Method and system for matching fish habitat in a watershed based on model simulation

Abstract

The application provides a kind of watershed fish habitat matching method and system based on model simulation, by obtaining the environmental monitoring data of different monitoring points in watershed in preset time period, the watershed environmental scene dataset is constructed in combination with watershed topography basic data;The data set is processed by multi-dimensional feature analysis, and the fish habitat demand feature set is generated;Cross-scale feature matching is carried out based on the environmental feature evolution sequence and the fish habitat demand feature set, and the habitat suitability correlation matrix is constructed;The habitat suitability correlation matrix is processed by time and space planning, and the habitat suitability index distribution map is generated;The optimal habitat space range of target fish is determined, and the ecological protection priority of the optimal habitat space range is sorted, and the watershed fish habitat planning result is generated.The practicability and pertinence of the watershed fish habitat planning result are improved by the application.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a method and system for matching fish habitats in a watershed based on model simulation. Background Technology

[0002] In the fields of aquatic ecological protection and water resource management, analyzing environmental monitoring data within a watershed and combining it with the survival needs of fish to assess habitat suitability can determine the spatial range suitable for fish survival and provide a basis for ecological protection planning. However, existing technologies often fail to effectively integrate the temporal dynamics and spatial distribution characteristics of environmental data in practical applications, resulting in an incomplete overall characterization of the watershed environment. Furthermore, the lack of a systematic characterization of fish habitat needs leads to insufficient analysis of the correlation between environmental supply and biological demand, making it difficult to accurately quantify the impact of environmental factors on habitat suitability at different temporal and spatial scales. Consequently, the generated habitat suitability assessment results fail to accurately reflect the suitability differences of different regional units at different time points, affecting the scientific validity and practicality of fish habitat planning results. Summary of the Invention

[0003] This invention provides a method and system for matching fish habitats in a watershed based on model simulation.

[0004] In a first aspect, embodiments of the present invention provide a method for matching fish habitats in a watershed based on model simulation. The method includes: acquiring environmental monitoring data from different monitoring points within a preset time period in the watershed; constructing a watershed environmental scene dataset containing spatial coordinate information and time series markers by combining watershed topographic and geomorphological basic data; performing multi-dimensional feature analysis processing on the watershed environmental scene dataset to generate an environmental feature evolution sequence describing the changing patterns of watershed environmental elements and a fish habitat demand feature set reflecting the survival needs of fish; performing cross-scale feature matching based on the environmental feature evolution sequence and the fish habitat demand feature set to construct a habitat suitability correlation matrix characterizing the correlation between environmental elements and habitat demand; performing spatiotemporal planning processing on the habitat suitability correlation matrix to generate a habitat suitability index distribution map containing different regional units of the watershed at different time nodes; determining the optimal habitat spatial range for the target fish species based on the habitat suitability index distribution map, and prioritizing the ecological protection of the optimal habitat spatial range to generate a watershed fish habitat planning result.

[0005] Secondly, embodiments of the present invention provide a computer system, including: a memory storing a computer program; and a processor for loading the computer program to implement the model simulation-based watershed fish habitat matching method as described above.

[0006] This invention provides a model-simulation-based watershed fish habitat matching method. By constructing a watershed environmental scene dataset containing spatial coordinate information and time-series labels, it achieves deep integration of environmental monitoring data and topographic data, providing data support for multi-dimensional feature analysis. Through multi-dimensional feature analysis, it simultaneously generates environmental feature evolution sequences and fish habitat demand feature sets, establishing a bidirectional feature correlation between environmental supply and biological demand, providing accurate input for cross-scale matching. Based on cross-scale feature matching, it constructs a habitat suitability correlation matrix, realizing a structured and quantitative expression of the correlation between environmental elements and habitat demand, providing a calculable basis for spatiotemporal planning. Through spatiotemporal planning, it generates a habitat suitability index distribution map, transforming the abstract correlation matrix into an intuitive spatiotemporal distribution index, providing a basis for determining the optimal habitat. Based on the distribution map, it determines the optimal habitat range and prioritizes protection, forming a complete technical link from data collection to planning decisions, improving the practicality and relevance of watershed fish habitat planning results. Attached Figure Description

[0007] Figure 1 This is a flowchart of a watershed fish habitat matching method based on model simulation provided in an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of the composition of a computer system provided in an embodiment of the present invention. Detailed Implementation

[0009] Please see Figure 1 This is a flowchart of a watershed fish habitat matching method based on model simulation provided by an embodiment of the present invention. The method can be executed by a computer system and may include the following steps: Step S100: Obtain environmental monitoring data from different monitoring points within the watershed within a preset time period, and construct a watershed environmental scene dataset containing spatial coordinate information and time series labels by combining the watershed topographic and geomorphological basic data.

[0010] Different monitoring points within the watershed refer to the locations for environmental data collection set up according to a scientific layout within the study watershed. These points need to cover different geographical units within the watershed to ensure data representativeness. The preset time period is a continuously defined time range for data collection and analysis, the length of which needs to be determined based on the timescale characteristics of the research object. Environmental monitoring data are the results obtained through quantitative measurement of watershed environmental elements using professional monitoring equipment, covering key environmental factors such as water temperature, dissolved oxygen, flow velocity, and water depth. Different environmental monitoring data reflect different attribute characteristics of the watershed environment. Basic topographic and geomorphological data of the watershed are fundamental information describing the surface morphology and geological structure of the watershed, including digital elevation models, river vector data, and topographic slope and aspect data, providing underlying support for spatial analysis. Spatial coordinate information is data used to accurately locate the geographical location of monitoring points, usually represented using a geographic coordinate system or a planar coordinate system to ensure the spatial uniqueness of each monitoring point. Time series markers are precise records of the collection time for each environmental monitoring data point, including time dimension information such as year, month, day, hour, and minute, used to construct the time series characteristics of the data. A watershed environmental scene dataset containing spatial coordinate information and time series labels is constructed. The specific implementation method is as follows: First, environmental monitoring data of each monitoring point within a preset time period is collected through the watershed environmental monitoring network, and each data point is accompanied by the collection time information; then, the environmental monitoring data is associated and matched with the spatial coordinate information of the corresponding monitoring point to form a data-space-time ternary structure; finally, the associated data is standardized, organized and stored to form a structured watershed environmental scene dataset, which can simultaneously support environmental feature analysis in both spatial and temporal dimensions.

[0011] Step S200: Perform multi-dimensional feature analysis on the watershed environmental scene dataset to generate an environmental feature evolution sequence describing the changing patterns of watershed environmental elements and a fish habitat demand feature set reflecting the survival needs of fish.

[0012] Multi-dimensional feature parsing refers to the in-depth analysis of watershed environmental scene datasets from multiple dimensions such as time, space, and environmental element types. It involves extracting feature information through the comprehensive application of data statistics, spatiotemporal analysis, and machine learning methods. The environmental feature evolution sequence is a set of feature data arranged chronologically to describe the dynamic changes of watershed environmental elements at different time points, reflecting the temporal evolution trend and periodic characteristics of environmental elements. The fish habitat demand feature set is a set of features describing the environmental element requirements of fish during their survival. It requires combining knowledge of fish biological characteristics and ecological needs to extract environmental element features closely related to fish survival from the environmental feature evolution sequence, including the suitable range and change patterns of environmental elements.

[0013] In one implementation, step S200 may specifically include the following steps S210 to S260: Step S210: Construct a watershed environment map network, using the monitoring points in the watershed environment scene dataset as network nodes, and determining the directed edges between nodes based on the river connectivity and terrain elevation differences in the topographic data. The edge attributes include the river distance and water flow direction between nodes.

[0014] A watershed environment map network is a network structure used to represent the spatial relationships of monitoring points in a watershed environment. It consists of nodes and directed edges. Network nodes correspond to monitoring points in the watershed environment scene dataset, and each node contains attribute information such as the spatial coordinates of the monitoring point and environmental monitoring data. Directed edges represent the spatial relationships between nodes. Their existence and direction are determined by the river connectivity and topographic elevation differences in the topographic data: if two monitoring points are located in connected rivers and have different topographic elevations, a directed edge is established between these two nodes, with the edge pointing from the node with higher elevation to the node with lower elevation, reflecting the natural flow direction of the water. Edge attributes describe the characteristic information of the directed edges. The river distance between nodes refers to the actual distance between the two monitoring points along the river, which can be calculated from the coordinate information in the river vector data; the flow direction refers to the direction in which the directed edge points, consistent with the flow direction determined by the topographic elevation difference. The specific method for constructing the watershed environment map network is as follows: First, each monitoring point in the watershed environment scene dataset is abstracted as a network node, and a unique identifier is assigned to each node; then, based on the river vector data in the topographic and geomorphological basic data, the connectivity of the river channels where each monitoring point is located is determined, and the topographic elevation difference is further calculated for the connected monitoring points; for connected monitoring points with elevation differences, directed edges are established between the corresponding nodes, with the direction of the edges from high to low; next, the river channel distance and water flow direction attributes between the nodes of each directed edge are calculated; finally, all nodes and directed edges are combined to form a complete watershed environment map network, which can intuitively reflect the spatial association structure and water flow direction characteristics of the watershed monitoring points.

[0015] Step S220: Extract node features from the environmental monitoring data corresponding to each node in the watershed environmental map network, arrange the node environmental monitoring data in ascending order according to the time series label to obtain a continuous time axis data sequence, determine the segment length according to the data collection interval, generate adjacent non-overlapping segmented data blocks, calculate the mean level, fluctuation range and trend direction of each data block, and combine them into a node environmental feature vector.

[0016] The environmental monitoring data corresponding to a node refers to all environmental monitoring data collected by each node (monitoring point) in the watershed environmental map network within a preset time period, including information on multiple environmental elements at different times for that point. Node feature extraction is the process of extracting key feature information from the environmental monitoring data of each node, aiming to transform the raw monitoring data into a representative feature vector. The continuous time axis data sequence is a sequence formed by arranging the node environmental monitoring data in ascending order according to the time series label, ensuring that the data is continuously distributed according to the collection time sequence without any time discontinuities. The data collection interval is the time difference between two adjacent environmental monitoring data points. The segment length needs to be determined according to the data collection interval to ensure the temporal rationality of the segmented data blocks. Adjacent non-overlapping segmented data blocks are time series blocks formed by dividing the continuous time axis data sequence according to the segment length. Each data block is temporally continuous and has no overlapping data points. The mean level is the arithmetic mean of all data points within a segmented data block, reflecting the average level of environmental factors during that period. The fluctuation range is the difference between the maximum and minimum values ​​of data points within a segmented data block, reflecting the magnitude of change in environmental factors. The trend direction is the direction of change obtained through trend analysis of data points within the segmented data block; the slope can be determined using linear regression to identify the trend direction. The node environmental feature vector is a vector formed by combining the mean level, fluctuation range, and trend direction characteristics of each segmented data block in chronological order, comprehensively describing the environmental characteristics of the node at different time periods.

[0017] In one implementation, step S220 may specifically include the following steps S221 to S226: Step S221: Arrange the environmental monitoring data of a single node in ascending order according to the time series label to obtain the environmental data sequence on the continuous time axis, so that the data points are distributed without interruption according to the order of collection.

[0018] Environmental monitoring data for a single node refers to all environmental monitoring data collected by a specific node in the watershed environmental map network within a preset time period, encompassing time-series data of multiple environmental elements. Ascending order of time series labels sorts data points from earliest to latest according to their collection time, ensuring the correct temporal sequence. Environmental data sequences on a continuous time axis are sorted time series where data points are continuously distributed in time, with the time interval between adjacent data points equal to the data collection interval, and no time breaks. Uninterrupted distribution of data points according to collection order means that each data point in the sequence (except for the first and last points) has a clearly defined preceding and succeeding data points, with consistent time intervals.

[0019] Step S222: Determine the segment length based on the data acquisition time interval. Starting from the beginning of the data sequence, continuously extract data segments according to the set step size to generate multiple adjacent and non-overlapping segmented data blocks. Each data block contains a preset number of environmental monitoring data points.

[0020] The data acquisition time interval is the time difference between two adjacent environmental monitoring data points and is the fundamental parameter for determining the segment length. The segment length must be determined based on the data acquisition interval, and its duration should reflect the typical change cycle of environmental elements. For example, the segment length can be set based on the daily, weekly, or monthly variation characteristics of environmental elements. The set step size is the movement step size when extracting data segments. To ensure that segmented data blocks are adjacent and non-overlapping, the set step size is equal to the number of data points contained in the segmented data block. The starting position of the data sequence is the position of the first data point in the continuous time-axis environmental data sequence, serving as the starting point for data extraction. Continuous data segment extraction begins from the starting position and proceeds sequentially according to the set step size, forming multiple segmented data blocks. Adjacent and non-overlapping segmented data blocks refer to data blocks that are continuously connected in time, and no data point simultaneously belongs to two data blocks. The preset number of environmental monitoring data points is the total number of data points contained in each segmented data block, determined by the ratio of the segment length to the data acquisition interval. The specific process for generating multiple adjacent and non-overlapping segmented data blocks by determining the segment length based on the data acquisition time interval is as follows: First, determine the data acquisition interval Δt, and set the segment length T (T is an integer multiple of Δt) according to the research requirements; then calculate the preset number N that each segmented data block should contain, N=N=T / Δt; next, starting from the beginning of the data sequence, extract the first N data points as the first segmented data block; then, starting from the next data point after the last data point of the first segmented data block, extract N data points as the second segmented data block, and so on, until the data sequence is completely extracted; finally, verify whether all segmented data blocks meet the requirements of being adjacent and non-overlapping, ensuring that there are no time breaks and data overlaps between data blocks, and that each data block contains a preset number of environmental monitoring data points.

[0021] Step S223: Calculate the arithmetic mean of all data points for each segmented data block as the mean level characteristic of environmental elements in that time period. The mean calculation covers all valid monitoring data points within the data block.

[0022] Each segmented data block is a data set generated in step S222, containing a preset number of environmental monitoring data points. The arithmetic mean is the sum of the values ​​of all data points within the data block divided by the number of data points, used to reflect the average level of environmental elements during that period. Valid monitoring data points refer to monitoring data points within the data block whose values ​​are complete, not missing, and within a reasonable range; the influence of outliers and missing values ​​must be excluded. Mean calculation covering all valid monitoring data points within the data block means that when calculating the arithmetic mean, only valid data points within the data block are used in the calculation, ensuring the accuracy of the mean level characteristics.

[0023] Step S224: Calculate the difference between the maximum and minimum values ​​in the segmented data block as the fluctuation range characteristic of the environmental element. The difference calculation is based on the extreme points of the actual monitoring data within the data block.

[0024] The maximum value in a segmented data block is the highest value among all data points within that block, and the minimum value is the lowest value. The fluctuation range characteristic is the difference between the maximum and minimum values, reflecting the magnitude of change in environmental factors within that time period. The difference calculation based on extreme points of actual monitored data within the data block means that when calculating the fluctuation range, the actual maximum and minimum values ​​within the data block are directly used for calculation, without data interpolation or smoothing, to preserve the fluctuation characteristics of the original data.

[0025] Step S225: Perform linear trend analysis on the segmented data blocks, fit a straight line by the coordinates of the data points, and determine the trend direction based on the positive or negative slope of the straight line. A positive slope indicates an increasing trend of environmental factors, while a negative slope indicates a decreasing trend. The absolute value of the slope reflects the rate of trend change.

[0026] Linear trend analysis of segmented data blocks uses mathematical methods to analyze the changing trends of environmental elements within the data block over time. Data point coordinates refer to the positions of data points in a Cartesian coordinate system with time as the x-axis and environmental element values ​​as the y-axis. The fitted straight line is obtained by performing linear regression on the data point coordinates using methods such as least squares; this line reflects the overall changing trend of the data points. The slope of the fitted line is an indicator of its inclination; its sign determines the trend direction: a positive slope indicates an increasing trend in environmental elements over time, while a negative slope indicates a decreasing trend. The absolute value of the slope reflects the rate of trend change; a larger absolute value indicates a faster rate of change, and a smaller absolute value indicates a slower rate of change.

[0027] Step S226: Combine the mean level feature, fluctuation range feature and trend direction feature of all segmented data blocks of the same node in chronological order to generate a node environment feature vector containing time dynamic characteristics. The vector dimension is consistent with the number of segmented data blocks.

[0028] All segmented data blocks of the same node refer to all segmented data blocks obtained by dividing a continuous time-axis data sequence of a single node. Mean level, fluctuation range, and trend direction are the three key features of each segmented data block, reflecting the average level, magnitude of change, and trend of environmental factors, respectively. Arranging features in chronological order means arranging the features of each segmented data block according to their chronological order on the time axis, ensuring that the chronological order of the feature vectors matches the chronological order of the data sequence. A node environmental feature vector containing time-dynamic characteristics means that the vector reflects the dynamic change of the node's environmental characteristics over time, with each dimension corresponding to a feature combination of a segmented data block. Maintaining consistency between vector dimension and the number of segmented data blocks means that the dimension of the feature vector equals the number of segmented data blocks, with each segmented data block corresponding to one dimension of the vector.

[0029] Step S230: Calculate the environmental element interaction weights of directed edges based on the river channel distance and water flow direction between nodes. Determine the weight value by the ratio of the difference in node environmental monitoring data to the river channel distance, and generate a weighted edge attribute matrix.

[0030] The river channel distance between nodes is one of the directed edge attributes determined in step S210, reflecting the spatial distance between two nodes (monitoring points) along the river channel. The water flow direction is the directional direction of the directed edge, reflecting the direction of water flow from one node to another. The environmental element interaction weight is an indicator used to quantify the degree of mutual influence between environmental elements between nodes; a larger weight value indicates a stronger interaction between environmental elements. Determining the weight value by the ratio of the difference in node environmental monitoring data to the river channel distance means that for two nodes on a directed edge, the absolute value of the difference in their environmental monitoring data at the same time period is calculated, and then divided by the river channel distance between the nodes to obtain the environmental element interaction weight of that directed edge. The weighted edge attribute matrix is ​​a matrix formed by organizing the environmental element interaction weights of all directed edges according to the node connection relationship; the elements in the matrix represent the weight values ​​of the directed edges between corresponding nodes. The specific method for calculating the environmental element interaction weights of directed edges based on the river channel distance and water flow direction between nodes, and generating a weighted edge attribute matrix, is as follows: First, traverse all directed edges in the watershed environmental graph network. For each directed edge, obtain the two nodes it connects (the starting node and the target node). Then, calculate the absolute value of the difference between the environmental monitoring data of the starting node and the target node in the same time period. Next, divide the absolute value of the difference by the river channel distance between the nodes to obtain the environmental element interaction weight of the directed edge. Finally, according to the connection relationship of the nodes, fill the weight values ​​of all directed edges into the corresponding positions of the edge attribute matrix to form a weighted edge attribute matrix. This matrix can quantitatively reflect the degree of interaction influence of environmental elements between nodes.

[0031] Step S240: Traverse each node in the watershed environment graph network. Based on the connection weights between the node and its neighboring nodes in the edge attribute matrix, calculate the weighted node environment feature vectors of each neighboring node with their corresponding connection weights. Aggregate all the weighted neighboring node environment feature vectors and fuse the aggregation result with the node's own node environment feature vector to generate enhanced node features with fused spatial association.

[0032] Traversing each node in the watershed environment graph network means processing all nodes in the network sequentially, ensuring that each node undergoes spatial association feature fusion. An adjacent node is a node directly connected to the current node via a directed edge, including in-degree and out-degree adjacent nodes. Connection weights are the interaction weights of environmental elements between the current node and its adjacent nodes in the edge attribute matrix. Weighted calculation multiplies the adjacent node's environment feature vector by its corresponding connection weight to obtain a weighted adjacent node environment feature vector. Aggregation involves summing or averaging all weighted adjacent node environment feature vectors to obtain the aggregation result. Fusion combines the aggregation result with the current node's own node environment feature vector to form an enhanced node feature with fused spatial associations. This feature simultaneously includes the node's own environmental features and the spatial association features of its adjacent nodes.

[0033] Step S250: Unify the time reference for all enhanced node features. By comparing the overlapping intervals of the time series markers of each node, adjust the time dimension of different node features to generate an environmental feature evolution sequence with a unified time axis.

[0034] All enhanced node features are the node feature vectors containing spatial correlation information generated in step S240. Time base unification refers to adjusting the feature vectors of different nodes to the same time base, ensuring consistency in the time dimension of the feature vectors. Overlapping interval comparison of time series labels for each node refers to comparing the time series labels of different node feature vectors to find their overlapping time intervals, which serve as a unified time base. Adjusting the time dimension of different node features refers to truncating or interpolating the node feature vectors based on the overlapping intervals, making the time dimension of all node feature vectors consistent with the unified time base. An environmental feature evolution sequence with a unified time axis means that after adjustment, the feature vectors of all nodes have the same time axis, enabling environmental feature evolution analysis at a unified time scale. The specific method for unifying the time base of all enhanced node features to generate an environmental feature evolution sequence is as follows: First, collect the time series markers of all enhanced node features and determine their time coverage range; then find the maximum overlap interval of all node time series markers as a unified time base; next, adjust the time dimension of the enhanced node feature vector of each node according to the unified time base, extract the feature data within the overlap interval, or interpolate and fill in the missing time points; finally, arrange all the adjusted node feature vectors in chronological order to generate an environmental feature evolution sequence with a unified time axis.

[0035] Step S260: Input the environmental feature evolution sequence into the fish demand association module, capture the synergistic relationship between environmental elements through multi-layer nonlinear transformation, extract the key element combination related to fish survival activities, and generate a fish habitat demand feature set that characterizes the fish demand pattern of environmental elements.

[0036] The environmental feature evolution sequence is a set of node feature vectors with a unified time axis generated in step S250, containing the temporal dynamic characteristics of environmental elements at each node. The fish demand correlation module is a functional module used to analyze the correlation between environmental elements and fish survival needs, and can be constructed using machine learning models such as neural networks and decision trees. Multi-level nonlinear transformation is the core processing mechanism of the fish demand correlation module, capturing complex collaborative relationships between environmental elements through multi-level nonlinear function transformation, such as mutual promotion or constraint relationships between elements. Extracting key element combinations related to fish survival activities refers to screening element combinations that have a significant impact on fish survival from environmental elements; these combinations are the core content of fish habitat demand. The fish habitat demand feature set, characterizing the fish's demand pattern for environmental elements, is a feature set containing information such as the demand threshold and suitable range of fish for key environmental element combinations, accurately reflecting the fish's habitat demand pattern.

[0037] In one implementation, step S260 may specifically include the following steps S261 to S266: Step S261: Standardize the environmental feature evolution sequence by dividing the deviation of the data value from the sequence mean by the sequence standard deviation, and map the feature value to a preset numerical range to generate a standardized environmental feature sequence.

[0038] An environmental feature evolution sequence is a sequence of data containing the temporal dynamic characteristics of environmental elements at each node. The feature values ​​of different environmental elements may have different dimensions and numerical ranges. Standardization is a crucial step in eliminating the influence of dimensions and unifying the numerical range. It involves converting the feature values ​​into standardized values ​​with a mean of 0 and a standard deviation of 1 by dividing the deviation of the data value from the sequence mean by the sequence standard deviation (i.e., the Z-score standardization method). The preset numerical range is a range of values ​​set according to the needs of subsequent model processing, such as [0,1] or [-1,1]. Further linear transformations can map the standardized values ​​to the preset numerical range. Generating a standardized environmental feature sequence refers to the environmental feature evolution sequence after standardization. The feature values ​​in this sequence have uniform dimensions and numerical ranges, facilitating subsequent model processing.

[0039] Step S262: Input the standardized environmental feature sequence into the feature interaction layer of the fish demand association module, capture the synergistic relationship between different environmental elements through multi-layer nonlinear transformation, and generate a feature interaction map containing element interaction information.

[0040] The standardized environmental feature sequence is the environmental feature sequence with uniform dimensions and numerical range generated in step S261. The feature interaction layer of the fish demand association module is a hierarchical structure specifically designed to analyze the interaction relationships between environmental elements. It can be implemented using architectures such as attention mechanisms and graph neural networks. Multi-layer nonlinear transformation is the core processing method of the feature interaction layer. By transforming the input features through multi-layer nonlinear functions (such as ReLU, Sigmoid, etc.), it can capture complex nonlinear synergistic relationships between environmental elements, such as the mutual influence between water temperature and dissolved oxygen. The feature interaction graph containing element interaction information treats environmental elements as graph nodes and synergistic relationships between elements as graph edges. The graph structure data generated after processing by the feature interaction layer shows that the weights of the edges in the graph reflect the strength of the synergistic effects between elements.

[0041] Step S263: Screen key elements in the feature interaction graph, retain the combination of elements that significantly affect fish survival activities through feature importance assessment, eliminate interference from minor elements, and generate a subset of key element interactions.

[0042] The feature interaction graph is a graph structure data containing information on the interactions of environmental elements generated in step S262. Key element screening is the process of selecting combinations of environmental elements that have a significant impact on fish survival activities from the feature interaction graph. Feature importance assessment is the core method of key element screening, and can employ model-based methods (such as feature importance in decision trees, weight analysis in neural networks) or statistical methods (such as correlation analysis, analysis of variance) to assess the degree of influence of each environmental element and element combination on fish survival activities. Retaining element combinations that significantly impact fish survival activities means retaining element combinations whose feature importance assessment scores are higher than a preset threshold. Eliminating minor element interference means excluding element combinations with low influence to reduce redundant information. The key element interaction subset is a subgraph structure containing combinations of elements with significant impact after screening, focusing on the interaction relationships of environmental elements crucial to fish survival.

[0043] Step S264: Extract fish demand patterns based on key element interaction subsets, determine the suitable element combination range for fish survival by analyzing the value range and variation pattern of element combinations, and generate demand pattern representation.

[0044] The key element interaction subset is the subgraph structure selected in step S263 that contains combinations of elements with significant influence. Fish demand patterns refer to the patterns of fish demand for combinations of environmental elements, including suitable value ranges and variation patterns. Analyzing the value range and variation patterns of element combinations involves statistically analyzing the element combinations in the key element interaction subset to determine their typical value ranges and temporal variation patterns in the fish's living environment. Determining the suitable element combination range for fish survival involves identifying the suitable range for fish survival based on fish ecological research findings and historical monitoring data, combined with the value range and variation patterns of element combinations. The element combinations within this range can meet the fish's survival needs. Demand pattern representation is the result of abstracting and representing fish demand patterns, which can be described using interval ranges, functional relationships, etc., to describe the characteristics of fish demand for combinations of environmental elements. The specific process of extracting fish demand patterns based on key element interaction subsets and generating demand pattern representations is as follows: First, historical monitoring data of key element combinations in the fish's living environment are collected; then, the value range and variation patterns of these data are analyzed to identify typical element combination patterns; next, combined with the biological characteristics of fish, the suitable range for fish survival under each pattern is determined; finally, the suitable range and variation patterns are abstracted and represented to generate demand pattern representations, which can accurately reflect the fish's demand patterns for key environmental element combinations.

[0045] Step S265: Quantify the demand pattern representation, convert the suitable element combination range into a numerical demand threshold range, including the lower limit, upper limit and optimal value center of the element, and generate quantitative demand characteristics.

[0046] The demand pattern representation is an abstract representation of the fish's demand pattern for environmental element combinations generated in step S264. Quantification is the process of converting the abstract demand pattern representation into a calculable numerical index. The suitable element combination interval is the range of suitable environmental element combinations for fish survival in the demand pattern representation. The numerical demand threshold range converts the suitable interval into a specific numerical range, including the lower limit (minimum value of the interval), upper limit (maximum value of the interval), and optimal value center (the numerical point within the interval most favorable for fish survival). Quantified demand features are a set of features containing the demand threshold ranges for each environmental element; each feature corresponds to the demand threshold range for one environmental element. The specific method for quantifying the demand pattern representation to generate quantitative demand features is as follows: First, analyze each suitable element combination interval in the demand pattern representation to determine its value range; then, determine the lower and upper limits of the element values ​​based on the interval range; next, determine the optimal value center by analyzing the optimal conditions for fish survival within the interval; finally, combine the lower limit, upper limit, and optimal value center of each environmental element into a numerical demand threshold range to form a quantitative demand feature. This feature can provide a quantitative demand standard for habitat suitability assessment.

[0047] Step S266: Classify and organize all quantitative demand features according to environmental element types to generate a fish habitat demand feature set that includes element type identifiers, demand threshold ranges, and descriptions of collaborative relationships.

[0048] All quantified demand features are the feature sets generated in step S265, containing the demand threshold ranges for each environmental element. Categorizing by environmental element type refers to grouping the quantified demand features according to the type of environmental element (such as water temperature, dissolved oxygen, flow rate, etc.) to ensure centralized management of features of the same type. Element type identifiers are labels used to distinguish different types of environmental elements, such as "water temperature" and "dissolved oxygen." Demand threshold ranges are the lower limit, upper limit, and optimal value center of the element values ​​included in the quantified demand features. Collaborative relationship descriptions are textual or numerical descriptions of the collaborative relationships between elements in the key element interaction subset, reflecting the mutual influence between elements. The fish habitat demand feature set is the categorized and organized feature set, containing the demand threshold ranges and collaborative relationship descriptions for all environmental elements. The specific method for generating a fish habitat demand feature set by classifying and organizing all quantitative demand features according to environmental element types is as follows: First, the quantitative demand features are grouped according to environmental element types, and a corresponding element type identifier is assigned to each group of features; then, the demand threshold range of each element is organized within each group of features; next, the synergistic relationship between the element and other elements is extracted from the key element interaction subset to form a synergistic relationship description; finally, the features of all groups are arranged in order of element type to generate a complete fish habitat demand feature set. This feature set includes element type identifiers, demand threshold ranges, and synergistic relationship descriptions, providing a comprehensive demand basis for matching fish habitats in the watershed.

[0049] Step S300: Based on the environmental feature evolution sequence and the fish habitat demand feature set, perform cross-scale feature matching to construct a habitat suitability correlation matrix that characterizes the relationship between environmental elements and habitat demand.

[0050] The environmental feature evolution sequence is a set of node feature vectors with a unified time axis generated in step S250, containing the temporal dynamic characteristics of environmental elements at each node. The fish habitat demand feature set is a set of features generated in step S260 containing patterns of fish demand for environmental elements. Cross-scale feature matching refers to matching and analyzing the environmental element features in the environmental feature evolution sequence with the demand features in the fish habitat demand feature set at different spatial and temporal scales to identify the correlation between the two. The habitat suitability correlation matrix is ​​a matrix used to characterize the correlation between environmental elements and habitat demand. The rows of the matrix represent environmental nodes or environmental elements, the columns represent fish habitat demand elements, and the matrix elements represent the degree of correlation between the corresponding environmental element and the demand element.

[0051] In one implementation, step S300 may specifically include the following steps S310 to S370: Step S310: Align the graph structure of the environmental feature evolution sequence and the fish habitat demand feature set, match the graph nodes of the environmental feature evolution sequence with the demand element nodes of the fish habitat demand feature set by type, and generate cross-scale graph node pairs so that the node types and quantities correspond one-to-one.

[0052] The graph nodes of the environmental feature evolution sequence refer to the monitoring point nodes in the watershed environmental map network, with each node corresponding to a set of environmental elements. The demand element nodes of the fish habitat demand feature set refer to the environmental element nodes in the fish habitat demand feature set, with each node corresponding to a fish demand element. Graph structure alignment refers to matching the graph structure of the environmental feature evolution sequence with the graph structure of the fish habitat demand feature set, ensuring consistency in node type and quantity. Type matching refers to matching nodes of the same type in the environmental feature evolution sequence with those in the fish habitat demand feature set; for example, matching the "water temperature" node in the environmental feature evolution sequence with the "water temperature" demand node in the fish habitat demand feature set. Cross-scale graph node pairs are the set of node pairs formed after matching; each node pair contains one environmental graph node and one demand element node, and the type and quantity of node pairs must correspond one-to-one.

[0053] In one implementation, step S310 may specifically include the following steps S311 to S316: Step S311: Parse the graph node type labels of the environmental feature evolution sequence, identify the included environmental element node types, each node type corresponds to an independent environmental element, and record the node type and quantity information.

[0054] The graph node type labels of an environmental feature evolution sequence are labels that identify the type of environmental element represented by each graph node, such as "water temperature" or "dissolved oxygen." Parsing the graph node type labels refers to extracting node type label information from the graph structure data of the environmental feature evolution sequence. Identifying the included environmental element node types means determining all environmental element types contained in the sequence based on the node type labels; each node type corresponds to an independent environmental element (such as water temperature or flow velocity). Recording node type and quantity information involves counting the number of nodes for each type of environmental element, forming a node type-quantity lookup table.

[0055] Step S312: Parse the demand element node labels of the fish habitat demand feature set, identify the types of demand elements included, and ensure that the types of demand elements are consistent with the types of environmental element nodes in the environmental feature evolution sequence.

[0056] The demand element node labels in the fish habitat demand feature set are tag information that identifies the type of fish demand represented by the demand element node. Parsing the demand element node labels refers to extracting the demand element node type labels from the graph structure data of the fish habitat demand feature set. Identifying the included demand element types means determining all demand element types included in the demand feature set based on the labels. Ensuring that the demand element types are consistent with the environmental element node types in the environmental feature evolution sequence means that the demand element types must completely match the environmental element node types in the environmental feature evolution sequence; there should be no situation where a demand element type has no corresponding environmental element node in the environmental feature evolution sequence, and vice versa. The specific method for parsing the demand element node labels of the fish habitat demand feature set to ensure that the demand element type is consistent with the environmental element node type is as follows: First, traverse all demand element nodes of the fish habitat demand feature set and extract the type label of each node; then identify the demand element type and compare it with the environmental element node type recorded in step S311; if there is a situation where the demand element type does not appear in the environmental element node type, or the environmental element node type does not appear in the demand element type, data correction is required to ensure that the two types are completely consistent; finally, confirm that the demand element type and the environmental element node type are consistent, providing a type basis for node matching.

[0057] Step S313: Perform type matching on environmental element nodes and demand element nodes, and bind environmental nodes and demand nodes of the same element type to obtain element node pairs. The binding process is based on node type label matching.

[0058] Environmental element nodes are graph nodes in the environmental feature evolution sequence, while demand element nodes are graph nodes in the set of fish habitat demand features. Type matching refers to matching environmental element nodes and demand element nodes according to the same element type based on node type labels. One-to-one binding means that each environmental element node is bound to only one demand element node of the same type, and each demand element node is also bound to only one environmental element node of the same type, forming a one-to-one element node pair. The binding process based on node type label matching means that the matching is based on the complete consistency of node type labels; only nodes with the same labels can be bound. The specific method for type matching of environmental element nodes and demand element nodes to obtain element node pairs is as follows: First, according to the node type labels, environmental element nodes and demand element nodes are grouped by type; then, for each type group, environmental element nodes and demand element nodes are bound one-to-one, for example, an environmental node of type "water temperature" is bound to a demand node of type "water temperature"; finally, a set of element node pairs is formed, each node pair containing one environmental element node and one demand element node, the type of the node pair is consistent with the label, and the binding process is strictly based on node type label matching.

[0059] Step S314: For each element node pair, extract the feature vector of the environmental node in the environmental feature evolution sequence to obtain the environmental node feature vector. The environmental node feature vector contains the temporal dynamic features and spatial correlation features of the node.

[0060] Each element node pair is a node pair generated in step S313, containing one environmental element node and one demand element node. The environmental node feature vector is the feature vector of the corresponding environmental element node in the environmental feature evolution sequence, containing the node's temporal dynamic features and spatial correlation features. Temporal dynamic features refer to the characteristics of environmental elements changing over time, such as mean level, fluctuation range, and trend direction; spatial correlation features refer to the interaction features between this node and its neighboring nodes' environmental elements, composed of the spatial correlation component in the enhanced node features. Extracting the environmental node feature vector means finding and extracting the corresponding feature vector from the environmental feature evolution sequence based on the environmental node identifier in the element node pair.

[0061] Step S315: Extract the demand feature vector of the demand node from the fish habitat demand feature set to obtain the demand node feature vector. The demand feature vector contains the demand threshold range and collaborative relationship features of the element.

[0062] The demand node feature vector is the feature vector of the corresponding demand element node in the fish habitat demand feature set, containing the demand threshold range and collaborative relationship features of that element. The demand threshold range is the numerical demand threshold range generated in step S265, including the lower limit, upper limit, and optimal value center of the element; the collaborative relationship features are the numerical representation of the collaborative relationship description generated in step S266, reflecting the strength of the collaborative effect between this element and other elements. Extracting the demand feature vector of a demand node refers to finding and extracting the corresponding feature vector from the fish habitat demand feature set based on the demand node identifier in the element node pair. The specific method for extracting the demand feature vector of this demand node in the fish habitat demand feature set is as follows: first, obtain the demand node identifier in the element node pair; then, locate the corresponding feature vector in the fish habitat demand feature set based on the node identifier; next, confirm that the feature vector contains the demand threshold range and collaborative relationship features; finally, extract the feature vector as the demand node feature vector.

[0063] Step S316: Combine the environmental node feature vector and the demand node feature vector into cross-scale graph node pairs. Each node pair contains feature type identifier, environmental feature data and demand feature data. The number of node pairs is consistent with the number of feature types.

[0064] A cross-scale map node pair is a combined set of node pairs. Each node pair contains three parts of information: feature type identifier (i.e., the type label of the node pair), environmental feature data (i.e., the environmental node feature vector), and demand feature data (i.e., the demand node feature vector). Maintaining the consistency between the number of node pairs and the number of feature types means that the number of node pairs in the cross-scale map equals the number of environmental feature node types, ensuring that each feature type has a corresponding node pair.

[0065] Step S320: Perform feature scale transformation on the cross-scale graph node pairs. By adjusting the node feature dimensions, the environmental features and the demand features have the same feature space dimension, generating cross-scale feature pairs with unified scale.

[0066] Cross-scale graph node pairs are the set of node pairs containing environmental feature data and demand feature data generated in step S316. Feature scaling refers to adjusting the dimensions of environmental and demand features to make them have the same feature space dimension, so as to facilitate feature matching. Node feature dimension adjustment is achieved by adjusting the dimensions of environmental and demand feature vectors through methods such as feature selection, feature dimensionality reduction, or feature dimensionality increase. Feature space dimension refers to the number of features contained in a feature vector; the same feature space dimension means that environmental and demand features have the same number of feature dimensions. Scale-unified cross-scale feature pairs are cross-scale graph node pairs where, after feature scaling, environmental and demand features have the same feature space dimension. The specific method for performing feature scale transformation on cross-scale graph node pairs to generate scale-uniform cross-scale feature pairs is as follows: First, determine the dimension of the target feature space, which must simultaneously meet the dimensional requirements of both environmental and demand features; then, adjust the dimensions of the environmental feature vector and demand feature vector in each cross-scale graph node pair. If the environmental feature dimension is higher than the target dimension, use dimensionality reduction methods such as principal component analysis to reduce the dimension; if the demand feature dimension is lower than the target dimension, use dimensionality increase methods such as feature expansion to increase the dimension; finally, verify whether the dimensions of the adjusted environmental feature vector and demand feature vector are consistent with the target dimension, generating scale-uniform cross-scale feature pairs that can support subsequent feature matching operations.

[0067] Step S330: Construct a cross-scale feature matching network, which includes an input layer, a feature fusion layer, a correlation calculation layer, and an output layer.

[0068] Cross-scale feature matching networks are neural network models used to achieve cross-scale matching of environmental features and demand features. They consist of multiple functional layers. The input layer, the first layer, receives cross-scale feature pairs of uniform scale as input data; the number of neurons in the input layer is consistent with the dimension of the feature space. The feature fusion layer, an intermediate layer, fuses the input environmental and demand features, capturing the correlation between them. Fusion methods such as concatenation, addition, or attention mechanisms can be used. The correlation calculation layer, the core layer, calculates the correlation between the fused features and the demand features. Methods such as cosine similarity, Euclidean distance, or inner product operations can be used to evaluate the feature matching degree. The output layer, the last layer, outputs the correlation calculation result. The output layer has only one neuron, and the output value represents the correlation score between the environmental and demand features.

[0069] Step S340: Input the cross-scale feature pairs into the input layer of the cross-scale feature matching network, and map the feature values ​​to the same numerical range through linear transformation of the feature values ​​to generate standardized cross-scale feature pairs.

[0070] Specifically, the environmental feature vector and the demand feature vector in the cross-scale feature pair can be extracted first. Then, for each feature vector, the minimum and maximum values ​​of its feature values ​​are calculated. Next, the feature values ​​are mapped to the target numerical range through a linear transformation formula. For example, for the feature value x, the transformed feature value x' = (x-min) / (max-min) maps the feature value to the [0,1] range. Finally, the transformed environmental feature vector and demand feature vector are combined into a standardized cross-scale feature pair and input into the input layer of the cross-scale feature matching network to provide standardized data for subsequent feature fusion and correlation calculation.

[0071] Step S350: Based on the feature fusion layer, the standardized cross-scale feature pairs are processed, and the correlation model between environmental elements and demand elements is constructed through node feature splicing and nonlinear transformation to generate fused feature vectors.

[0072] The feature fusion layer is an intermediate layer in the cross-scale feature matching network, used to process standardized cross-scale feature pairs. Node feature concatenation combines the environmental and demand feature vectors from the standardized cross-scale feature pairs along their feature dimensions, forming a longer feature vector whose dimension is the sum of the environmental and demand feature dimensions. Nonlinear transformation uses nonlinear activation functions (such as ReLU and Sigmoid) to transform the concatenated feature vector, capturing the nonlinear relationship between environmental and demand elements. The relationship model between environmental and demand elements is the model structure formed after processing by the feature fusion layer, reflecting the relationship between the two. The fused feature vector is the feature vector generated after processing by the feature fusion layer, containing fused information from environmental and demand features; its dimension is determined by the concatenated dimension and the parameters of the nonlinear transformation.

[0073] Step S360: Based on the correlation calculation layer, perform similarity evaluation on the fused feature vectors, determine the matching degree between environmental elements and demand elements through vector inner product operation, and generate correlation score.

[0074] The correlation calculation layer is the core layer in the cross-scale feature matching network, used to evaluate the similarity of the fused feature vectors. Similarity evaluation measures the degree of similarity between the fused feature vector and the demand feature vector; a higher similarity indicates a stronger match between the environmental and demand elements. The vector dot product operation assesses similarity by calculating the dot product value between the fused and demand feature vectors; a larger dot product value indicates a more consistent direction between the two vectors, resulting in higher similarity. The degree of matching refers to the extent to which the environmental element features satisfy the fish habitat demand features; a higher degree of matching leads to a higher correlation score. The correlation score is the quantitative result of the similarity evaluation, representing the degree of match between the environmental and demand elements.

[0075] In one implementation, step S360 may specifically include the following steps S361 to S366: Step S361: Perform dimension normalization on the fused feature vector to make the vector dimension consistent with the preset correlation calculation dimension. The normalization process adjusts the vector length through feature mapping to maintain the integrity of feature information.

[0076] Dimensional normalization refers to adjusting the dimensions of the fused feature vector to match the preset correlation calculation dimension, ensuring the feasibility of subsequent inner product operations. The preset correlation calculation dimension is a vector dimension pre-defined according to the correlation calculation requirements and must be consistent with the dimension of the required feature vector. Feature mapping adjusts the dimension of the fused feature vector to the target dimension through linear or nonlinear transformations, such as using a fully connected layer to transform the dimensions of the fused feature vector. Maintaining the integrity of feature information means preserving as much key information as possible in the fused feature vector during the dimension normalization process, avoiding information loss.

[0077] Step S362: Calculate the inner product of the fused feature vector and the required feature vector. The inner product operation covers all dimensions of the vector, and the result reflects the consistency of the directions of the two vectors.

[0078] The fused feature vector and the required feature vector are feature vectors with the same dimensions after dimensionality normalization. The inner product operation covers all dimensions of the vector, meaning that all elements of the vector are used in the calculation, ensuring that the result comprehensively reflects the overall similarity between the two vectors. The result reflects the directional consistency of the two vectors because the magnitude of the inner product is proportional to the cosine of the angle between the two vectors; a larger inner product value indicates a more consistent direction and higher similarity between the two vectors.

[0079] Step S363: Normalize the inner product value, calibrate it by the maximum inner product value, and map the result to the 0-1 interval. The normalized value represents the degree of matching between environmental elements and demand elements.

[0080] Normalization transforms inner product values ​​into a uniform range for comparison and evaluation. Calibration using the maximum inner product value means dividing the inner product by the maximum value to obtain the normalized result. Mapping to the 0-1 interval means the normalized result ranges from 0 to 1, with 0 representing the lowest matching degree and 1 representing the highest. The normalized value indicates the degree of matching between environmental and demand elements; a higher normalized value indicates a more consistent match between the environmental element characteristics and the fish habitat demand characteristics, signifying a higher degree of matching.

[0081] Step S364: Determine the correlation score based on the normalized inner product value. The higher the score, the more the environmental elements meet the needs of fish. The score calculation is based on the direct conversion of the normalization result.

[0082] The correlation score is an indicator used to quantify the degree of matching between environmental elements and fish habitat requirements. A higher score indicates that the environmental element better meets the needs of fish. The correlation score is positively correlated with the normalized inner product value; a higher score indicates that the environmental element characteristics better meet the habitat requirements of fish. The score calculation is based on a direct conversion of the normalization result, meaning the correlation score equals the normalized inner product value, without requiring additional conversion calculations.

[0083] Step S365: Validate the relevance scores, remove invalid scores caused by missing or abnormal data, and ensure that the scores reflect the true matching relationship.

[0084] The correlation score is an indicator of the degree of matching between environmental elements and demand elements as determined in step S364. Validity verification refers to checking whether the correlation score is reasonable and whether it can truly reflect the matching relationship between environmental elements and demand elements. Data missing or anomalies refer to score anomalies caused by data missingness, noise interference, or model errors during feature extraction, fusion, or calculation. Eliminating invalid scores means excluding correlation scores that do not meet the validity criteria and not including them in subsequent habitat suitability assessments. Ensuring the score reflects the true matching relationship means that the verified correlation score can accurately reflect the actual degree of matching between environmental elements and fish demand.

[0085] Step S366: Record the verified correlation scores in the order of feature node pairs, and generate a correlation score list containing the matching degree of all node pairs. The list order is consistent with the node pair order of the cross-scale map.

[0086] The correlation score list is a list that records the verified correlation scores in the order of feature node pairs. Each element in the list corresponds to the correlation score of a feature node pair. The list order must match the node pair order of the cross-scale map to ensure that each score accurately corresponds to its corresponding node pair.

[0087] Step S370: Arrange the correlation scores of all cross-scale map node pairs according to the order of environmental map network nodes and the order of demand feature set elements to generate a two-dimensional matrix form of habitat suitability correlation matrix. The matrix rows represent environmental nodes, the columns represent demand elements, and the elements represent the correlation scores between the corresponding nodes and elements.

[0088] The order of nodes in the environmental map network refers to the arrangement of environmental nodes in the watershed environmental map network. The order of elements in the demand feature set refers to the arrangement of demand elements in the fish habitat demand feature set. The habitat suitability correlation matrix is ​​a two-dimensional matrix of correlation scores. The rows of the matrix correspond to environmental nodes in the environmental map network, and the columns correspond to demand elements in the demand feature set. Each element in the matrix represents the correlation score between the corresponding environmental node and the demand element.

[0089] Step S400: Perform spatiotemporal planning processing on the habitat suitability correlation matrix to generate a distribution map of habitat suitability index for different regional units of the watershed at different time points.

[0090] In one implementation, step S400 may specifically include the following steps S410 to S450: Step S410: Convert the habitat suitability correlation matrix into a spatiotemporal graph network. Nodes represent the combination of spatial regional units and time nodes within the watershed, and edges represent the correlation between adjacent spatiotemporal units. Edge attributes include the comprehensive spatiotemporal distance index and the index transmission coefficient.

[0091] The habitat suitability correlation matrix is ​​a two-dimensional matrix containing the correlation scores between environmental nodes and demand elements. The spatiotemporal graph network is a graph network structure that combines spatial and temporal dimensions to represent the spatiotemporal correlations of habitat suitability within a watershed. Nodes represent combinations of spatial regional units and temporal nodes within the watershed; each node corresponds to a spatial region and a temporal node, and the node identifier contains information about both the spatial regional unit and the temporal node. Edges represent the correlations between adjacent spatiotemporal units. Adjacent spatiotemporal units include spatially adjacent regional units or spatiotemporal units corresponding to consecutive temporal nodes. Edges can be directed or undirected, depending on the direction of the correlation. Edge attributes include a comprehensive spatiotemporal distance index and an index transmission coefficient. The comprehensive spatiotemporal distance index is an index calculated by combining spatial and temporal distances, reflecting the spatiotemporal proximity of two spatiotemporal units. The index transmission coefficient is a coefficient reflecting the degree of mutual influence between the suitability indices of two spatiotemporal units; a larger coefficient indicates a stronger mutual influence.

[0092] In one implementation, step S410 may specifically include the following steps S411 to S416: Step S411: Analyze the basic topographic and geomorphological data of the watershed, and divide the watershed into regular regional units with fixed spatial resolution. Each regional unit has unique spatial coordinates and boundary range.

[0093] Watershed topographic and geomorphological baseline data are fundamental data describing the surface morphology of a watershed, including digital elevation models and river vector data. Analyzing this baseline data involves extracting information such as watershed boundaries and topographic features to provide a basis for regional unit division. Fixed spatial resolution means that the size of the regional unit is fixed, such as 1km × 1km or 500m × 500m; the resolution is determined based on research needs and data accuracy. Regular regional units are those with regular shapes, such as squares or regular hexagons; regular shapes facilitate spatial analysis and calculation. Unique spatial coordinates and boundary extent mean that each regional unit has a unique geographic coordinate identifier and a clearly defined boundary extent, ensuring the uniqueness of its spatial location.

[0094] Step S412: Divide the entire research period into equally spaced time nodes according to the preset time period. Each time node corresponds to a fixed time scale, and the time nodes cover the entire preset time period.

[0095] The preset time period is the time range set in step S100 for data collection and analysis. Equally spaced time nodes mean dividing the preset time period into multiple time nodes of equal length, such as dividing a year into 12 monthly time nodes, each node corresponding to one month. Fixed time scale means that each time node corresponds to a specific point in time or time period, such as "January 2022," "February 2022," etc. Time nodes covering the entire preset time period means that the sum of the time ranges of all time nodes equals the preset time period, ensuring no time breaks.

[0096] Step S413: Combine the regional units with the time nodes to obtain spatiotemporal units. Each spatiotemporal unit serves as a node in the spatiotemporal graph network, and the node identifier contains the coordinates of the regional unit and the information of the time node.

[0097] A spatiotemporal unit is a combination of a regional unit and a time node. Each spatiotemporal unit corresponds to a spatial region and a time node, representing the habitat status of the watershed within that spatial region and at that time node. Nodes in a spatiotemporal graph network refer to spatiotemporal units as the basic building blocks of the graph network; each node corresponds to one spatiotemporal unit. Node identifiers containing regional unit coordinates and time node information mean that each node's identifier consists of the spatial coordinates of the regional unit and the time scale of the time node, for example, "Regional Unit (100, 200) _ Time Node (January 2022)", ensuring the uniqueness of node identifiers.

[0098] Step S414: Determine the relationship between adjacent spatiotemporal units. Spatially adjacent regional units or spatiotemporally consecutive time nodes are considered as adjacent nodes, and directed edges are generated between nodes.

[0099] Adjacent spatiotemporal units refer to spatiotemporal units that are spatially adjacent or temporally continuous. Spatially adjacent regional units are those that share a common boundary or are less than a threshold distance. Temporally continuous time nodes are those that are consecutive in time (e.g., January and February), and their corresponding spatiotemporal units are considered temporally adjacent. Adjacent nodes are spatiotemporal unit nodes that have a relationship. A directed edge is an edge connecting adjacent nodes, and the direction of the edge indicates the direction of the relationship, such as from an earlier node in time to a later node, or from an upstream node in space to a downstream node. The specific method for determining the relationship between adjacent spatiotemporal units and generating directed edges is as follows: First, traverse all spatiotemporal unit nodes; then, for each node, check the spatiotemporal units corresponding to its spatially adjacent regional units (the same time node) and the spatiotemporal units corresponding to its temporally continuous time nodes (the same regional unit); then, treat these spatiotemporal units as adjacent nodes; finally, generate directed edges between adjacent nodes, with the direction of the edges determined according to the direction of the relationship (e.g., temporal direction or spatial flow direction), forming the edge structure of the spatiotemporal graph network.

[0100] Step S415: Calculate the spatial distance and temporal distance between adjacent nodes, and normalize the spatial distance and temporal distance respectively to obtain normalized spatial distance value and normalized temporal distance value; calculate the comprehensive spatiotemporal distance index based on the normalized spatial distance value and normalized temporal distance value.

[0101] Spatial distance between adjacent nodes refers to the spatial distance between regional units corresponding to two adjacent spatiotemporal units, which can be calculated using methods such as Euclidean distance or Manhattan distance. Temporal distance refers to the time interval between time nodes corresponding to two adjacent spatiotemporal units; for example, the time distance between two consecutive monthly nodes is one month. Normalization methods can include min-max standardization or Z-score standardization. The normalized spatial distance value and the normalized temporal distance value are distance values ​​after normalization processing, with a value range of, for example, [0,1]. The comprehensive spatiotemporal distance index is an indicator that combines spatial distance and temporal distance, reflecting the spatiotemporal proximity of two spatiotemporal units. It can be calculated using methods such as weighted summation or geometric mean; for example, the comprehensive spatiotemporal distance index = α × normalized spatial distance value + (1-α) × normalized temporal distance value, where α is a weighting coefficient.

[0102] Step S416: Determine the exponential transfer coefficient of the edge based on the comprehensive spatiotemporal distance index and the environmental similarity of the regional unit. The smaller the distance and the higher the similarity of the node pair, the larger the transfer coefficient. The value of the exponential transfer coefficient reflects the degree of mutual influence of the suitability index between nodes.

[0103] The environmental similarity of a regional unit refers to the degree of similarity in environmental features between two adjacent spatiotemporal units, which can be calculated through the similarity of environmental feature vectors (such as cosine similarity). The edge's exponential transmission coefficient reflects the degree of mutual influence between the suitability indices of two spatiotemporal unit nodes; a larger transmission coefficient indicates a stronger influence of one node's suitability index on another. A larger transmission coefficient indicates a node pair with a smaller overall spatiotemporal distance index (closer in spatiotemporal proximity) and higher environmental similarity. The exponential transmission coefficient value reflects the degree of mutual influence between the suitability indices of nodes; the magnitude of the transmission coefficient directly determines the strength of the influence of the suitability indices between nodes. A positive transmission coefficient indicates a positive influence, and a negative coefficient indicates a negative influence. The specific method for determining the exponential transfer coefficient of an edge based on the comprehensive spatiotemporal distance index and environmental similarity is as follows: First, calculate the environmental similarity of the regional units corresponding to adjacent nodes; then, calculate the exponential transfer coefficient using a functional relationship (such as an exponential function or a linear function) based on the comprehensive spatiotemporal distance index and environmental similarity, for example, transfer coefficient = exp(-β × comprehensive spatiotemporal distance index) × environmental similarity, where β is the attenuation coefficient; finally, ensure that the smaller the distance and the higher the similarity between nodes, the larger the transfer coefficient. The value of the exponential transfer coefficient reflects the degree of mutual influence of the suitability index between nodes.

[0104] Step S420: Normalize the area size, historical fish distribution density, and ecological function importance of the spatial region unit to generate dimensionless spatial weights; assign and normalize the fish life history stages corresponding to the time nodes to generate dimensionless temporal weights; perform weighted processing on each node of the spatiotemporal graph network, multiply the node's correlation score by the corresponding spatial weight and temporal weight to generate a weighted habitat suitability index.

[0105] The area size of a spatial unit refers to its geographical area; historical fish distribution density refers to the historical fish density of that unit; and ecological function importance refers to the functional importance of that unit within the watershed ecosystem. These three factors jointly influence the habitat suitability weight of the unit. The normalization method can use min-max standardization. The spatial weight is a dimensionless weight obtained by combining the normalized values ​​of these three factors; a larger weight indicates a higher importance of the unit to fish habitat. The fish life history stage corresponding to a time node refers to the stage in the fish's life history at that time node (e.g., spawning period, fattening period, overwintering period, etc.). Different life history stages have different habitat requirements, thus requiring different weights. Assignment and normalization involves assigning values ​​to time nodes based on the importance of life history stages, and then normalizing to obtain dimensionless time weights. A larger weight indicates a more important time node for fish survival. The correlation score of each node in the spatiotemporal graph network is the correlation score between the corresponding environmental node and the demand element in the habitat suitability correlation matrix. The weighted processing involves multiplying the correlation score by the spatial and temporal weights to obtain a weighted habitat suitability index. The higher the index value, the higher the habitat suitability of the spatiotemporal unit.

[0106] Step S430: Spatially expand the weighted habitat suitability index, predict the suitability index of unmonitored areas through the correlation between nodes, and generate a spatially continuous index distribution covering the entire watershed.

[0107] In one implementation, step S430 may specifically include the following steps S431 to S436: Step S431: Set up a regular interpolation grid that covers the entire watershed. The grid cells have a fixed spatial resolution, and the grid range is consistent with the watershed boundary to ensure coverage of all regional cells.

[0108] A regular interpolation grid is a regular grid used for spatial interpolation, and the grid cell is the basic unit that makes up the grid. Fixed spatial resolution means that the size of the grid cell is fixed and must be consistent with or higher than the resolution of the regional cells to ensure interpolation accuracy. Grid extent coincides with the watershed boundary, meaning that the grid boundary completely coincides with the geographical boundary of the watershed, ensuring coverage of all areas within the watershed. Covering all regional cells means that the grid cell can completely contain all regional cells within the watershed, without omission.

[0109] Step S432: Extract the spatial coordinates and weighted suitability index of all nodes in the spatiotemporal graph network, and generate a set of sampling points containing coordinates and index values. The sampling points cover all monitoring area units within the watershed.

[0110] In the spatiotemporal graph network, the spatial coordinates of all nodes refer to the spatial coordinates of the corresponding regional unit, and the weighted suitability index is the habitat suitability index of the node. The sampling point set is a data set containing spatial coordinates and corresponding suitability index values, with each sampling point corresponding to a spatiotemporal node of a monitoring regional unit. Covering all monitoring regional units within the watershed means that the sampling point set includes the suitability indices of all regional units with monitoring data, ensuring the completeness of the basic data for interpolation.

[0111] Step S433: Construct a node association model. Calculate the association weights based on the spatial distance and environmental similarity between sampling points. Sampling points that are close to each other and have high similarity have greater weights. The weight values ​​determine the degree of influence of the sampling points on the interpolation results.

[0112] The node association model is used to calculate the association weights between sampling points. These weights reflect the influence of sampling points on the interpolation results for unmonitored areas. Spatial distance between sampling points refers to the spatial distance between the corresponding regional units of two sampling points, while environmental similarity refers to the similarity of environmental features between the corresponding regional units of two sampling points. The calculation of association weights is based on spatial distance and environmental similarity; sampling points that are closer together and have higher similarity have greater association weights and stronger influence on the interpolation results. The weight value determines the influence of a sampling point on the interpolation results; a sampling point with a larger weight has a greater proportion in the interpolation calculation. The specific method for constructing the node association model is as follows: First, design a function to calculate the association weights, for example, association weight = exp(-γ × spatial distance) × environmental similarity, where γ is the distance decay coefficient; then, for each unmonitored grid unit, calculate its spatial distance and environmental similarity with all sampling points; next, calculate the association weight of each sampling point according to the calculation function; finally, construct the node association model, which can calculate the association weights based on the spatial distance and environmental similarity of sampling points, providing a weight calculation method for spatial interpolation.

[0113] Step S434: For each grid cell in the interpolation grid, calculate the weighted average index of the surrounding sampling points according to the node association model. The weights are distributed according to the distance decay law, and the weighted average covers all valid sampling points.

[0114] Surrounding sampling points refer to sampling points near unmonitored grid cells, typically determined based on a spatial distance threshold. The weighted average index is the result of averaging the suitability indices of surrounding sampling points using the association weights calculated based on the node association model. Weights are distributed according to a distance decay law, meaning the association weight of a sampling point decreases with increasing spatial distance from the grid cell; the closer the distance, the greater the weight. Covering all valid sampling points means the weighted average calculation includes all valid sampling points that meet the criteria, ensuring the accuracy of the interpolation results.

[0115] Step S435: Perform boundary constraint processing on the interpolation results, setting the index values ​​of cells that exceed the watershed boundary in the grid to zero, to ensure that the interpolation results conform to the actual watershed range.

[0116] Boundary constraint processing refers to correcting the index values ​​of grid cells that extend beyond the watershed boundary, ensuring that the interpolation results only include areas within the watershed. Setting the index value of cells extending beyond the watershed boundary to zero means that for grid cells not within the watershed boundary, their suitability index value is set to zero, indicating that the area does not belong to the watershed's habitat range. Ensuring that the interpolation results conform to the actual watershed range means that after boundary constraint processing, the spatial range of the interpolation results is completely consistent with the watershed boundary, with no areas exceeding the range. The specific method for boundary constraint processing of the interpolation results is as follows: First, obtain the geographic coordinate data of the watershed boundary; then, traverse all grid cells and determine whether each cell is within the watershed boundary; next, set the suitability index value of grid cells extending beyond the watershed boundary to zero; finally, obtain the interpolation results that conform to the actual watershed range, ensuring the spatial accuracy of the interpolation results.

[0117] Step S436: Smooth the interpolation results by averaging the index values ​​of adjacent grid cells to eliminate local abnormal fluctuations and generate a spatially continuous and smooth suitability index distribution. Each grid cell value in the suitability index distribution represents the habitat suitability index of the corresponding spatial location.

[0118] The interpolation result is a suitability index for grid cells after boundary constraint processing, and may contain local anomalies (such as outliers caused by noise or interpolation errors). Smoothing involves averaging the index values ​​of adjacent grid cells to eliminate these local anomalies, making the index distribution more continuous and smooth. Adjacent grid cells are those spatially adjacent to the target grid cell (e.g., cells in the four directions: up, down, left, and right). A spatially continuous and smooth suitability index distribution means that after smoothing, the index values ​​change continuously in space without significant jumps or anomalies. Each grid cell value represents the habitat suitability index for a corresponding spatial location; the value of each grid cell in the smoothed index distribution represents the degree of habitat suitability for that spatial location.

[0119] Step S440: Perform time slicing on the spatial continuous exponential distribution, extract the spatial distribution data at the corresponding time nodes, and generate spatial distribution maps of suitability index at different time nodes.

[0120] Time-slicing refers to dividing the spatially continuous exponential distribution according to time nodes and extracting the spatial distribution data corresponding to each time node. The spatial distribution data at a given time node refers to the suitability index values ​​of all grid cells at that time node, reflecting the spatial distribution of watershed habitat suitability at that moment. The spatial distribution map of suitability index at different time nodes is the result of graphically displaying the spatial distribution data of each time node, which can intuitively reflect the spatial changes in watershed habitat suitability at different times.

[0121] Step S450: Combine the spatial distribution maps of the suitability index of all time points in chronological order to generate a habitat suitability index distribution map containing different regional units of the watershed at different time points. The map retains the spatial coordinates and time markers of each regional unit.

[0122] Arranging spatial distribution maps in chronological order refers to arranging these maps according to the order of time nodes to form a time-series map set. The habitat suitability index distribution map is the combined map set, containing suitability index information for different regional units in the watershed at different time nodes. Retaining the spatial coordinates and time markers of each regional unit means that each data point in the map contains corresponding spatial coordinates and time node information, ensuring that the spatiotemporal location of each index value can be accurately traced. The specific method for combining the spatial distribution maps of all time nodes to generate the distribution map is as follows: First, obtain the spatial distribution maps of suitability indices for all time nodes and sort them in chronological order; then, integrate these distribution maps into a comprehensive map, where the index value at each location corresponds to spatial coordinates and time nodes; finally, ensure that the map retains the spatial coordinates and time markers of each regional unit, generating a habitat suitability index distribution map containing different regional units in the watershed at different time nodes. This map can comprehensively reflect the spatiotemporal dynamic changes of habitat suitability in the watershed, providing a visual basis for the planning of fish habitats in the watershed.

[0123] Step S500: Determine the optimal habitat spatial range for the target fish species based on the habitat suitability index distribution map, prioritize the ecological protection of the optimal habitat spatial range, and generate the watershed fish habitat planning results.

[0124] In one implementation, step S500 may specifically include the following steps S510-S560: Step S510: Extract index distribution data for different time periods from the habitat suitability index distribution map. Based on the habitat demand characteristics of fish at different life history stages, set corresponding index screening conditions for each life history stage, identify continuous spatial areas where the index values ​​for each time period meet the corresponding conditions, and generate a multi-time period candidate habitat area set containing spatial coordinate range and time period index value distribution.

[0125] The habitat suitability index distribution map contains index distribution data at different time points, where different time points refer to different time nodes or time intervals within the map. Different life history stages of fish refer to different stages in the fish's life cycle, such as spawning, hatching, juvenile fattening, adult fattening, and overwintering. Each stage has different habitat requirements. Habitat requirement characteristics refer to the environmental requirements of fish at the corresponding life history stage, including the required range of suitability index. Index screening criteria are suitability index thresholds or ranges set based on the requirements of fish life history stages; for example, the suitability index for the spawning period must be higher than 0.8. Continuous spatial regions refer to areas where the index values ​​satisfy the screening criteria are spatially continuous; continuous regions are more conducive to fish survival and activity. The multi-time period candidate habitat region set is a collection of continuous spatial regions that meet the screening criteria at different time periods. Each region includes a spatial coordinate range and the corresponding time period's index value distribution.

[0126] Step S520: Perform a spatiotemporal overlay operation on the multi-time period candidate habitat area set, count the proportion of each area's occurrence in all time periods to the total number of time periods, calculate the fluctuation range of the index value in the area, and mark the areas whose occurrence ratio reaches the preset ratio condition and whose fluctuation range meets the preset fluctuation condition as core candidate areas.

[0127] Spatiotemporal overlay refers to spatially overlaying candidate habitat areas from different time periods and counting the number of times each spatial location is marked as a candidate area across all time periods. The proportion of occurrences to the total number of time periods is the area's occurrences across all time periods divided by the total number of time periods; a higher proportion indicates that the area is suitable for fish survival in multiple time periods. The fluctuation range of the index value refers to the range of variation in the suitability index value across different time periods within the area. It can be measured by calculating the difference between the maximum and minimum values ​​or the standard deviation of the index value; a smaller fluctuation range indicates more stable suitability. The preset proportion condition refers to the threshold that the occurrence proportion must reach; for example, ≥0.7 indicates that the area is suitable for fish survival in more than 70% of the time periods. The preset fluctuation condition refers to the range that the index value fluctuation range must meet; for example, a standard deviation ≤0.1 indicates that the area's suitability index fluctuates less. Core candidate areas are areas whose occurrence proportion and fluctuation range both meet the preset conditions, representing long-term suitable habitats for fish. The specific method for spatiotemporally overlaying candidate habitat regions across multiple time periods and marking core candidate regions is as follows: First, spatially overlay candidate habitat regions across all time periods and count the occurrence frequency of each spatial location; then calculate the occurrence ratio of each region (occurrence frequency / total number of time periods); next, calculate the fluctuation range of the index value within each region; then mark regions with an occurrence ratio ≥ a preset ratio condition and a fluctuation range ≤ a preset fluctuation condition as core candidate regions; finally, generate a core candidate region set, which contains long-term stable and suitable habitats for fish.

[0128] Step S530: Adjust the spatial morphology of the core candidate region. By calculating the minimum bounding polygon of the region boundary, merge the boundaries of adjacent core candidate regions whose spatial distance meets the preset merging conditions, delete independent regions whose internal area meets the preset hole conditions, and generate an integrated habitat region with regular morphology.

[0129] Spatial morphology adjustment refers to optimizing the spatial shape of core candidate areas to better meet the actual functional needs of habitats. The minimum bounding polygon is the smallest convex polygon or smallest enclosing rectangle containing all boundary points of the area; calculating the minimum bounding polygon simplifies the boundary shape of the area. Preset merging conditions refer to the spatial distance threshold for merging adjacent core candidate areas; for example, merging occurs when the boundary distance between two areas is ≤500 meters, resulting in a larger, continuous habitat. Preset perforation conditions refer to the area threshold for independent areas (perforations) within the area; for example, perforations with an area ≤1 square kilometer are deleted, resulting in a more complete area morphology. The integrated habitat area, generated after morphological adjustment, has a regular shape and clear boundaries, making it more conducive to fish activity and conservation management.

[0130] Step S540: Calculate the effective habitat area and spatial connectivity index for each integrated habitat region. The effective habitat area is the total area of ​​all grid cells in the region whose index values ​​meet the screening conditions for the corresponding time period. The spatial connectivity index is the ratio of the average shortest path length between any two points in the region to the area of ​​the region.

[0131] Effective habitat area refers to the total area of ​​all grid cells within a region whose index values ​​meet the selection criteria for the corresponding time period. These cells are truly suitable areas for fish survival. Spatial connectivity index measures the degree of connectivity between different locations within a region; higher connectivity indicates easier movement for fish within the region. The average shortest path length between any two points is calculated by randomly selecting multiple pairs of points within the region, calculating the shortest path length between each pair (considering obstacles such as terrain and waterways), and then averaging the results. Region area refers to the total area of ​​the integrated habitat region. The spatial connectivity index is the ratio of the average shortest path length to the region area; a smaller ratio indicates higher connectivity (shorter paths and larger area).

[0132] Step S550: Normalize the effective habitat area and spatial connectivity index to generate dimensionless area score and connectivity score; construct habitat quality assessment rules, assign preset weights to the area score and connectivity score respectively, calculate the weighted sum of the two as the regional quality score, sort the integrated habitat areas from high to low according to the score, and select the area with the highest score as the optimal habitat spatial range for the target fish.

[0133] Effective habitat area and spatial connectivity index are quantitative indicators calculated in step S540. Since their dimensions and numerical ranges differ, they require normalization. The min-max normalization method can be used. The area score is the normalized result of the effective habitat area; a higher score indicates a larger effective habitat area. The connectivity score is the normalized result of the spatial connectivity index. Because the connectivity index is a ratio, the normalization direction needs attention (e.g., a smaller index indicates higher connectivity; therefore, 1 - index / maximum index can be used for normalization). The habitat quality assessment rules are rules for evaluating regional quality by combining area score and connectivity score, including weight allocation and calculation methods. The preset weights are weights assigned based on the importance of area and connectivity to fish habitat quality; for example, the area weight is 0.6, and the connectivity weight is 0.4. The regional quality score is the weighted sum of the area score and connectivity score; a higher score indicates higher habitat quality. Sorting by score from highest to lowest means arranging the quality scores of all integrated habitat regions in descending order and selecting the region with the highest score as the optimal habitat spatial range.

[0134] Step S560: Divide the optimal habitat spatial range into multiple sub-regional units of equal area according to a fixed area standard. Each sub-regional unit has a unique spatial index and boundary coordinates. Construct an assessment dimension that includes habitat area, index mean, disturbance distance, and ecosystem integrity. Perform multi-dimensional quantitative scoring on each sub-regional unit. Generate an ecological protection priority list by sorting the scores from high to low. Integrate the coordinate information of the optimal habitat spatial range with the priority list to generate the watershed fish habitat planning results.

[0135] In one implementation, step S560 may specifically include the following steps S561 to S566: Step S561: Calculate the total area enclosed by the boundary coordinates of the optimal habitat spatial range, determine the number of divisions according to the preset sub-region area standard, and distribute the total area equally into multiple square sub-region units with the same area. Each sub-region unit generates a unique spatial index identifier and a corresponding set of boundary coordinates.

[0136] The boundary coordinates of the optimal habitat spatial extent refer to the geographic boundary coordinates of the region. The total enclosed area refers to the area enclosed by the boundary coordinates, which can be calculated using the polygon area calculation formula. The preset sub-region area standard refers to the fixed area size of each sub-region unit, such as 1 square kilometer. The number of divisions is the integer obtained by dividing the total area by the sub-region area standard; for example, if the total area is 10 square kilometers and the sub-region area standard is 1 square kilometer, then the number of divisions is 10. A square sub-region unit with equal area means that each sub-region unit is square in shape, and its area is equal to the sub-region area standard. The square shape facilitates calculation and management. A unique spatial index identifier means that each sub-region unit has a unique number or identifier, such as "SubRegion_001" or "SubRegion_002". The corresponding boundary coordinate set refers to the coordinates of the four vertices or boundary line of each sub-region unit, used to determine the spatial extent of the sub-region unit.

[0137] Step S562: Extract the index values ​​of all grid cells in the habitat suitability index distribution map for each sub-regional unit, and calculate the arithmetic mean of these index values. This mean reflects the overall environmental suitability level of the sub-regional unit and serves as the assessment dimension for the mean of the suitability index.

[0138] The habitat suitability index distribution map includes the suitability index value for each grid cell, and each sub-regional unit consists of multiple grid cells. Extracting the index values ​​of all grid cells within each sub-regional unit involves selecting all grid cells belonging to that sub-regional unit from the map and extracting their index values. The arithmetic mean is the sum of these index values ​​divided by the number of grid cells. The overall environmental suitability level refers to the comprehensive suitability of environmental factors within the sub-regional unit for fish survival; a higher arithmetic mean indicates a higher overall suitability level. The suitability index mean assessment dimension uses the arithmetic mean as a dimension to assess the ecological protection priority of the sub-regional unit; the dimension value is the arithmetic mean.

[0139] Step S563: Obtain spatial distribution data of human activity disturbance sources in the watershed, including the coordinates of industrial facilities, waterways, and sewage outlets. Calculate the straight-line distance between the centroid coordinates of each sub-region unit and the coordinates of the nearest disturbance source. This distance reflects the degree of human activity impact on the sub-region unit and serves as a dimension for disturbance distance assessment.

[0140] Human disturbance sources in a watershed refer to human facilities or areas that negatively impact fish habitats, including industrial facilities, waterways, and sewage outlets. The spatial distribution data of these sources includes their geographic coordinates. Centroid coordinates are the geometric center coordinates of a sub-regional unit, calculated from its boundary coordinates. For example, for a square sub-regional unit, the centroid coordinates are ((x1+x2) / 2, (y1+y2) / 2), where (x1,y1) and (x2,y2) are the coordinates of the diagonal vertices. The nearest disturbance source coordinates are the coordinates of the disturbance source closest to the centroid of the sub-regional unit among all disturbance source coordinates. Straight-line distance is the Euclidean distance between the centroid coordinates and the nearest disturbance source coordinates. The degree of human impact indicates that the closer the source, the greater its impact on the sub-regional unit; the farther away, the smaller the impact. The disturbance distance assessment dimension uses straight-line distance as a dimension to assess the ecological protection priority of a sub-regional unit. The dimension value is the straight-line distance; a larger distance indicates a smaller degree of disturbance and a higher protection priority.

[0141] Step S564: Collect biological community structure data, food chain integrity data, and native vegetation cover data for each sub-regional unit. Integrate and correlate these data, and generate a quantitative value that comprehensively reflects the regional ecological status through the analysis of the synergistic relationship between the data, which serves as a dimension for assessing ecosystem integrity.

[0142] Biological community structure data refers to data such as species composition, species richness, and community diversity within a sub-regional unit, reflecting the structural characteristics of the biological community. Food chain integrity data refers to the completeness of the food chain within a sub-regional unit, including the presence and interrelationships of producers, consumers, and decomposers, reflecting the energy flow and material cycling status of the ecosystem. Native vegetation cover data refers to the proportion of native vegetation (natural vegetation undisturbed by humans) within a sub-regional unit, reflecting the ecosystem's natural recovery capacity and stability. Correlation integration refers to the comprehensive analysis of these three types of data, considering their interrelationships. For example, the more complex the biological community structure, the more complete the food chain, and the higher the native vegetation cover, the higher the ecosystem integrity. Synergistic relationship analysis refers to analyzing positive or negative correlations between data; for example, high native vegetation cover usually promotes a complex biological community structure and a complete food chain. The quantitative value comprehensively reflecting the regional ecological status is a comprehensive indicator obtained through weighted summation or model calculation; a higher value indicates higher ecosystem integrity. The ecosystem integrity assessment dimension uses this quantitative value as a dimension to assess the ecological protection priority of the sub-regional unit; the dimension value is the comprehensive quantitative value.

[0143] Step S565: For the four assessment dimensions of habitat area, mean suitability index, disturbance distance and ecosystem integrity, determine the corresponding weight values ​​by comparing the degree of impact of each dimension on fish survival. The sum of the weight values ​​is a set proportion, among which habitat area and ecosystem integrity have higher weights than the other two dimensions.

[0144] The four assessment dimensions are habitat area, mean suitability index, disturbance distance, and ecosystem integrity. Each dimension has a different impact on fish survival, therefore requiring different weight values. The weight value is an indicator of the importance of each dimension; a larger weight value indicates a greater impact of that dimension on ecological protection priorities. Setting a percentage means the total weight values ​​are 1 (or 100%), ensuring the rationality of the weight allocation. The higher weights for habitat area and ecosystem integrity compared to the other two dimensions mean that their weight values ​​are greater than those for the mean suitability index and disturbance distance. This is because habitat area is the basic space for fish survival, ecosystem integrity is crucial for maintaining habitat function, while the mean suitability index and disturbance distance are secondary influencing factors.

[0145] Step S566: Convert the raw data of the four evaluation dimensions of each sub-region unit into standard scores with uniform dimensions. During the conversion process, the value range is determined by the maximum and minimum values ​​of each dimension. The raw data is mapped to the corresponding range proportionally. When mapping the interference distance dimension, the greater the distance, the higher the score.

[0146] The raw data for the four assessment dimensions refer to the original quantitative values ​​of habitat area, mean suitability index, disturbance distance, and ecosystem integrity. Since their dimensions and numerical ranges differ, they need to be converted to standardized scores with unified dimensions (e.g., a range of [0,1]). Standardized scores with unified dimensions mean that the scores for all dimensions fall within the same numerical range, facilitating weighted summation. Determining the value range using the maximum and minimum values ​​of each dimension involves collecting the raw data from all sub-regional units for each dimension, determining the maximum (max) and minimum (min) values, and setting the value range to [min, max]. Proportional mapping to the corresponding range involves converting the raw data x into a standardized score s using the formula: s = (x - min) / (max - min). For the disturbance distance dimension, since a larger distance results in a higher score, the conversion formula remains consistent (because min is the minimum distance and max is the maximum distance; a larger distance results in a larger x and a larger s). The higher the score for a larger distance when mapping the disturbance distance dimension means that the standardized score for that dimension is positively correlated with the distance, meeting the requirements of protection priority (the greater the distance, the less disturbance, and the higher the priority).

[0147] In one implementation, step S566 may specifically include the following steps S5661 to S5665: Step S5661: Collect the raw habitat area data of all sub-region units, determine the maximum and minimum values, divide the difference between the raw area value and the minimum value by the difference between the maximum and minimum values, and then perform proportional mapping to obtain the standard score of habitat area for each sub-region unit, ensuring that the score is within the corresponding interval.

[0148] The raw habitat area data consists of the effective habitat area value for each sub-region unit. After collecting this data from all sub-region units, the maximum value (A) is determined. max ) and minimum value (A) min The difference between the original area value and the minimum area value is (A). i -A min The difference between the maximum and minimum values ​​is (A) max -A min ). Scale mapping is to map (A) i -A min ) divided by (A max -A min ), to obtain the standard score S of habitat area A =(A i -A min ) / (A max -A min Ensuring the score is within the corresponding range means S. A The value range of A is [0,1]. i =A min At that time, S A =0; when A i =A max At that time, S A =1, other values ​​are mapped proportionally.

[0149] Step S5662: Process the suitability index mean assessment dimension data based on the same maximum and minimum value comparison method. Divide the difference between the original mean and minimum value by the difference between the maximum and minimum value, and then perform proportional mapping to obtain the standard score of the suitability index mean. The score interval is consistent with the habitat area score. When processing the disturbance distance assessment dimension data, first determine the boundary of the influence range of the disturbance source within the watershed. When the distance from the sub-region unit to the disturbance source exceeds the boundary of the influence range, the score is the upper limit of the interval; when it does not exceed the boundary, the score is obtained by interval mapping through the ratio of distance to the boundary of the influence range.

[0150] The data processing method for the suitability index mean assessment dimension is the same as that for the habitat area dimension, that is, determining the maximum value of this dimension (M). max ) and minimum value (M) min ), calculate the standard score S M =(M i -M min ) / (Mmax -M min The scoring interval is also [0,1], consistent with the habitat area score. The processing of disturbance distance assessment data needs to consider the boundary of the disturbance source's influence range. The influence range boundary refers to the maximum distance at which the disturbance source has a significant impact on fish habitat, which can be determined based on ecological research or expert experience (e.g., 5 kilometers). When the distance D from the sub-region unit to the nearest disturbance source... i Exceeding the boundary of the influence area D bound When D indicates that the sub-region unit is minimally affected by the interference source, the standard score is set to the upper limit of the interval, which is 1; when D i Not exceeding D bound At that time, interval mapping is performed using the ratio of distance to the boundary of the influence range, i.e., S D =D i / D bound The score range is [0,1], and the greater the distance, the higher the score.

[0151] Step S5663: Perform outlier checks on the standard scores of all evaluation dimensions. When the score of a certain dimension deviates from the average score of that dimension by a set deviation level, adjust it to the boundary value of the score range of that dimension to ensure that the score distribution conforms to the overall data characteristics.

[0152] The standard scores for all assessment dimensions refer to the standard scores for the four dimensions: habitat area, mean suitability index, disturbance distance, and ecosystem integrity. Outliers are extreme scores that deviate from the overall data distribution, possibly due to data errors or special circumstances. Setting the deviation level refers to the threshold for deviation from the average score, such as ±2 standard deviations of the average score. Scores exceeding this range are considered outliers. Adjusting to the boundary values ​​of the score interval means adjusting outliers to the minimum (0) or maximum (1) value of the score interval for that dimension. For example, if the average score for a dimension is 0.5 and the standard deviation is 0.1, setting the deviation level to ±2 standard deviations means that scores <0.3 or >0.7 are considered outliers and adjusted to 0 or 1. Ensuring that the score distribution conforms to the overall data characteristics means that after outlier adjustment, the score distribution is more reasonable and can reflect the actual situation of the sub-regional units.

[0153] Step S5664: Multiply the habitat area standard score, the average suitability index standard score, the disturbance distance standard score, and the ecosystem integrity standard score of each sub-region unit by their respective weight values, and then sum the products to obtain the comprehensive evaluation score of the sub-region unit.

[0154] Habitat area standard score (S A ), the mean standard score of the suitability index (S) M Interference distance standard score (S) D ) and ecosystem integrity standard score (SE The standardized score obtained in steps S5661~S5663 corresponds to the weight value (W). A W M W D W E The weight values ​​are those determined in step S565, and the sum of the weight values ​​is 1. The comprehensive evaluation score is the result of multiplying the standard score of each dimension by its corresponding weight value and then summing the results. The calculation formula is: Comprehensive Score = S A ×W A +S M ×W M +S D ×W D +S E ×W E This score comprehensively reflects the ecological protection priority of the sub-regional unit; a higher score indicates a higher priority.

[0155] Step S5665: Arrange the comprehensive assessment scores of all sub-regional units in descending order. When the comprehensive assessment scores are the same, sort them again in descending order according to the ecosystem integrity standard score to generate the final ecological protection priority ranking result.

[0156] The comprehensive assessment score of all sub-regional units is the quantitative score calculated in step S5664. Arranging them in descending order means sorting the sub-regional units in descending order of their comprehensive scores, with higher scores indicating higher priority. The secondary sorting for units with the same comprehensive assessment score involves comparing their ecosystem integrity standard scores when two or more sub-regional units have the same score, sorting them in descending order of ecosystem integrity standard scores, as ecosystem integrity is more important for the long-term stability of fish habitats. The final ecological protection priority ranking result is a list of sub-regional units obtained after these two sorting processes. The order of the list represents the priority of ecological protection, with the sub-regional units listed earlier receiving priority protection.

[0157] Please see Figure 2This is a schematic diagram of a computer system provided in an embodiment of the present invention. The computer system includes at least a processor 101, a communication interface 102, and a memory 103. The processor 101, communication interface 102, and memory 103 can be connected via a bus or other means. The processor 101 (or Central Processing Unit, CPU) is the computing and control core of the computer system, capable of parsing various instructions and processing various data within the computer system. The communication interface 102 may optionally include standard wired interfaces or wireless interfaces (such as Wi-Fi, mobile communication interfaces, etc.), and can be used for sending and receiving data under the control of the processor 101; the communication interface 102 can also be used for data transmission and interaction within the computer system. The memory 103 is a memory device in the computer system used to store programs and data. It is understood that the memory 103 here can include the computer system's built-in memory, or it can include extended memory supported by the computer system. The memory 103 provides storage space, which stores the computer system's operating system; this invention does not limit this. In one embodiment, the processor 101 executes the watershed fish habitat matching method based on model simulation provided above in the embodiments of the present invention by running a computer program in the memory 103.

Claims

1. A method for matching fish habitats in a watershed based on model simulation, characterized in that, The method includes: Environmental monitoring data from different monitoring points within the watershed are acquired within a preset time period. Combined with basic topographic and geomorphological data of the watershed, a watershed environmental scene dataset containing spatial coordinate information and time series labels is constructed. A watershed environment map network is constructed, with monitoring points in the watershed environment scene dataset as network nodes. Directed edges between nodes are determined based on river connectivity and terrain elevation differences in the basic topographic data. Edge attributes include river distance and water flow direction between nodes. Node features are extracted from the environmental monitoring data corresponding to each node in the watershed environmental map network. The node environmental monitoring data are arranged in ascending order according to the time series label to obtain a continuous time axis data sequence. The segment length is determined according to the data collection interval to generate adjacent non-overlapping segmented data blocks. The mean level, fluctuation range and trend direction of each data block are calculated and combined into a node environmental feature vector. The environmental element interaction weights of directed edges are calculated based on the river distance and water flow direction between nodes. The weight values ​​are determined by the ratio of the difference in node environmental monitoring data to the river distance, and a weighted edge attribute matrix is ​​generated. Traverse each node in the watershed environment graph network, and calculate the node environment feature vector of each neighboring node by weighting it with the corresponding connection weight according to the connection weight of the node and its neighboring nodes in the edge attribute matrix. Aggregate all the weighted neighboring node environment feature vectors, and fuse the aggregation result with the node's own node environment feature vector to generate enhanced node features with fused spatial association. The time reference of all enhanced node features is unified. By comparing the overlapping intervals of the time series markers of each node, the time dimension of different node features is adjusted to generate an environmental feature evolution sequence with a unified time axis. The environmental feature evolution sequence is input into the fish demand association module. Through multi-layer nonlinear transformation, the synergistic relationship between environmental elements is captured, the key element combination related to fish survival activities is extracted, and a fish habitat demand feature set representing the fish demand pattern of environmental elements is generated. Based on the environmental feature evolution sequence and the fish habitat demand feature set, cross-scale feature matching is performed to construct a habitat suitability correlation matrix that characterizes the relationship between environmental elements and habitat demand; The habitat suitability correlation matrix is ​​subjected to spatiotemporal planning processing to generate a habitat suitability index distribution map containing different regional units of the watershed at different time nodes; The optimal habitat spatial range for the target fish species is determined based on the habitat suitability index distribution map, and the ecological protection priority of the optimal habitat spatial range is ranked to generate the watershed fish habitat planning results.

2. The method according to claim 1, characterized in that, The process involves extracting node features from the environmental monitoring data corresponding to each node in the watershed environmental map network, arranging the node environmental monitoring data in ascending order according to time series labels to obtain a continuous time axis data sequence, determining the segment length according to the data collection interval, generating adjacent non-overlapping segmented data blocks, calculating the mean level, fluctuation range, and trend direction of each data block, and combining them into a node environmental feature vector, including: The environmental monitoring data of a single node are arranged in ascending order according to the time series label to obtain an environmental data sequence on a continuous time axis, so that the data points are distributed without interruption according to the order of collection. The segment length is determined based on the data acquisition time interval. Starting from the beginning of the data sequence, data segments are continuously extracted according to the set step size to generate multiple adjacent and non-overlapping segmented data blocks. Each data block contains a preset number of environmental monitoring data points. For each segmented data block, calculate the arithmetic mean of all data points, which serves as the mean level characteristic of the environmental elements for the corresponding time period of that segmented data block; The difference between the maximum and minimum values ​​in the segmented data blocks is calculated as a characteristic of the fluctuation range of environmental elements. The difference calculation is based on the extreme points of the actual monitoring data within the data block. Linear trend analysis is performed on segmented data blocks. A straight line is fitted by the coordinates of the data points, and the trend direction is determined by the positive or negative slope of the line. A positive slope indicates an increasing trend of environmental factors, while a negative slope indicates a decreasing trend. The absolute value of the slope reflects the rate of trend change. The mean level, fluctuation range, and trend direction features of all segmented data blocks of the same node are combined sequentially in chronological order to generate a node environment feature vector containing time dynamics. The vector dimension is consistent with the number of segmented data blocks.

3. The method according to claim 2, characterized in that, The environmental feature evolution sequence is input into the fish demand association module, which captures the synergistic relationship between environmental elements through multi-layer nonlinear transformation, extracts key element combinations related to fish survival activities, and generates a fish habitat demand feature set characterizing the fish's demand pattern for environmental elements, including: The environmental feature evolution sequence is standardized to generate a standardized environmental feature sequence. The standardized environmental feature sequence is input into the feature interaction layer of the fish demand association module. Through multi-layer nonlinear transformation, the synergistic relationship between different environmental elements is captured, and a feature interaction graph containing element interaction information is generated. The feature interaction graph is filtered for key elements to generate a subset of key element interactions; Fish demand patterns are extracted based on key element interaction subsets. By analyzing the value range and variation patterns of element combinations, the suitable element combination range for fish survival is determined, and a demand pattern representation is generated. The demand pattern representation is quantified, and the appropriate combination range of elements is converted into a numerical demand threshold range, including the lower limit, upper limit and optimal value center of the elements, to generate quantitative demand characteristics. All quantitative demand features are categorized and organized according to environmental element types to generate a fish habitat demand feature set that includes element type identifiers, demand threshold ranges, and descriptions of collaborative relationships.

4. The method according to claim 1, characterized in that, The method of constructing a habitat suitability correlation matrix, which characterizes the relationship between environmental elements and habitat demand, by performing cross-scale feature matching based on the environmental feature evolution sequence and the fish habitat demand feature set, includes: The environmental feature evolution sequence and the fish habitat demand feature set are aligned with a graph structure. The graph nodes of the environmental feature evolution sequence are matched with the demand element nodes of the fish habitat demand feature set according to type, generating cross-scale graph node pairs so that the node types and quantities correspond one-to-one. The feature scale transformation is performed on the cross-scale graph node pairs. By adjusting the node feature dimension, the environmental features and the demand features have the same feature space dimension, generating cross-scale feature pairs with unified scale. Construct a cross-scale feature matching network, which includes an input layer, a feature fusion layer, a correlation calculation layer, and an output layer; The cross-scale feature pairs are input into the input layer of the cross-scale feature matching network, and the feature values ​​are mapped to the same numerical range through linear transformation of the feature values ​​to generate standardized cross-scale feature pairs. Based on the feature fusion layer, standardized cross-scale feature pairs are processed, and a correlation model between environmental elements and demand elements is constructed through node feature splicing and nonlinear transformation to generate a fused feature vector. Based on the aforementioned correlation calculation layer, the similarity of the fused feature vectors is evaluated, and the degree of matching between environmental elements and demand elements is determined through vector inner product operation to generate a correlation score. Arrange the correlation scores of all cross-scale map node pairs according to the order of environmental map network nodes and the order of demand feature set elements to generate a two-dimensional matrix form of habitat suitability correlation matrix. The matrix rows represent environmental nodes, the columns represent demand elements, and the elements represent the correlation scores between the corresponding nodes and elements.

5. The method according to claim 4, characterized in that, The step of aligning the environmental feature evolution sequence and the fish habitat demand feature set with graph structures, matching the graph nodes of the environmental feature evolution sequence with the demand element nodes of the fish habitat demand feature set by type, and generating cross-scale graph node pairs to ensure a one-to-one correspondence between node types and quantities, includes: The graph node type labels of the environmental feature evolution sequence are analyzed to identify the types of environmental element nodes included. Each node type corresponds to an independent environmental element, and the node type and quantity information are recorded. Analyze the node labels of the demand elements in the fish habitat demand feature set to identify the types of demand elements included; Type matching is performed on environmental element nodes and demand element nodes, and environmental nodes of the same element type are bound to demand nodes one-to-one to obtain element node pairs. For each element node pair, the feature vector of the environmental node in the environmental feature evolution sequence is extracted to obtain the environmental node feature vector, which contains the temporal dynamic features and spatial correlation features of the node. Extract the demand feature vector of the demand node from the fish habitat demand feature set to obtain the demand node feature vector. The demand feature vector contains the demand threshold range and collaborative relationship features of the element. The feature vectors of environmental nodes and feature vectors of demand nodes are combined into cross-scale graph node pairs. Each node pair contains feature type identifier, environmental feature data and demand feature data. The number of node pairs is consistent with the number of feature types.

6. The method according to claim 1, characterized in that, The process of performing spatiotemporal planning on the habitat suitability correlation matrix to generate a habitat suitability index distribution map containing different regional units of the watershed at different time points includes: The habitat suitability correlation matrix is ​​converted into a spatiotemporal graph network. Nodes represent the combination of spatial regional units and time nodes within the watershed, and edges represent the correlation between adjacent spatiotemporal units. Edge attributes include the comprehensive spatiotemporal distance index and the index transmission coefficient. The area size, historical distribution density of fish, and ecological function importance of spatial units are normalized to generate dimensionless spatial weights; the fish life history stages corresponding to time nodes are assigned values ​​and normalized to generate dimensionless temporal weights; each node of the spatiotemporal graph network is weighted, and the correlation score of the node is multiplied by the corresponding spatial weight and temporal weight to generate a weighted habitat suitability index. Spatially expand the weighted habitat suitability index, predict the suitability index of unmonitored areas through the correlation between nodes, and generate a spatially continuous index distribution covering the entire watershed; The spatial continuous exponential distribution is processed by time slicing, and the spatial distribution data at the corresponding time point is extracted to generate a spatial distribution map of suitability index at different time points. By combining the spatial distribution maps of the suitability index at all time points in chronological order, a habitat suitability index distribution map containing different regional units of the watershed at different time points is generated.

7. The method according to claim 6, characterized in that, The habitat suitability correlation matrix is ​​converted into a spatiotemporal graph network, where nodes represent combinations of spatial regional units and time nodes within the watershed, edges represent the correlation between adjacent spatiotemporal units, and edge attributes include a comprehensive spatiotemporal distance index and an index transmission coefficient, including: The watershed's topographic and geomorphological data are analyzed, and the watershed is divided into regular regional units with fixed spatial resolution. Each regional unit has unique spatial coordinates and boundary range. The entire research period is divided into equally spaced time nodes according to the preset time period. Each time node corresponds to a fixed time scale, and the time nodes cover the entire preset time period. By combining regional units with time nodes, we obtain spatiotemporal units. Each spatiotemporal unit serves as a node in the spatiotemporal graph network, and the node identifier contains the coordinates of the regional unit and the information of the time node. Determine the relationship between adjacent spatiotemporal units. Spatially adjacent regional units or spatiotemporally consecutive time nodes are considered as adjacent nodes, and directed edges are generated between nodes. Calculate the spatial and temporal distances between adjacent nodes, and normalize the spatial and temporal distances respectively to obtain normalized spatial and temporal distance values; calculate the comprehensive spatiotemporal distance index based on the normalized spatial and temporal distance values; The exponential transfer coefficient of the edge is determined based on the comprehensive spatiotemporal distance index and the environmental similarity of the regional unit. The smaller the distance and the higher the similarity of the node pair, the larger the transfer coefficient. The value of the exponential transfer coefficient reflects the degree of mutual influence between the suitability indices of the nodes.

8. The method according to claim 1, characterized in that, The process of determining the optimal habitat spatial range for the target fish species based on the habitat suitability index distribution map, prioritizing the ecological protection of the optimal habitat spatial range, and generating watershed fish habitat planning results includes: Extract index distribution data for different time periods from the habitat suitability index distribution map, set corresponding index screening conditions for each life stage based on the habitat demand characteristics of fish at different life stages, identify continuous spatial areas where the index values ​​of each time period meet the corresponding conditions, and generate a multi-time period candidate habitat area set containing spatial coordinate range and time period index value distribution. A spatiotemporal overlay operation is performed on the set of candidate habitat areas for multiple time periods. The proportion of each area's occurrence in all time periods is counted to the total number of time periods. The fluctuation range of the index value in the area is calculated. Areas whose occurrence proportion reaches a preset proportion condition and whose fluctuation range meets a preset fluctuation condition are marked as core candidate areas. The spatial morphology of the core candidate region is adjusted by calculating the minimum bounding polygon of the region boundary, merging adjacent core candidate regions whose spatial distance meets the preset merging conditions, deleting independent regions whose internal area meets the preset hole conditions, and generating an integrated habitat region with regular morphology. Calculate the effective habitat area and spatial connectivity index for each integrated habitat region. The effective habitat area is the total area of ​​all grid cells in the region whose index values ​​meet the screening criteria for the corresponding time period. The spatial connectivity index is the ratio of the average shortest path length between any two points in the region to the area of ​​the region. The effective habitat area and spatial connectivity index are normalized to generate dimensionless area scores and connectivity scores. Habitat quality assessment rules are constructed, and the area scores and connectivity scores are assigned preset weights. The weighted sum of the two is calculated as the regional quality score. The integrated habitat areas are sorted from high to low scores, and the area with the highest score is selected as the optimal habitat spatial range for the target fish. The optimal habitat spatial range is divided into multiple sub-regional units of equal area according to a fixed area standard. Each sub-regional unit has a unique spatial index and boundary coordinates. An evaluation dimension including habitat area, index mean, disturbance distance and ecosystem integrity is constructed. Each sub-regional unit is quantitatively scored in multiple dimensions. An ecological protection priority list is generated by sorting the scores from high to low. The coordinate information of the optimal habitat spatial range and the priority list are integrated to generate the watershed fish habitat planning results.

9. A computer system, characterized in that, include: A memory, wherein a computer program is stored; A processor for loading the computer program to implement the model simulation-based watershed fish habitat matching method as described in any one of claims 1-8.

Images