the nonlinear dimensionality reduction vector and a preset dynamic impact assessment algorithm for forest ecological integrity on water quality. The algorithm for assessing the dynamic impact of forest ecological integrity on water quality is as follows:

[0123] In the formula, FQ This is a dynamic impact assessment value. It is a non-linear dimensionality reduction vector.

[0124] The identification device identifies water quality treatment units based on the dynamic impact assessment values ​​of each grid cell and constructs a topology map of the water quality treatment units.

[0125] S5: Identify water quality treatment units based on the dynamic impact assessment values ​​of each grid unit, and construct a topology map of water quality treatment units.

[0126] In this embodiment, the identification device identifies water quality treatment units based on the dynamic impact assessment values ​​of each grid cell, constructs a topology map of water quality treatment units, quantifies the dynamic impact of forest ecological integrity on water quality, and improves the accuracy of water quality treatment unit identification.

[0127] Please see Figure 9 , Figure 9 The flowchart of step S5 in the water quality management unit identification method based on forest ecological integrity provided in the first embodiment of this application includes steps S51 to S52, as follows: S51: Using a spatiotemporal clustering method, cluster analysis is performed based on the preset spatial neighborhood radius, temporal neighborhood, minimum number of points, and the dynamic impact assessment value of each grid unit to obtain several clusters.

[0128] In this embodiment, the identification device uses the T-DBSCAN spatiotemporal clustering method to perform clustering analysis based on the preset spatial neighborhood radius, temporal neighborhood, minimum number of points, and the dynamic impact evaluation value of each grid unit to obtain several clusters, wherein each cluster includes several grid units.

[0129] S52: Based on the dynamic impact assessment values ​​of several grid cells in each cluster, extract the target grid cell as the water quality treatment unit from the several grid cells and construct a water quality treatment unit topology map.

[0130] In this embodiment, the identification device extracts target grid cells as water quality treatment units from several grid cells in each cluster based on the dynamic impact assessment values, and constructs a water quality treatment unit topology map. Specifically, the identification device calculates the difference in dynamic impact assessment values ​​of several grid cells in each cluster, that is, it calculates the average of the dynamic impact assessment values ​​of all grid cells within the same cluster, subtracts the average from the dynamic impact assessment value of each grid cell to obtain the difference in dynamic impact assessment values ​​of each cluster. If all grid cells in the cluster are less than a preset deviation constraint value, the cluster is designated as the target cluster, and the soil consistency of each grid cell in the target cluster is obtained. If the soil consistency is greater than a preset soil consistency percentage threshold, the grid cell is extracted as the target grid cell.

[0131] Please refer to Figure 10 , Figure 10This is a schematic diagram of the structure of a water quality management unit identification device based on forest ecological integrity provided in the fifth embodiment of this application. This device can be implemented in whole or in part through software, hardware, or a combination of both. The water quality management unit identification device 10 based on forest ecological integrity includes: The data acquisition module 101 is used to acquire multi-source data and a dynamic coupling model of the forest area. The multi-source data includes forest ecological integrity data and water quality data of several grid cells. The dynamic coupling model includes a feature matrix construction module, a feature fusion module, and an impact assessment module. The matrix construction module 102 is used to input the forest ecological integrity data and the water quality data of each grid cell into the feature matrix construction module to construct the feature matrix, thereby obtaining the forest ecological integrity feature matrix and the water quality feature matrix of each grid cell. The matrix fusion module 103 is used to input the forest ecological integrity feature matrix and the water quality feature matrix of each grid cell into the feature fusion module, and perform feature fusion using a spatiotemporal attention fusion mechanism to obtain the dynamic coupling matrix of each grid cell. The grid impact assessment module 104 is used to input the dynamic coupling matrix of each grid cell into the impact assessment module to conduct a dynamic impact assessment of the forest ecological integrity on water quality and obtain the dynamic impact assessment value of each grid cell. The water quality treatment unit identification module 105 is used to identify water quality treatment units based on the dynamic impact assessment values ​​of each grid unit and to construct a water quality treatment unit topology map.

[0132] In this embodiment, a data acquisition module obtains multi-source data and a dynamic coupling model of the forest area. The multi-source data includes forest ecological integrity data and water quality data for several grid cells. The dynamic coupling model includes a feature matrix construction module, a feature fusion module, and an impact assessment module. The matrix construction module inputs the forest ecological integrity data and water quality data of each grid cell into the feature matrix construction module to construct feature matrices, obtaining forest ecological integrity feature matrices and water quality feature matrices for each grid cell. The matrix fusion module inputs the forest ecological integrity feature matrices and water quality feature matrices of each grid cell into the feature fusion module, and performs feature fusion using a spatiotemporal attention fusion mechanism to obtain the dynamic coupling matrix of each grid cell. The grid impact assessment module inputs the dynamic coupling matrix of each grid cell into the impact assessment module to assess the dynamic impact of forest ecological integrity on water quality, obtaining dynamic impact assessment values ​​for each grid cell. The water quality management unit identification module identifies water quality management units based on the dynamic impact assessment values ​​of each grid cell, constructing a water quality management unit topology map. The dynamic impact of forest ecological integrity on water quality was quantified, improving the accuracy of water quality management unit identification.

[0133] Please refer to Figure 11 , Figure 11 This is a schematic diagram of the structure of a computer device provided in the sixth embodiment of this application. The computer device 11 includes: a processor 111, a memory 112, and a computer program 113 stored in the memory 112 and executable on the processor 111. The computer device can store multiple instructions, which are applicable to the steps of the embodiments shown in the first to fourth embodiments above being loaded and executed by the processor 111. For the specific execution process, please refer to the specific description of the embodiments shown in the first to fourth embodiments, which will not be repeated here.

[0134] The processor 111 may include one or more processing cores. The processor 111 connects to various parts of the server using various interfaces and lines. It executes instructions, programs, code sets, or instruction sets stored in the memory 112, and calls data from the memory 112 to perform various functions and process data in the water quality management unit identification device 10 based on forest ecological integrity. Optionally, the processor 111 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 111 may integrate one or more of the following: a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), and a modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content displayed on the touch screen; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor 111.

[0135] The memory 112 may include random access memory (RAM) or read-only memory. Optionally, the memory 112 may include a non-transitory computer-readable storage medium. The memory 112 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 112 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch instructions), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 112 may also be at least one storage device located remotely from the aforementioned processor 111. This application embodiment also provides a storage medium that can store multiple instructions, which are applicable to being loaded by a processor and executed by the method steps of the first to fourth embodiments described above. For the specific execution process, please refer to the specific description of the first to fourth embodiments, which will not be repeated here.

[0136] Those skilled in the art will understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the functions described above can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. In the above embodiments, the descriptions of each embodiment have different focuses; parts not described or recorded in detail in a certain embodiment can be referred to the relevant descriptions of other embodiments.

[0137] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the algorithm. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0138] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and 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. Furthermore, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Additionally, in the various embodiments of this invention, each functional unit can be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units.

[0139] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms.

[0140] This invention is not limited to the above-described embodiments. If any modifications or variations to this invention do not depart from the spirit and scope of this invention, and if such modifications and variations fall within the scope of the claims and equivalent technologies of this invention, then this invention also intends to include such modifications and variations.

Claims

1. A method for identifying water quality management units based on forest ecological integrity, characterized in that, Includes the following steps: The system obtains multi-source data and a dynamic coupling model for forest areas. The multi-source data includes forest ecological integrity data and water quality data for several raster units. The dynamic coupling model includes a feature matrix construction module, a feature fusion module, and an impact assessment module. The forest ecological integrity data and water quality data of each grid cell are input into the feature matrix construction module to construct the feature matrix, thereby obtaining the forest ecological integrity feature matrix and water quality feature matrix of each grid cell. The forest ecological integrity feature matrix and the water quality feature matrix of each grid cell are input into the feature fusion module, and the spatiotemporal attention fusion mechanism is used to perform feature fusion to obtain the dynamic coupling matrix of each grid cell. The dynamic coupling matrix of each grid cell is input into the impact assessment module to assess the dynamic impact of forest ecological integrity on water quality, and the dynamic impact assessment value of each grid cell is obtained. Water quality treatment units are identified based on the dynamic impact assessment values ​​of each grid cell, and a topology map of water quality treatment units is constructed.

2. The method for identifying water quality management units based on forest ecological integrity according to claim 1, characterized in that, The step of inputting the forest ecological integrity data and water quality data of each grid cell into the feature matrix construction module for feature matrix construction to obtain the forest ecological integrity feature matrix and water quality feature matrix of each grid cell includes the following steps: Local voxel feature extraction is performed on the forest ecological integrity data to obtain local voxel feature representations; Based on the local voxel feature representation, a forest ecological integrity heterogeneous map is constructed to obtain the forest ecological integrity heterogeneous map; using the meta-path mapping method, a matrix is ​​constructed based on the forest ecological integrity heterogeneous map to obtain the forest ecological integrity heterogeneous map matrix. A gated aggregation feature matrix is ​​constructed based on the label vectors of several meta-paths in the forest ecological integrity heterogeneous graph matrix; the gated aggregation feature matrix is ​​then subjected to spatial three-dimensional compression to obtain a compressed feature matrix. Based on the compressed feature matrix and the heterogeneous primitive path data of forest ecological integrity, a matrix is ​​constructed to obtain the forest ecological integrity feature matrix.

3. The method for identifying water quality management units based on forest ecological integrity according to claim 1, characterized in that, The step of inputting the forest ecological integrity data and water quality data of each grid cell into the feature matrix construction module for feature matrix construction to obtain the forest ecological integrity feature matrix and water quality feature matrix of each grid cell includes the following steps: Water quality characteristic analysis is performed on the water quality data to obtain water quality characteristic analysis data, wherein the water quality characteristic analysis data includes a sequence of dissolved oxygen drop patterns in precipitation events, a topological map of water quality parameter hysteresis response, and a multimodal water quality confidence tensor. Obtain the hydrological topology node matrix; calculate spatial features based on the hydrological topology node matrix and the dissolved oxygen drop pattern analysis sequence of precipitation events to obtain spatial feature vectors; perform spatiotemporal dynamic coding based on the water quality parameter hysteresis response topology map and the multimodal water quality confidence tensor to obtain a spatiotemporal dynamic coding matrix. Feature fusion is performed based on the spatial feature vector, spatiotemporal dynamic coding matrix, and dissolved oxygen drop pattern analysis sequence of precipitation events to obtain a fused node feature matrix; based on the fused node feature matrix and a preset graph attention algorithm, adjacent node features are aggregated to obtain an aggregated node feature matrix. The precipitation intensity matrix is ​​obtained by mapping the feature matrix of the aggregated nodes using the hyperspherical projection method. The water quality feature matrix is ​​then constructed based on the precipitation intensity matrix, the nitrogen and phosphorus content data of litter in the forest ecological integrity data, and the preset nitrogen and phosphorus release kinetic model of litter.

4. The method for identifying water quality management units based on forest ecological integrity according to claim 3, characterized in that, The step of performing water quality characteristic analysis on the water quality data to obtain water quality characteristic analysis data includes the following steps: The pH, temperature parameters, and conductivity of the water quality data at each time point are nonlinearly fused and three-dimensionally reconstructed to obtain the three-dimensional latent feature matrix at each time point. Based on the three-dimensional hidden state feature matrix at each time point and the preset rain intensity mask matrix, the event correlation parameters are calculated to obtain the event correlation parameter set at each time point. The event correlation parameter set includes correlation parameters for several events, including correlation parameters for acid dissolution events, correlation parameters for temperature-density stratification events, and correlation parameters for salt-induced oxygen transport events. The event label matrix at each time step is obtained by performing label vector transformation and matrix construction based on the event association parameter set at each time step; the spatial weight distribution is calculated based on the three-dimensional latent feature matrix, the event label matrix, and the preset spatial feature algorithm at each time step to obtain the spatial feature vector at each time step. Spatiotemporal convolutional co-mapping is performed on the event label matrix at each time step to obtain the channel gating signal at each time step; feature aggregation is performed based on the three-dimensional latent feature matrix, the channel gating signal, and the preset spatiotemporal feature aggregation algorithm at each time step to obtain the spatiotemporal aggregated features at each time step. Based on the spatial feature vectors, spatiotemporal aggregation features, event label vectors, and the preset warning vector algorithm at each time point, the warning vectors at each time point are obtained; based on the warning vectors at each time point, a sequence is constructed to obtain the dissolved oxygen drop pattern analysis sequence of precipitation events.

5. The method for identifying water quality management units based on forest ecological integrity according to claim 3, characterized in that, The step of performing water quality characteristic analysis on the water quality data to obtain water quality characteristic analysis data includes the following steps: The nitrogen and phosphorus speciation data at each time point in the water quality data are encoded to obtain nitrogen speciation encoding sequences and phosphorus speciation encoding sequences. The nitrogen and phosphorus form coding sequences are input into a bidirectional gated recurrent unit for effect state feature calculation to obtain the effect state feature sequence; based on the effect state feature sequence and a preset pollutant effect classification algorithm, pollutant effect classification analysis is performed to obtain the pollutant effect category sequence. Based on the pollutant effect category sequence, a multi-temporal coupling analysis of pollutants was performed to construct a water quality lag response topology map.

6. The method for identifying water quality management units based on forest ecological integrity according to claim 3, characterized in that, The step of performing water quality characteristic analysis on the water quality data to obtain water quality characteristic analysis data includes the following steps: Sequences were constructed for the biochemical oxygen demand (BOD) and initial permanganate index at each time point in the water quality data to obtain the initial BOD sequence and the initial permanganate index sequence. The initial biochemical oxygen demand (BOD) sequence is modeled for composite noise and the signal is separated to obtain the real BOD sequence. Asymmetric dilated convolution processing is performed on the real biochemical oxygen demand (BOD) sequence to obtain a reliable BOD sequence. Anomaly masking is performed on the initial permanganate index sequence to obtain a corrected permanganate index sequence; a credibility index matrix is ​​constructed based on the reliable biochemical oxygen demand sequence and the corrected permanganate index sequence to obtain the credibility index matrix. Based on the credibility index matrix and the preset bidirectional attention gating algorithm, forward attention calculation and backward attention calculation are performed to obtain the forward attention vector and backward attention vector at each time step; based on the forward attention vector and backward attention vector at each time step, deep fusion and tensor generation are performed to obtain the multimodal water quality credibility tensor.

7. The method for identifying water quality management units based on forest ecological integrity according to claim 1, characterized in that, The step of using a spatiotemporal attention fusion mechanism to perform feature fusion and obtain the dynamic coupling matrix of each grid unit includes the following steps: Obtain a terrain-weighted mask; perform Hadamard product operation based on the forest ecological integrity feature matrix and the terrain-weighted mask to construct a terrain-enhanced forest ecological field; Hydrological response modeling is performed on the water quality feature matrix to obtain precipitation time-delay sequence; based on the water quality feature matrix, precipitation time-delay sequence and preset time-varying water quality decay feature algorithm, time-varying water quality decay feature matrix is ​​obtained. Based on the terrain-enhanced forest ecological field, the time-varying water quality decay feature matrix, and the preset spatiotemporal attention fusion algorithm, feature fusion is performed to obtain a fused feature matrix, which serves as the dynamic coupling matrix.

8. The method for identifying water quality management units based on forest ecological integrity according to claim 1, characterized in that: The impact assessment module includes a feature coupling module and a fully connected module; The step of inputting the dynamic coupling matrix of each grid cell into the impact assessment module to conduct a dynamic impact assessment of forest ecological integrity on water quality and obtaining the dynamic impact assessment value of each grid cell includes the following steps: The dynamic coupling matrix is ​​input into the feature coupling module for parameter extraction to obtain the scale parameter and translation parameter of the dynamic coupling matrix; based on the scale parameter and translation parameter, the dynamic coupling matrix is ​​decoupled to obtain the feature decoupling matrix; The feature decoupling matrix is ​​subjected to feature compression to obtain a compressed feature matrix; The compressed feature matrix is ​​input into the fully connected module for nonlinear dimensionality reduction to obtain a nonlinear dimensionality reduction vector; based on the nonlinear dimensionality reduction vector and a preset dynamic impact assessment algorithm for forest ecological integrity on water quality, a dynamic impact assessment value is obtained.

9. The method for identifying water quality management units based on forest ecological integrity according to claim 1, characterized in that, The step of identifying water quality treatment units based on the dynamic impact assessment values ​​of each grid cell and constructing a water quality treatment unit topology map includes the following steps: A spatiotemporal clustering method is used to perform cluster analysis based on the preset spatial neighborhood radius, temporal neighborhood, minimum number of points, and the dynamic impact evaluation value of each grid cell to obtain several clusters, wherein each cluster includes several grid cells. Based on the dynamic impact assessment values ​​of several grid cells in each cluster, target grid cells are extracted from several grid cells as water quality treatment units, and a water quality treatment unit topology map is constructed.

10. A water quality treatment unit identification device based on forest ecological integrity, characterized in that, include: The data acquisition module is used to acquire multi-source data and a dynamic coupling model of the forest area. The multi-source data includes forest ecological integrity data and water quality data of several raster units. The dynamic coupling model includes a feature matrix construction module, a feature fusion module, and an impact assessment module. The matrix construction module is used to input the forest ecological integrity data and the water quality data of each grid cell into the feature matrix construction module to construct the feature matrix, thereby obtaining the forest ecological integrity feature matrix and the water quality feature matrix of each grid cell. The matrix fusion module is used to input the forest ecological integrity feature matrix and the water quality feature matrix of each grid cell into the feature fusion module, and perform feature fusion using a spatiotemporal attention fusion mechanism to obtain the dynamic coupling matrix of each grid cell. The grid impact assessment module is used to input the dynamic coupling matrix of each grid cell into the impact assessment module to conduct a dynamic impact assessment of the forest ecological integrity on water quality and obtain the dynamic impact assessment value of each grid cell. The water quality treatment unit identification module is used to identify water quality treatment units based on the dynamic impact assessment values ​​of each grid unit and to construct a water quality treatment unit topology map.

Images