A method and system for dynamic assessment of the impact of engineering construction on biodiversity.

Abstract

This invention relates to the field of biodiversity assessment technology, and discloses a method and system for dynamically assessing the impact of engineering construction on biodiversity. The method includes acquiring a basic ecological dataset; constructing an ecological patch map structure and performing hydrological connectivity edge construction based on the dataset to obtain a composite ecological map model; performing patch functional heterogeneity aggregation and disturbance link simulation to obtain ecological propagation paths and sensitivity profiles of each node; performing spectral domain decomposition and key propagation node identification to obtain ecological intermediary nodes and propagation weights; performing reversibility modeling and path entropy backtracking analysis to obtain a community reconstruction trend index and functional stability decay curve; and performing ecological response persistence analysis to obtain a time-series biodiversity impact index. This invention overcomes the problems of static, fragmented, and poorly coupled nature of existing technologies, significantly improving the accuracy, timeliness, and predictive ability of dynamic response assessment to ecological disturbances.

Description

Technical Field

[0001] This invention relates to the field of biodiversity assessment technology, and more specifically, to a method and system for dynamic assessment of the impact of engineering construction on biodiversity. Background Technology

[0002] In existing technologies, biodiversity impact assessments largely rely on static monitoring data, macro-level indicator systems, or species count methods, supplemented by expert experience for disturbance intensity grading and regional sensitivity classification. These methods struggle to reveal the dynamic transmission mechanisms of disturbances within ecological networks, particularly failing to characterize the spatiotemporal fragmentation process of wetland patch community structure caused by changes in water connectivity. Furthermore, traditional methods fail to fully integrate multi-source heterogeneous information such as high-resolution topography, hydrological flow, community function, and species interactions, resulting in assessments that are often one-sided and outdated. This makes them unsuitable for targeted ecological restoration designs and unable to predict potential ecological cascading effects before geological disturbances occur.

[0003] Therefore, there is an urgent need for a dynamic assessment method and system for the impact of engineering construction on biodiversity to address the above issues. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for dynamically assessing the impact of engineering construction on biodiversity, thereby addressing the aforementioned problems. To achieve this objective, the technical solution adopted by this invention is as follows:

[0005] Firstly, this application provides a dynamic assessment method for the impact of engineering construction on biodiversity, including:

[0006] Acquire a basic ecological dataset, which includes a water system distribution map, a forest wetland patch structure change map, and multi-time period quadrat community data;

[0007] Based on the aforementioned basic ecological dataset, ecological patch map structure construction and hydrological connectivity edge construction are performed to obtain a composite ecological map model that reflects spatial structural variations.

[0008] Based on the composite ecological graph model, patch functional heterogeneity aggregation and disturbance link simulation are performed to obtain the ecological propagation path and the sensitivity spectrum of each node.

[0009] Based on the ecological propagation path and the sensitivity spectrum of each node, spectral domain decomposition and key propagation node identification are performed to obtain the ecological intermediary nodes and propagation weights in the disturbance response network.

[0010] Based on the aforementioned ecological intermediary nodes and propagation weights, reversibility modeling and path entropy backtracking analysis are performed to obtain the community reconstruction trend index and functional stability decay curve.

[0011] Based on the community reconstruction trend index and functional stability decay curve, ecological response persistence analysis was performed to obtain the time series biodiversity impact index.

[0012] Secondly, this application also provides a dynamic assessment system for the impact of engineering construction on biodiversity, including:

[0013] The acquisition unit is used to acquire basic ecological datasets, which include water system distribution maps, forest wetland patch structure change maps, and multi-time period quadrat community data.

[0014] The processing unit is used to construct the ecological patch map structure and perform hydrological connectivity edge construction based on the basic ecological dataset to obtain a composite ecological map model that reflects spatial structure variation.

[0015] The simulation unit is used to perform patch functional heterogeneity aggregation and disturbance link simulation processing based on the composite ecological graph model to obtain the ecological propagation path and the sensitivity spectrum of each node.

[0016] The identification unit is used to perform spectral domain decomposition and key propagation node identification processing based on the ecological propagation path and the sensitivity spectrum of each node, so as to obtain the ecological intermediary nodes and propagation weights in the disturbance response network.

[0017] The backtracking unit is used to perform reversibility modeling and path entropy backtracking analysis based on the ecological intermediary nodes and propagation weights to obtain the community reconstruction trend index and functional stability decay curve.

[0018] The analysis unit is used to perform ecological response sustainability analysis based on the community reconstruction trend index and functional stability decay curve to obtain the time series biodiversity impact index.

[0019] The beneficial effects of this invention are as follows:

[0020] This invention takes the travertine-forest-wetland complex ecosystem as a representative scenario and proposes a comprehensive algorithm scheme covering the entire process from patch structure, hydrological connectivity, disturbance propagation, response identification, to the construction of impact indices. This method constructs a composite ecological graph model, integrates hydrological pathways and heterogeneous ecological function edge weights, and combines graph neural networks to model patch structure evolution, thereby identifying ecological propagation pathways and sensitive lineages. Finally, it outputs a time-series biodiversity index reflecting the intensity and lag of disturbance effects. This technology overcomes the static, fragmented, and poorly coupled problems of existing technologies, significantly improving the accuracy, timeliness, and predictive ability of dynamic response assessment to ecological disturbances, and possesses broad adaptability and practical value for ecological protection.

[0021] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the dynamic assessment method for the impact of engineering construction on biodiversity described in this embodiment of the invention;

[0024] Figure 2 This is a schematic diagram of the dynamic assessment system structure for the impact of engineering construction on biodiversity as described in this embodiment of the invention.

[0025] In the diagram: 701, Acquisition Unit; 702, Processing Unit; 703, Simulation Unit; 704, Identification Unit; 705, Backtracking Unit; 706, Analysis Unit. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0027] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0028] Example 1

[0029] This embodiment provides a method for dynamically assessing the impact of engineering construction on biodiversity.

[0030] See Figure 1 The figure shows that the method includes steps S1, S2, S3, S4, S5 and S6.

[0031] Step S1: Obtain the basic ecological dataset, which includes a water system distribution map, a forest wetland patch structure change map, and multi-time period quadrat community data;

[0032] Understandably, this step first involves acquiring a basic ecological dataset with temporal series characteristics and spatial heterogeneity, tailored to the unique landforms and ecological patterns. This dataset includes water system distribution maps, forest and wetland patch structure change maps, and multi-time-phase quadrat community data. Water system extraction relies on the fusion of high-precision LiDAR topographic data and multispectral remote sensing imagery to accurately reconstruct irregular water system structures formed by geomorphic processes such as travertine deposition, faulting, and closed depressions. These structures play a crucial role in hydrological connectivity and ecological function migration. Forest and wetland patch structures are dynamically identified through object-oriented remote sensing classification techniques and temporal change detection algorithms, dynamically recognizing patch fragmentation, marginalization, and migration trends caused by geological disaster control projects. Quadrat community data originates from multi-time-phase survey plots established in typical habitats, recording species composition, functional taxa classification, and their response characteristics, providing empirical support for subsequent modeling of ecological node semantics. This step, by breaking down the barriers between spatial layer information and biological attribute data, lays a unified data foundation for conducting spatial-functional-temporal multi-dimensional ecological map structure modeling. Its technical effectiveness is reflected in its ability to accurately reflect the structural evolution potential of complex regional ecosystems under disturbance backgrounds, and to provide high-quality input for subsequent graph-based dynamic ecological propagation mechanisms, thereby improving the spatial resolution and ecological interpretability of the entire process assessment.

[0033] Step S2: Based on the basic ecological dataset, construct the ecological patch map structure and perform hydrological connectivity edge construction to obtain a composite ecological map model that reflects spatial structural variations.

[0034] Understandably, this step, based on the acquired basic ecological dataset, integrates the spatial attributes of ecological patches with the dynamic relationship of hydrological processes to construct a composite ecological map structure with spatiotemporal semantics. It transforms the traditional static patch distribution map into a dynamically evolving ecological atlas with multimodal information fusion capabilities, effectively capturing structural variation trends within the scenic area driven by engineering disturbances, hydrological reconstruction, and community migration, significantly enhancing the modeling ability and prediction accuracy of ecological response processes. In this step, step S2 includes steps S21, S22, S23, and S24.

[0035] Step S21: Based on the water system distribution map in the basic ecological dataset and the topographic data of the preset digital elevation model, perform hydrological connectivity identification processing, and calculate the minimum cumulative cost path of surface runoff based on the CostDistance model to obtain the hydrological connectivity attenuation matrix between patches.

[0036] It is understandable that the water system in this step is a unique geomorphic phenomenon resulting from travertine biodeposition. Its water distribution is jointly controlled by surface calcareous crust, micro-topographic undulations, and underground currents. Traditional hydrological models based on the main water system are insufficient to accurately reflect its local hydrological processes. Therefore, this step innovatively introduces a hydrological modeling method based on the minimum cumulative cost path. This method couples the slope aspect and gradient from the topographic data of the digital elevation model with the geomorphic structure of the travertine formation area to construct a high-resolution surface flow resistance field. Based on this, the system calculates the cumulative hydrological cost of reaching each other via the minimum resistance path between any two ecological patches, resulting in an asymmetric, directional hydrological connectivity attenuation matrix. This matrix not only quantifies the accessibility of potential transport paths for water bodies or disturbance factors (such as pollution and sediment) between ecological units but also reflects the declining flow capacity influenced by factors such as travertine barriers and slope closure. This process effectively solves the problem of difficult hydrological accessibility modeling in highly geomorphically complex areas. This step provides a connectivity representation that integrates micro-topography, special travertine structures, and ecological patch layouts. This makes the edge weights in subsequent ecological graph structure modeling no longer simple spatial distances or artificial settings, but rather dynamic coupling indicators that reflect real geomorphological and hydrological processes. This enhances the structural credibility and evolutionary explanatory power of the graph model in ecological process modeling.

[0037] This step first uses terrain preprocessing algorithms (such as depression filling, flow direction calculation, and slope weight extraction) to construct hydrological consistency in the digital elevation model. Hydrological consistency refers to ensuring that, through terrain data and reasonable model parameter settings, the flow path, watershed division, and hydrological process simulation conform to actual surface topographic features and accurately reflect natural water flow behavior. Then, each ecological patch in the patch layer is defined as a source-sink pair, and the actual watershed division in the river system distribution map is used as the constraint boundary to form local flow units. The system then uses the CostDistance algorithm to establish a minimum resistance path model for water propagation. This model assigns each grid cell a composite resistance value composed of slope, travertine backwater factor, and topographic roughness, and uses a variant of the Dijkstra algorithm to calculate a minimum "cost" path between each pair of patches; the cost is the cumulative sum of hydrological resistance.

[0038] Ultimately, each patch corresponds to a minimum hydrological path, and the cost of all paths constitutes a hydrological connectivity decay matrix. Smaller matrix elements indicate higher path accessibility and smoother potential flow, while larger elements indicate a significant trend of blockage or isolation. Furthermore, to enhance ecological interpretability, this matrix can be further integrated and normalized with attributes such as patch area and boundary shape to improve its adaptability to ecological processes (such as nutrient dispersal and wetland connectivity restoration).

[0039] Step S22: Based on the hydrological connectivity attenuation matrix, an ecological node map is constructed. The forest wetland patch structure change map and multi-time period quadrat community data are fused using a regional fusion method to extract spatial node sets representing different functional groups, thereby obtaining an ecological node map with community function labels.

[0040] Understandably, this step not only identifies the location and boundaries of patches but also aggregates spatial units with similar ecological functions, enabling them to participate in subsequent propagation modeling as semantically consistent and functionally coordinated nodes in the graph. This process employs a region fusion (graph-based region fusion) method, which, unlike traditional object-oriented patch classification, can dynamically adjust the ecological semantic units constructed by nodes while preserving the continuity of ecological structure boundaries.

[0041] This step begins by treating all spatial patches in the forest wetland patch structure change map as initial subdivision graph nodes, and assigning primary attribute features to each node, including patch morphology index (Fractal dimension), physical location, adjacent hydrological pathways, and evolutionary trend (obtained through multi-time period patch overlap analysis). Simultaneously, multi-time period quadrat community data are mapped to the corresponding patch regions, and functional attribute labels such as species composition and functional taxa (e.g., nitrogen-fixing plants, emergent plants, key nodes in the food chain) are annotated.

[0042] Subsequently, a regional fusion strategy was introduced, using functional similarity (based on the Tanimoto coefficient or attribute distance) and structural connectivity (based on low-cost edges in the hydrological attenuation matrix) as dual constraints to gradually merge functionally similar patches with hydrological connectivity tendencies in the original map. This process not only preserves the true spatial distribution and hydrological coupling relationships among ecological units under micro-topography, but also achieves accurate merging of ecological functions of graph nodes through joint structure-attribute optimization. The final node set reflects both spatial clustering characteristics and embeds biodiversity functional labels from quadrat data, constituting a functional ecological node map with ecological semantic expression capabilities.

[0043] Step S23: Perform edge weight modeling based on the ecological node graph and hydrological connectivity decay matrix, and construct multivariate heterogeneous edges using spectral clustering algorithm to obtain an ecological relationship graph with joint structure-function information;

[0044] Understandably, this step first constructs an initial edge connection framework using the cost value in the hydrological connectivity decay matrix, selecting node pairs with significant accessibility along the hydrological path as candidate edge targets. Subsequently, three core features are extracted between each node pair: consistency of water flow direction (based on the flow direction grid to determine whether the path conforms to the overall watershed confluence trend), differences in geomorphic resistance (such as slope difference and travertine deposition influence factors), and the degree of community functional redundancy (the degree of overlap of the community functional structures of the two nodes, calculated based on the Jaccard coefficient). These features, after standardization, are used as multi-dimensional attribute vectors, representing the multiple ecological coupling characteristics of each candidate edge.

[0045] The system then introduces a spectral clustering model to perform preliminary clustering analysis on the node graph, identifying subgraph regions with locally dense connections in structure or significant group characteristics in function. These clustering results serve as structural guidelines for edge weight construction. Based on this, a multivariate weighted combination function is used to synthesize edge weights from the three core attributes. The weight coefficients can be determined using information gain or minimum redundancy maximum relevance (mRMR) algorithms to enhance the ecological representativeness of the edges. The resulting ecological relationship graph not only preserves the basic topological connections in the graph structure but also encodes the complexity of ecological processes. The heterogeneity of the edges is manifested in the mixed existence of hydrophysical edges, functional compensation edges, and coupling edges.

[0046] Step S24: Perform graph structure perturbation expression processing based on the ecological relationship graph. By introducing a gated graph neural network to dynamically modulate and learn the edge weight changes, capture the changing trend of forest wetland patch connection strength before and after geological disaster management, and obtain a composite ecological graph model that reflects spatial structure variation.

[0047] Understandably, this step first inputs the ecological relationship graphs constructed under multiple time slices (corresponding to key node time points before, during, and after governance) as a dynamic graph sequence, and retains the weight changes of each edge at different times. The gated graph neural network dynamically controls the retention of historical states and the introduction of new states through a gating mechanism (the reset gate and update gate in the GRU structure), enabling the system to effectively learn the edge weight change patterns before and after disturbance events. For example, for node pairs whose hydrological flow is interrupted due to governance, their edge weights will decrease abruptly over time, while for artificially restored water connectivity areas, their edge weights will show an increasing trend after the disturbance. This network not only updates the state of the nodes themselves but also propagates and modulates the structure of their adjacent edges, enabling the entire graph to achieve "disturbance-sensitive reconstruction" in its structure.

[0048] The output is a temporally enhanced composite ecological graph model, in which node attributes retain ecological semantics, edge weights express the connection strength after temporal evolution, and the overall structure reflects the connection change trend of spatial patches under the background of governance disturbances. The spatial graph structure of this model will serve as the basis for subsequent disturbance path extraction and sensitivity analysis.

[0049] Step S3: Based on the composite ecological graph model, perform patch functional heterogeneity aggregation and disturbance link simulation processing to obtain the ecological propagation path and the sensitivity spectrum of each node;

[0050] Understandably, this step utilizes the established composite ecological graph model to focus on the propagation path of disturbances in the ecosystem and the modeling of node response differences, in order to reveal the transmission channels and impact diffusion mechanisms between functionally heterogeneous patches. This step is the first to integrate ecological function clustering with graph structure path propagation models, breaking the limitations of separate modeling of spatial structure and ecological attributes, and enhancing the ecological semantic expression ability of disturbance paths. In this step, step S3 includes steps S31, S32, S33, and S34.

[0051] Step S31: Extract patch functional indicators based on the composite ecological graph model. Obtain node functional feature tensors by applying a feature encoding mechanism based on community functional redundancy, species sensitivity and habitat rarity to the graph nodes.

[0052] Understandably, this system first uses quadrat community data, combined with the Grime triangular model of plant functional classification, to classify the species within each node functionally. Community functional redundancy is calculated based on the balance of species numbers within the same functional group; a higher value indicates stronger substitutability and greater resilience of the node in the functional dimension. Species sensitivity is calculated by weighting and summing a known list of environmentally sensitive species with their frequency of occurrence in the quadrat. The weights can be set based on local stress response thresholds; a higher value indicates a stronger response to minor disturbances and greater ecological vulnerability. Habitat rarity is extracted by calculating the standard deviation of the geographical distribution density and patch area of ​​nodes across the entire ecograph. Rare, small-area, isolated nodes score higher on this indicator, reflecting their irreplaceability within the community network.

[0053] The three indicators mentioned above are encoded into three-dimensional feature channels at the node level, constructing a node functional feature tensor that structurally corresponds one-to-one with the nodes in the ecological graph and semantically integrates information from multiple ecological processes. This tensor will serve as the core input variable in subsequent propagation path identification and response spectrum analysis, supporting multi-scale graph neural networks to efficiently model the influence and functional heterogeneity between nodes.

[0054] Step S32: Based on the node functional feature tensor, perform patch functional heterogeneity aggregation processing. By using the adaptive fuzzy C-means clustering algorithm, the nodes are divided into functional response subgroups. The aggregation boundary is adjusted by combining the edge density and local connectivity in the graph structure to obtain a functional heterogeneous sub-graph set.

[0055] Understandably, this step first relies on the node functional feature tensor constructed in the previous step, which includes high-dimensional ecological indicators such as community functional redundancy, species sensitivity, and habitat rarity, to establish a multi-dimensional ecological state description vector for each node. Considering the fuzzy boundaries and local cross-influences among these ecological indicators, the system introduces an adaptive fuzzy C-means clustering method to perform soft classification and aggregation of nodes.

[0056] The adaptive fuzzy C-means clustering method in this step introduces a structure guiding term and a semantic entropy adjustment factor into the distance function. This ensures that the membership degree of a node among multiple cluster centers is constrained by both ecological function similarity and dynamically adjusted based on its sensitivity to structural disturbances. This mechanism is particularly suitable for situations in ecosystems where nodes have similar functions but significantly different responses, such as two wetland nodes with nitrogen-fixing vegetation, but one of which is structurally isolated and more sensitive to disturbances. The adaptive fuzzy C-means clustering method can cluster it into a response-priority subclass.

[0057] The structure-guided clustering process considers graph structure features in the perturbation propagation path, such as path betweenness, coupling degree, and centrality, making it easier for adjacent high-propagation nodes to be assigned to the same functional response subset, thereby improving the propagation consistency and ecological connectivity interpretability of the results. The semantic entropy adjustment factor measures the change in global membership entropy caused by assignment to a cluster, limiting the excessive divergence or convergence of a node's membership distribution across all clusters, maintaining a reasonable range of ecological fuzziness, preventing problems such as "structural jump clustering" or "excessive intra-cluster ecological heterogeneity," and improving the semantic stability of ecological clusters.

[0058] Among them, habitat rarity is an important indicator used to measure the spatial scarcity and ecological substitutability of a certain ecological unit (such as species habitat, functional patch, ecological node) in the whole ecological network. It is a key factor in assessing its ecological protection priority, response vulnerability and sensitivity to disturbance propagation. It is represented by functional substitutability, which is the opposite of habitat rarity. Functional substitutability represents the similarity to other habitats and is calculated using a cosine function.

[0059] After initial clustering, the system further fine-tunes the boundaries by combining edge density, edge weights (determined by hydrological connectivity and functional similarity), and local connectivity indicators (such as local clustering coefficients) in the ecological graph structure. This process ensures that the subgraph partitioning is not only rationally aggregated in the functional space but also forms cohesive regions along the propagation paths in the graph structure. In the final output set of functionally heterogeneous subgraphs, each subgraph represents a relatively consistent ecological response unit in both the functional and structural spaces, characterized by strong cohesion and clear external disturbance propagation paths.

[0060] Step S33: Perform disturbance link simulation processing based on the functional heterogeneity sub-graph set. By constructing a structural disturbance model based on random walk and combining it with construction disturbance time series data to simulate path crossing probability, a potential disturbance propagation path network is obtained.

[0061] Understandably, this step first treats each functionally heterogeneous subgraph as a propagation unit, identifying all nodes in the graph that overlap with the construction disturbance space or exhibit drastic changes in edge weights, as the set of disturbance source nodes. These nodes correspond to patches of areas that, during construction, disrupt travertine hydrological connectivity, compress wetland boundaries, or induce geomorphological alteration. Subsequently, for each source node, the system performs a weighted biased random walk operation within its own subgraph and adjacent subgraphs. The random walk transition probability considers not only the edge weights of the graph structure (determined by the ecological relationship graph with joint structure-function information) but also integrates the following two disturbance perception factors:

[0062] Disturbance impact factor time series weight: Based on the timestamps and disturbance intensity records of the construction stages (such as early blasting, mid-term slope support, and late-term water diversion), a disturbance time series decay function (such as exponential decay or memory gating adjustment) is assigned to each transfer.

[0063] Functional imbalance propagation tendency: By adjusting the path jump probability through the functional distance between nodes (derived from the tensor features of S31), edges with strong functional heterogeneity are more likely to propagate their influence, reflecting the weighted reinforcement of the ecological "imbalance channel".

[0064] Each path accumulates perturbation traversal probabilities during its journey, forming a perturbation propagation probability field radiating outward from the source node. The system performs multiple rounds of random sampling and aggregation on the propagation paths within each subgraph, ultimately summing them into a potential perturbation propagation path network. Its structure is characterized by a perturbation diffusion map with strong directionality, uneven weighting, and high path overlap.

[0065] Compared to traditional shortest path or full-graph walk methods, this step introduces a structural perturbation perception mechanism, which matches path selection with the ecological vulnerability between patches. Moreover, by integrating functional distance and propagation tendency, the path network output by the system is not only a connected graph in terms of graph structure, but also an explicit expression of potential ecological collapse chains, providing key support for subsequent sensitive lineage identification and regulation strategy design.

[0066] Step S34: Perform sensitivity spectrum identification processing based on the disturbance propagation path network. By jointly ranking the maximum increment rate of node edge weight evolution and functional contribution, use spectral clustering to divide the high-response node subset to obtain the set of high-sensitivity nodes and their sensitivity spectrum in the ecological propagation path.

[0067] Understandably, this step first performs edge weight evolution analysis on all nodes in the disturbance propagation path network. Based on the edge weight time series learned by the graph neural network, the maximum increment rate of adjacent edges of each node is calculated, i.e., the maximum magnitude or maximum slope of the edge weight change before and after the disturbance, serving as an indicator of the node's disturbance response intensity in structural propagation. Simultaneously, the system calls the encoded node functional feature tensor and calculates the functional contribution of each node using principal component analysis. This value reflects the node's ecological maintenance weight within its respective functional module.

[0068] ,in, This represents the maximum increment rate of the adjacent edges of node i. This indicates that all adjacent edges of node i have been traversed. This indicates the location of the maximum increment in the time series (starting from time 2). Let ε represent the edge weight between node i and its neighboring node j at time t (output by the graph neural network), let ε represent the small constant to prevent the removal of zero, and let T represent the total time step (e.g., the number of perturbation time sequence frames).

[0069] Subsequently, the system constructs a joint ranking matrix based on these two dimensions as the main axes, performs dual-index score ranking on the nodes, and identifies the set of candidate sensitive nodes with the highest scores. On this basis, a spectral clustering method is further introduced. Based on the Laplace matrix of the perturbation propagation path network, its eigenvectors are calculated and an embedding space is constructed. The candidate sensitive nodes are projected into this space and clustered according to response intensity and relative topological position, thereby dividing a subset of high-response nodes with strong response synergy and similar functional positions.

[0070] The final output is a set of highly sensitive nodes and its sensitivity spectrum: each highly sensitive node has clear functional attributes (such as high dependence on wetland stability or hydrological connectivity), structural attributes (located at the bottleneck of the disturbance path), and its position in the cluster spectrum, constituting the most vulnerable and critical response unit in the ecological network.

[0071] Step S4: Based on the ecological propagation path and the sensitivity spectrum of each node, perform spectral domain decomposition and key propagation node identification to obtain the ecological intermediary nodes and propagation weights in the disturbance response network.

[0072] Understandably, this step, based on the previously obtained ecological propagation path network and its corresponding node sensitivity spectrum, further identifies ecological intermediary nodes in the disturbance transmission process—that is, structural cores that play a bridging, guiding, or diffusion role in multiple propagation paths. Simultaneously, each intermediary node is assigned a precise propagation weight to characterize its true ecological niche and control in system-level disturbance diffusion. In this step, step S4 includes steps S41, S42, S43, and S44.

[0073] Step S41: Based on the ecological propagation path and the sensitivity spectrum of each node, perform Laplace spectral decomposition processing on the propagation path map. By constructing the normalized Laplace matrix of the perturbation propagation map and calculating its eigenvalue-eigenvector pairs, a set of spectral domain features for path energy distribution identification is obtained.

[0074] Understandably, this step first represents the ecological propagation path network G=(V,E) as a weighted graph, where nodes V represent functional patches, edges E represent perturbation propagation paths, and edge weights incorporate hydrological connectivity, functional redundancy, and propagation probability (obtained from perturbation link simulation). Subsequently, the normalized Laplace matrix of this network is constructed. This Laplace matrix helps maintain the stability of spectral domain characteristics under variations in node number or graph density, making it suitable for structural analysis of heterogeneous patch networks.

[0075] Next, the system performs eigenvalue decomposition on the Laplace matrix to obtain the first k eigenvalue-eigenvector pairs. The eigenvalues ​​represent the frequency characteristics of the perturbation propagation path in the graph, while the eigenvectors represent the structural weight of each node in that propagation mode. Low-order eigenvalues ​​correspond to global graph connectivity, reflecting slow diffusion and globally stable processes; high-order eigenvalues ​​capture boundary breaks and local abrupt change paths, serving as important criteria for identifying chain perturbations and structural imbalances.

[0076] The system combines the projection values ​​of all nodes into these spectral vectors to form a graph domain feature set, i.e., a spectral matrix of nodes × modes. This set can be used for subsequent analyses such as spectral energy calculation, propagation energy focusing identification, and intermediate node selection. In particular, the system also retains the magnitude of spectral feature changes before and after the disturbance to measure the "frequency activation effect" of the disturbance event on the network propagation basis.

[0077] This step projects the disturbance propagation path from the graph space to the frequency domain space, providing pattern recognition capabilities for structural transmission mechanisms, enhancing the visualization and quantitative explanation of system-level infectious paths, and strengthening the analytical power of the ecological graph structure for dynamic responses to disturbances. It avoids relying solely on shallow indicators such as node degree and adjacency, and instead uses the spectral morphology to comprehensively characterize the propagation resilience and instability threshold of the ecosystem.

[0078] Step S42: Calculate the energy concentration of propagation nodes based on the spectral domain feature set. Identify propagation path segments with significantly concentrated spectral energy using the maximum spectral energy density projection method to obtain local path sub-graphs with perturbation aggregation effects greater than a preset threshold.

[0079] Understandably, this step involves the system first calculating the energy of the spectral projection values ​​of each node in the graph domain feature set across various eigenvectors (corresponding to different propagation modes). The algorithm used is the maximum spectral energy density projection method, which calculates the spectral energy density of each node in the spectral vector set as follows: Where E(i) represents the maximum spectral energy density at node i. Let represent the component value at node i in the l-th eigenvector (i.e., the projection of the node onto the l-th mode), and k represent the number of spectral modes taken.

[0080] Subsequently, the system extracts propagation path segments using a sliding window across the entire map and aggregates the average total energy density of nodes within each path segment as the perturbation clustering index for that segment. All path segments are sorted, and an adaptive energy threshold (e.g., mean + 1.5 × standard deviation) based on the statistical characteristics of the energy distribution across the entire map is set to filter out path segments with significantly higher energy densities than the background, thus forming a set of highly perturbation clustered paths.

[0081] These highly perturbation segments are re-aggregated into local path subgraphs with continuous structure and smooth edge weight transfer, which are the regions in the propagation graph that exhibit strong concentrated perturbation responses in the frequency domain. This step constructs local structural units with the strongest perturbation effects, the largest response amplitudes, and the most prominent control significance, providing a structural enclosure and energy focusing basis for key node identification.

[0082] Step S43: Based on the local path subgraph, ecological intermediary node identification processing is performed. Through the joint screening mechanism of betweenness centrality and path coupling degree, key nodes with bridging effect in multiple propagation paths are extracted to obtain the set of ecological intermediary nodes in the perturbation response network.

[0083] Understandably, this step involves the system first calculating the betweenness centrality of nodes in each local path subgraph. This metric measures the frequency with which a node acts as a transit node across all shortest paths. In the ecological propagation graph, nodes with high betweenness centrality often serve as connecting hubs between different paths, controlling the main channels through which disturbances spread from one functional subdomain to another. Simultaneously, the system introduces a path coupling index to measure the degree of overlap in a node's participation across multiple disturbance paths. Specifically, it is defined as follows: for each node, the frequency with which it appears simultaneously in every independent disturbance path is counted, and weighted by the functional heterogeneity between these paths (such as differences in the functions of the subgraphs connected by the paths). A higher value indicates a more significant bridging effect in "cross-functional propagation."

[0084] Next, the system normalizes and integrates betweenness centrality and path coupling degree into a score, and uses a linear weighting method to jointly rank all candidate nodes, selecting nodes with scores exceeding the network mean plus standard deviation as the final set of intermediate nodes. To enhance ecological explanatory power, the system also retains the list of participating paths, bridging function type (e.g., forest-wetland, travertine-bare slope), and edge weight change rate for each intermediate node, facilitating subsequent visualization modeling and attribution of influencing factors. This accurately identifies key nodes with bridging effects in multipath propagation, providing precise control points for disrupting disturbance links or guiding ecological restoration paths.

[0085] Step S44: Perform propagation weight modeling based on the set of ecological intermediary nodes. Using a multi-factor data normalization method based on node redundancy and path coupling coefficient, assign response weights to each intermediary node on the engineering construction disturbance path to obtain the ecological intermediary nodes and propagation weights in the disturbance response network.

[0086] Understandably, this step begins by constructing the propagation effect factor of each intermediary node within the engineering construction path, based on two core dimensions. Firstly, the system calculates node redundancy to measure whether a node can be replaced within its functional subgraph. This calculation is based on the redundancy of the node's functional group (e.g., the number of nodes with the same functional type) and the similarity of its ecological status (e.g., the overlap of functional axis scores). Nodes with low redundancy are typically indispensable control points on a particular functional path, and their propagation weight should be relatively high. Engineering construction disturbance paths refer to the propagation channels through which certain nodes in the ecosystem (e.g., wetlands, forests) are affected by engineering activities (e.g., construction, infrastructure development). These paths will vary depending on the sensitivity and structural changes of the ecosystem.

[0087] On the other hand, the system calculates the path coupling coefficient of the node in the propagation path, quantifying the strength of the synergistic effect formed by the node's participation as a bridging node in the convergence of multiple propagation paths. The path coupling coefficient comprehensively considers two factors: how many paths with strong functional heterogeneity the node appears on (cross-functional coupling), and the synchronicity of edge weight changes between these paths (perturbation cooperation).

[0088] ,in, This represents the path coupling coefficient of node v. This represents the set of all perturbation propagation paths that include node v. and belong Two different disturbance propagation paths.

[0089] Subsequently, the system employs a multi-factor data normalization method to allocate the propagation capability of each intermediary node on each perturbation path. This algorithm constructs the initial propagation impact score by multiplying the inverse of node redundancy (representing non-substitutability) by the path coupling coefficient. Then, through local path-level normalization and full-graph-level mapping scaling, it ensures that the propagation weights possess comparable, non-linear scaling, and context-adaptive characteristics. Finally, each intermediary node is assigned a propagation weight value between 0 and 1 on each participating path, forming a propagation weight matrix with the structure {node ID, path ID, weight value}.

[0090] Finally, the set of ecological intermediary nodes and their propagation weight distribution on each path in the disturbance response network are obtained. This structure is used as the core input for subsequent disturbance backtracking modeling, structural reversibility learning, and stability trend quantitative analysis.

[0091] Step S5: Based on the ecological intermediary nodes and propagation weights, perform reversibility modeling and path entropy backtracking analysis to obtain the community reconstruction trend index and functional stability decay curve.

[0092] It is understandable that this step, based on the set of ecological intermediary nodes and their propagation weight matrix obtained in the previous step, further conducts disturbance reversibility modeling and path entropy backtracking analysis, aiming to quantitatively characterize the ecological network's recovery potential, reconstruction trend, and functional stability evolution under disturbance at the system level. This step not only reveals whether the disturbance can be spontaneously absorbed or repaired by the ecosystem, but also outputs actionable assessment indicators: the community reconstruction trend index and the functional stability decay curve, providing key support for ecological intervention design and time window optimization. In this step, step S5 includes steps S51, S52, S53, and S54.

[0093] Step S51: Based on the ecological intermediary nodes and propagation weights, perform perturbation reversibility state modeling processing, and simulate the deletion and reconstruction process of perturbation nodes by constructing a weighted graph convolutional autoencoder to obtain the reconstruction difference tensor of the ecological network structure before and after perturbation.

[0094] Understandably, this step first uses the original graph structure in the perturbation response network as input, where nodes represent ecological patches or functional units, edges are perturbation propagation paths, and edge weights are provided by the propagation weight matrix, expressing the strength and susceptibility of the perturbation signal propagating along each path. Each node in the set of ecological intermediary nodes is regarded as a key perturbation entry point, simulating its weakened, ineffective, or disconnected state during a perturbation event, thus constructing a perturbation graph. That is, based on the original graph, the connection structure of the intermediary nodes is reduced or the edge weights are attenuated.

[0095] Subsequently, the original and perturbed graphs are simultaneously input into a weighted graph convolutional autoencoder framework. This model consists of a graph convolutional encoder and a graph decoder: the encoder extracts the latent space representation of nodes through multi-layer graph convolution, fusing their functional properties with structural adjacency features; the decoder attempts to reconstruct the adjacency matrix or edge weight tensor based on the encoded representation. The model is trained separately on the two graphs, allowing it to learn the structural difference patterns before and after perturbing, and outputs the connection reconstruction differences for each pair of nodes in the two states.

[0096] To more clearly express the spatial distribution characteristics of perturbation reversibility, the system defines a structural reconstruction difference tensor: Among them, the D structure reconstructs the difference tensor, a three-dimensional tensor structure with dimensions of N X N X T , N Indicates the total number of network nodes. T Indicates the number of time steps or the number of disturbance simulation stages. This represents the structural reconstruction difference between node i and node j at the t-th perturbation time step. This represents the value of the adjacency matrix element at the t-th time step before the disturbance occurs. This represents the adjacency matrix element value at the t-th time step after the disturbance occurs. The larger the difference in the tensor, the more sensitive the structural connections are to disturbances and the poorer the recovery ability; the smaller the difference, the stronger the functional resilience and structural reconfigurability of the system.

[0097] Step S52: Perform path perturbation information entropy evolution analysis based on the reconstructed difference tensor. By constructing a path information entropy sequence based on node state transition probability and combining the perturbation cascade amplitude change, calculate the entropy diffusion rate of the perturbation propagation path to obtain a set of path entropy backtracking functions.

[0098] Understandably, this step first extracts the connection difference values ​​between node pairs in each disturbance propagation path based on the structural reconstruction difference tensor, and constructs the local state space formed by the structural changes before and after the disturbance.

[0099] Next, the system defines its information entropy evolution sequence on a path-by-path basis: ,in, This represents the structural perturbation information entropy of path p at the t-th perturbation time step. This represents all adjacent node pairs in path p. Let represent the normalized value of the connection difference of node pair (i, j) in the t-th step of the perturbation, and p represent a certain perturbation propagation path.

[0100] To further measure the propagation and dissipation rate characteristics of disturbances along different paths, the system introduces an entropy diffusion rate index, performing first-order difference and trend fitting on the entropy sequence of each path. High disturbance information entropy indicates that the disturbance spreads rapidly along the path and is difficult to recover, while low entropy indicates that the path information dissipates slowly or has a certain structural resilience.

[0101] Finally, the system packages the entropy sequence, diffusion rate, stability fluctuation, and other parameters of each path, outputting a set of path entropy backtracking functions. This set will provide basic data support for subsequent steps such as community reconstruction trend identification, functional decay curve fitting, and risk zoning.

[0102] Step S53: Perform community response trend clustering processing based on the path entropy backtracking function set, and extract the structural response pattern by jointly embedding the entropy change rate and community structural redundancy within the time window into a preset sparse representation model to obtain the community reconstruction trend index.

[0103] Understandably, this step first performs time window warping on the information entropy evolution sequence of each path to ensure that different paths are comparable on a time scale. To this end, a dynamic time warping algorithm is introduced to perform multi-window alignment processing on the entropy curve, solving the scale differences caused by the lag and asynchrony of path disturbance response.

[0104] Next, the system uses the entropy evolution sequence of each path as the feature vector input and employs a pre-defined sparse representation model to perform low-rank encoding on all paths. By optimizing the reconstruction error between paths and the lexicon sparsity, the system extracts the most discriminative perturbation trend subspace. Each principal component axis in this subspace represents a typical perturbation evolution mode, such as "rapid diffusion-slow recovery type", "continuous high-entropy oscillation type", and "local rebound stability type".

[0105] The system then projects each path onto the perturbation trend subspace and performs spectral clustering, classifying them into minor perturbation response categories based on the path similarity matrix. For each path category, the system reverse maps the node community region it covers, generates community response trend labels, and statistically analyzes the structural connectivity fluctuations, functional stability residues, and response reversibility exhibited by the community during the perturbation process, forming a community reconstruction trend index.

[0106] The community reconstruction trend index is shown below: ,in, CRTI (C k ) represents community C k The reconstruction trend index Represents community C k The fluctuation amplitude of the structural connectivity during the disturbance process. Res (C k ) represents community C k Functional stability residual ratio, Rev (C k ) represents community C k The response reversibility index is given by α, β, and γ, which are weighting parameters representing the relative importance of the three factors.

[0107] The index combines path-level disturbance intensity, community-level functional redundancy, and structural volatility to assign a comprehensive value. The higher the value, the greater the possibility that the community in the region will undergo reorganization, breakup, or irreversible migration during disturbances, and the more complex the ecological evolution trend. The lower the value, the more likely the system has the potential to return to a steady state or maintain functional conservation.

[0108] Step S54: Quantify the ecological function stability based on the community reconstruction trend index. By constructing a multi-scale response amplitude function and performing wavelet reconstruction analysis, the functional stability decay curve is obtained.

[0109] Understandably, this step first collects time-series data of functional variables of each community node during the perturbation simulation process. Specifically, this includes, but is not limited to, the abundance changes of species functional groups (such as nitrogen fixation and water regulation), functional redundancy, and system interaction (such as species coexistence and energy flux density), etc., and uses these as multi-channel time-series tensors. , where n is the number of communities, d is the functional dimension, and T is the perturbation time step.

[0110] The system uses principal component analysis to reduce the dimensionality of the functional time series of each community, extracting the perturbation-dominated functional fluctuation pattern and constructing a one-dimensional representative variable. Then, wavelet transform analysis is introduced to analyze the series, extracting its changing trends and abrupt change points across multiple time scales, thereby constructing a continuous function curve describing the evolution of functional state over time, defined as: , among which, S k(t) represents community C k The continuous curve of functional state evolution. This represents the fluctuation sequence of the dominant function of the community. Let the wavelet basis function be represented by scale a and translation b. The integrand time variable is used to represent the historical information of the integration of the original sequence, t represents the current time step, and T is the number of perturbation time steps.

[0111] Step S6: Based on the community reconstruction trend index and functional stability decay curve, perform ecological response sustainability analysis to obtain the time series biodiversity impact index.

[0112] It is understandable that this step integrates the community reconstruction trend index and functional stability decay curve generated in the previous steps to conduct an ecological response sustainability analysis oriented towards the dynamic change trend of the ecosystem after disturbance. The goal is to eventually construct an index function that can continuously reflect the intensity, duration and carrying capacity of the ecosystem's functions, namely the time series biodiversity impact index. In this step, step S6 includes steps S61, S62, S63 and S64.

[0113] Step S61: Based on the community reconstruction trend index and functional stability decay curve, perform ecological response delay modeling. By constructing a variable structure state equation based on a time-delay nonlinear dynamic system, capture the lag response time period of the biological community after disturbance, and obtain a set of multi-timescale ecological response delay functions.

[0114] Understandably, this step first constructs a two-dimensional input tensor based on the reconstructive trend index of each community and its corresponding stability decay curve, serving as the time-space function input for the ecosystem's dynamic response. Building upon this, the system introduces a time-delay nonlinear dynamic system modeling framework to construct the response function: ,in, Represents community C k The response function, Represents community C k The function state evolution is a continuous curve, where λ represents the natural decay coefficient, the rate at which the control system returns to a steady state, μ represents the response inertia weight, affecting the feedback strength of the time delay term to the current state, and σ represents the sigmoid activation function used for nonlinear modeling. Let X be the integrand time variable, representing the historical information of the integration of the original sequence; let η represent the external driving strength coefficient, controlling the direct influence of the input tensor on the response; and let X be the integrand time variable. k，t Represents community C k Trend-decay joint input value at time step t.

[0115] Then, the delay response functions of all nodes are unified and organized to form a set of ecological response delay functions with multiple time scales, which are used to support the prediction window of disturbance impact, the identification of key turning points and the prediction of recovery time.

[0116] Step S62: Perform disturbance effect persistence estimation processing based on the set of ecological response delay functions. By calculating the multi-window alignment error between the trend index and the stability curve under dynamic time warping, obtain the ecological effect persistence distribution map at different stages after disturbance.

[0117] Understandably, in this step, the system first performs multi-time-window sliding analysis on each delay function, introducing window length and disturbance response threshold to calculate whether the delay response value of each community exceeds the ecological disturbance sensitivity threshold in different time periods. The following function is constructed for each community node: , where D k (t) represents community C k At time t, whether the state is in a highly sensitive disturbance response state is indicated by 1 for yes and 0 for no. ω represents the sliding window length (unit: time step), z represents the offset step number within the sliding window, and S... k (tz) represents community C k At time step tz, the response delay state value is θ, which represents the sensitivity threshold for ecological disturbance response (such as a preset structure-function tolerance upper limit).

[0118] The function outputs whether the disturbance effect is active for each time period. After the sliding window is shifted, a binary sequence representing the continuous activity of the disturbance is formed, and then the duration, frequency, maximum span, and other time characteristics are calculated.

[0119] Next, the system introduces a dynamic time warping method to align and compare the perturbation delay functions between different communities, identify which communities have similar perturbation effect persistence patterns, and construct a clustering heatmap based on function similarity to reveal the synchronous / asynchronous diffusion pattern of perturbation effects in space.

[0120] Step S63: Based on the ecological effect persistence distribution map, the species function contribution is reconstructed. By constructing a community response clustering model based on fuzzy hierarchical weighted clustering, key functional groups in each persistence stage are identified, and a set of biodiversity structure vectors under time slices is obtained.

[0121] It is understandable that this step is based on the data on the persistence of perturbation response and the species functional performance matrix to construct a correlation matrix of the co-change between each species and the community perturbation response.

[0122] Then, the functional contribution function for each species during the duration of the perturbation is defined as follows: Among them, G wC represents the overall functional contribution during the duration of species w disturbance. k,w Indicates species w in community C k Functional relevance during disturbance response This represents the functional performance vector of species w. This represents the functional response strength of the community at time step t. This represents the sensitive time period for disturbance in a community, and k represents the k-th community.

[0123] By clustering the functional contribution function curves of each species, species with high functional response, high stability, or functional lag recovery are identified. Key functional supporting species are then selected by combining indicators such as betweenness centrality and functional redundancy in the community structure map to form a pool of species that prioritize recovery intervention (key functional groups). Subsequently, the collective response of member species under each functional group in the perturbation situation is integrated into a vector representation, thereby generating a set of biodiversity structure vectors for each time slice.

[0124] Step S64: Perform time series index construction processing based on the biodiversity structure vector set to obtain the biodiversity impact index.

[0125] Understandably, this system defines a weighted aggregation function at each time step to fuse the structural state vectors from the previous few time slices, and introduces a perturbation-sensitive factor as an adjustment term, enabling the index construction process to adaptively adjust the smoothing window weights when the perturbation intensity fluctuates significantly. This mechanism ensures that the generated exponential curve reflects both the system's short-term drastic response and preserves long-term structural trend changes, ultimately outputting a biodiversity impact index.

[0126] The biodiversity impact index is shown below: ,

[0127] Wherein, BII(t) represents the biodiversity impact index, indicating the degree of functional stability of the ecosystem relative to the reference state at time step t, with a value range of [0,1]. t Here, represents the normalization factor, and s represents the index of any time step within the time window. Indicates the length of the time window, α s V represents the perturbation-sensitive weighting factor. s V represents the biodiversity structure vector. ref Indicates the reference state structure vector.

[0128] Example 2

[0129] like Figure 2As shown in the figure, this embodiment provides a dynamic assessment system for the impact of engineering construction on biodiversity. The system includes an acquisition unit 701, a processing unit 702, a simulation unit 703, an identification unit 704, a backtracking unit 705, and an analysis unit 706.

[0130] The acquisition unit 701 is used to acquire a basic ecological dataset, which includes a water system distribution map, a forest wetland patch structure change map, and multi-time period quadrat community data.

[0131] The processing unit 702 is used to construct the ecological patch map structure and perform hydrological connectivity edge construction based on the basic ecological dataset to obtain a composite ecological map model that reflects spatial structure variation.

[0132] The simulation unit 703 is used to perform patch functional heterogeneity aggregation and disturbance link simulation processing based on the composite ecological graph model to obtain the ecological propagation path and the sensitivity spectrum of each node.

[0133] The identification unit 704 is used to perform spectral decomposition and key propagation node identification processing based on the ecological propagation path and the sensitivity spectrum of each node, so as to obtain the ecological intermediary nodes and propagation weights in the disturbance response network.

[0134] The backtracking unit 705 is used to perform reversibility modeling and path entropy backtracking analysis based on the ecological intermediary nodes and propagation weights to obtain the community reconstruction trend index and functional stability decay curve.

[0135] Analysis unit 706 is used to perform ecological response persistence analysis based on the community reconstruction trend index and functional stability decay curve to obtain the time series biodiversity impact index.

[0136] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0137] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0138] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for dynamically assessing the impact of engineering construction on biodiversity, characterized in that, include: Acquire a basic ecological dataset, which includes a water system distribution map, a forest wetland patch structure change map, and multi-time period quadrat community data; Based on the aforementioned basic ecological dataset, ecological patch map structure construction and hydrological connectivity edge construction are performed to obtain a composite ecological map model that reflects spatial structural variations. Based on the composite ecological graph model, patch functional heterogeneity aggregation and disturbance link simulation are performed to obtain the ecological propagation path and the sensitivity spectrum of each node. Based on the ecological propagation path and the sensitivity spectrum of each node, spectral domain decomposition and key propagation node identification are performed to obtain the ecological intermediary nodes and propagation weights in the disturbance response network. Based on the aforementioned ecological intermediary nodes and propagation weights, reversibility modeling and path entropy backtracking analysis are performed to obtain the community reconstruction trend index and functional stability decay curve. Based on the community reconstruction trend index and functional stability decay curve, ecological response sustainability analysis was performed to obtain the time series biodiversity impact index. The process, based on the composite ecological map model, includes patch functional heterogeneity aggregation and perturbation link simulation, including: Based on the composite ecological graph model, patch functional indicators are extracted and processed. By applying a feature encoding mechanism based on community functional redundancy, species sensitivity and habitat rarity to the graph nodes, node functional feature tensors are obtained. Based on the node functional feature tensor, patch functional heterogeneity aggregation processing is performed. The nodes are divided into functional response subgroups by using an adaptive fuzzy C-means clustering algorithm, and the aggregation boundary is adjusted by combining the edge density and local connectivity in the graph structure to obtain a functional heterogeneous sub-graphet. Based on the functional heterogeneity sub-graphet, disturbance link simulation processing is performed. By constructing a structural disturbance model based on random walk and combining it with construction disturbance time series data to simulate path crossing probability, a potential disturbance propagation path network is obtained. Sensitivity spectrum identification is performed on the perturbation propagation path network. The nodes are jointly ranked by the maximum increment rate of the node edge weight evolution and the functional contribution. High-response node subsets are divided by spectral clustering to obtain the set of high-sensitivity nodes and their sensitivity spectrum in the ecological propagation path. The process includes spectral decomposition and key propagation node identification based on the ecological propagation path and the sensitivity spectrum of each node, including: Based on the ecological propagation path and the sensitivity spectrum of each node, the propagation path graph is subjected to Laplace spectral decomposition. By constructing the normalized Laplace matrix of the perturbation propagation graph and calculating its eigenvalue-eigenvector pairs, a set of spectral domain features for path energy distribution identification is obtained. Based on the spectral domain feature set, the energy concentration of propagation nodes is calculated and processed. By using the maximum spectral energy density projection method, propagation path segments with significant spectral energy concentration are identified, and local path sub-graphs with perturbation aggregation effects greater than a preset threshold are obtained. Based on the local path subgraph, ecological intermediary nodes are identified. Through a joint screening mechanism of betweenness centrality and path coupling, key nodes with bridging effects in multiple propagation paths are extracted to obtain the set of ecological intermediary nodes in the perturbation response network. Based on the set of ecological intermediary nodes, propagation weight modeling is performed. By using a multi-factor data normalization method based on node redundancy and path coupling coefficient, the response weight of each intermediary node on the engineering construction disturbance path is assigned, thereby obtaining the ecological intermediary nodes and propagation weights in the disturbance response network.

2. The dynamic evaluation method according to claim 1, characterized in that... Based on the aforementioned basic ecological dataset, ecological patch map structure construction and hydrological connectivity edge establishment are performed, including: Based on the water system distribution map in the basic ecological dataset and the topographic data of the preset digital elevation model, hydrological connectivity identification processing is performed. The minimum cumulative cost path of surface runoff is calculated based on the CostDistance model to obtain the hydrological connectivity attenuation matrix between patches. Based on the hydrological connectivity attenuation matrix, an ecological node map is constructed. The forest wetland patch structure change map and multi-time period quadrat community data are fused using a regional fusion method to extract spatial node sets representing different functional groups, thus obtaining an ecological node map with community function labels. Based on the ecological node graph and the hydrological connectivity attenuation matrix, edge weight modeling is performed, and multivariate heterogeneous edges are constructed using spectral clustering algorithm to obtain an ecological relationship graph with joint structure-function information. Based on the ecological relationship diagram, graph structure perturbation expression processing is performed. By introducing a gated graph neural network to dynamically modulate and learn edge weight changes, the changing trend of forest wetland patch connectivity before and after geological disaster management is captured, resulting in a composite ecological graph model reflecting spatial structure variation.

3. The dynamic evaluation method according to claim 1, characterized in that... Based on the aforementioned ecological intermediary nodes and propagation weights, reversibility modeling and path entropy backtracking analysis are performed, including: Based on the ecological intermediary nodes and propagation weights, the perturbation reversibility state modeling process is performed. By constructing a weighted graph convolutional autoencoder, the process of deleting and reconstructing perturbation nodes is simulated, and the reconstruction difference tensor of the ecological network structure before and after perturbation is obtained. The path perturbation information entropy evolution analysis is performed based on the reconstructed difference tensor. By constructing a path information entropy sequence based on the node state transition probability and combining the change in the amplitude of the perturbation cascade, the entropy diffusion rate of the perturbation propagation path is calculated, and a set of path entropy backtracking functions is obtained. The community response trend clustering is performed based on the path entropy backtracking function set. The structural response pattern is extracted by jointly embedding the entropy change rate and community structural redundancy within the time window into a preset sparse representation model to obtain the community reconstruction trend index. The ecological function stability is quantified based on the community reconstruction trend index. By constructing a multi-scale response amplitude function and performing wavelet reconstruction analysis, the functional stability decay curve is obtained.

4. A dynamic assessment system for the impact of engineering construction on biodiversity, characterized in that, include: The acquisition unit is used to acquire basic ecological datasets, which include water system distribution maps, forest wetland patch structure change maps, and multi-time period quadrat community data. The processing unit is used to construct the ecological patch map structure and perform hydrological connectivity edge construction based on the basic ecological dataset to obtain a composite ecological map model that reflects spatial structure variation. The simulation unit is used to perform patch functional heterogeneity aggregation and disturbance link simulation processing based on the composite ecological graph model to obtain the ecological propagation path and the sensitivity spectrum of each node. The identification unit is used to perform spectral domain decomposition and key propagation node identification processing based on the ecological propagation path and the sensitivity spectrum of each node, so as to obtain the ecological intermediary nodes and propagation weights in the disturbance response network. The backtracking unit is used to perform reversibility modeling and path entropy backtracking analysis based on the ecological intermediary nodes and propagation weights to obtain the community reconstruction trend index and functional stability decay curve. The analysis unit is used to perform ecological response persistence analysis based on the community reconstruction trend index and functional stability decay curve to obtain the time series biodiversity impact index. The simulation unit includes: The first simulation subunit is used to extract patch functional indicators based on the composite ecological graph model. By applying a feature encoding mechanism based on community functional redundancy, species sensitivity and habitat rarity to the graph nodes, the node functional feature tensor is obtained. The second simulation subunit is used to perform patch functional heterogeneity aggregation processing based on the node functional feature tensor. It divides the nodes into functional response subgroups by using an adaptive fuzzy C-means clustering algorithm, and adjusts the aggregation boundary by combining edge density and local connectivity in the graph structure to obtain a functional heterogeneous sub-graph set. The third simulation subunit is used to perform disturbance link simulation processing based on the functional heterogeneity sub-graphet. By constructing a structural disturbance model based on random walk and combining it with construction disturbance time series data to simulate path crossing probability, a potential disturbance propagation path network is obtained. The fourth simulation subunit is used to perform sensitivity spectrum identification processing based on the disturbance propagation path network. It sorts nodes by the maximum increment rate of edge weight evolution and functional contribution, and uses spectral clustering to divide high-response node subsets to obtain the set of high-sensitivity nodes and their sensitivity spectrum in the ecological propagation path. The identification unit includes: The first identification subunit is used to perform Laplace spectral decomposition of the propagation path graph according to the ecological propagation path and the sensitivity spectrum of each node. By constructing the normalized Laplace matrix of the perturbation propagation graph and calculating its eigenvalue-eigenvector pairs, a set of spectral domain features for path energy distribution identification is obtained. The second identification subunit is used to calculate the energy concentration of propagation nodes based on the set of spectral domain features. By using the maximum spectral energy density projection method, it identifies propagation path segments with significantly concentrated spectral energy and obtains local path sub-graphs with a disturbance aggregation effect greater than a preset threshold. The third identification subunit is used to identify ecological intermediary nodes based on the local path subgraph. Through a joint screening mechanism of betweenness centrality and path coupling, key nodes with bridging effects in multiple propagation paths are extracted to obtain the set of ecological intermediary nodes in the perturbation response network. The fourth identification subunit is used to perform propagation weight modeling based on the set of ecological intermediary nodes. By using a multi-factor data normalization method based on node redundancy and path coupling coefficient, the response weight of each intermediary node on the engineering construction disturbance path is assigned to obtain the ecological intermediary nodes and propagation weights in the disturbance response network.

5. The dynamic evaluation system according to claim 4, characterized in that, The processing unit includes: The first processing subunit is used to perform hydrological connectivity identification processing based on the water system distribution map in the basic ecological dataset and the topographic data of the preset digital elevation model. It calculates the minimum cumulative cost path of surface runoff based on the CostDistance model to obtain the hydrological connectivity attenuation matrix between patches. The second processing subunit is used to construct an ecological node map based on the hydrological connectivity attenuation matrix. It integrates the forest wetland patch structure change map and multi-time period quadrat community data through a regional fusion method, extracts a set of spatial nodes representing different functional groups, and obtains an ecological node map with community function labels. The third processing subunit is used to perform edge weight modeling based on the ecological node graph and the hydrological connectivity attenuation matrix, and to construct multivariate heterogeneous edges through spectral clustering algorithm to obtain an ecological relationship graph with joint structure-function information. The fourth processing subunit is used to perform graph structure perturbation expression processing based on the ecological relationship graph. By introducing a gated graph neural network to dynamically modulate and learn the edge weight changes, it captures the changing trend of forest wetland patch connection strength before and after geological disaster management, and obtains a composite ecological graph model that reflects spatial structure variation.

6. The dynamic evaluation system according to claim 4, characterized in that, The backtracking unit includes: The first backtracking subunit is used to perform perturbation reversibility state modeling based on the ecological intermediary nodes and propagation weights. By constructing a weighted graph convolutional autoencoder, it simulates the deletion and reconstruction process of perturbation nodes and obtains the reconstruction difference tensor of the ecological network structure before and after perturbation. The second backtracking subunit is used to perform path perturbation information entropy evolution analysis based on the reconstructed difference tensor. By constructing a path information entropy sequence based on node state transition probability and combining the perturbation cascade amplitude change, the entropy diffusion rate of the perturbation propagation path is calculated to obtain a set of path entropy backtracking functions. The third backtracking subunit is used to perform community response trend clustering based on the path entropy backtracking function set. By jointly embedding the entropy change rate and community structural redundancy within the time window into a preset sparse representation model, the structural response pattern is extracted to obtain the community reconstruction trend index. The fourth retrospective subunit is used to quantify the ecological function stability based on the community reconstruction trend index. By constructing a multi-scale response amplitude function and performing wavelet reconstruction analysis, the functional stability decay curve is obtained.

Images