A small and micro wetland ecological restoration optimization method based on multi-modal data

Abstract

This invention relates to the field of wetland ecological regulation technology, specifically a method for optimizing the ecological restoration of small wetlands based on multimodal data. The method includes: collecting multimodal monitoring information such as remote sensing image sequences, time-series data of physical and chemical parameters, and ecological sample data of the target area; constructing a dynamic three-dimensional ecological field; assimilating and fusing multi-source data using a spatial grid as a basis to form a comprehensive ecological state attribute value with timestamps; utilizing an improved ecological cellular automata algorithm, combined with an ecological restoration intervention library and the attribute differences of adjacent grids, completing the synchronous evolution calculation of the ecological state of the entire grid, and generating a spatialized ecological evolution field; selecting spatially connected stable grid clusters and implementing connectivity enhancement and structural optimization; and outputting restoration layout, implementation timeline, and engineering quantity configuration content that matches site conditions. This method achieves spatiotemporal integration of multi-source heterogeneous ecological data, refines the accuracy of regional ecological evolution prediction, and enhances the ability to regulate the spatial structure of wetlands.

Description

Technical Field

[0001] This invention relates to the field of wetland ecological regulation technology, and in particular to a method for optimizing the ecological restoration of small wetlands based on multimodal data. Background Technology

[0002] Small wetlands are an important component of regional ecosystems. Currently, ecological restoration assessments and scheme formulations often rely on analysis and calculations based on single types of monitoring data. Remote sensing images, ground sensor time-series data, and field ecological sample data from different sources are usually stored and analyzed independently. There is a lack of a unified spatial framework among various types of data, making it impossible to integrate and process multi-source information at a unified scale, and hindering the formation of a coordinated expression of spatiotemporal information.

[0003] Conventional ecological evolution simulations generally employ traditional cellular automata models, with simplistic evolution rule settings that rely solely on fixed neighborhood rules for state deduction, failing to adapt and adjust rules to integrate with ecological restoration intervention systems. Grid cell state evolution calculations neglect spatial differences in ecological attributes within a region, resulting in evolutionary results that lack refined spatial distribution characteristics and cannot output restoration layout content with spatial correspondence.

[0004] In the process of small-scale wetland restoration, traditional analysis models are insufficient to fully reconstruct the spatiotemporal evolution of regional ecosystems, and decentralized data processing methods result in insufficient dimensions for representing ecological status. Fixed evolutionary calculation rules are ill-suited to the diverse changes in wetland environments, and methods for identifying stable regions are rather crude, making it difficult to accurately identify spatially connected stable ecological units and to generate refined restoration plans with quantitative layout and temporal sequence. Therefore, it is necessary to establish a unified spatiotemporal data fusion architecture, optimize ecological evolution projection rules, and improve the refined generation model of wetland restoration plans. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an optimization method for the ecological restoration of small wetlands based on multimodal data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for optimizing the ecological restoration of small wetlands based on multimodal data, comprising: Acquire multimodal monitoring data of the target small wetland ecological restoration area. The multimodal monitoring data includes regularly collected remote sensing image sequences, continuous time-series data of physical and chemical parameters recorded by ground sensor networks, and ecological sample data from periodic field surveys. A dynamic three-dimensional ecological field is constructed to characterize the spatiotemporal evolution of the ecological restoration area. In the dynamic three-dimensional ecological field, multimodal monitoring data is assimilated and fused into a comprehensive ecological state attribute value with a timestamp, based on a spatial grid. The comprehensive ecological state attribute value is input into the improved ecological cellular automaton algorithm. The improved ecological cellular automaton algorithm performs synchronous evolution calculation on the comprehensive ecological state attribute value of each grid at the next moment based on the difference between the comprehensive ecological state attribute values ​​of the ecological restoration intervention measures library and the adjacent grids in the dynamic three-dimensional ecological field, and generates an ecological evolution field containing spatial explicit restoration schemes. An ecological steady-state network is extracted from the ecological evolution field. The ecological steady-state network consists of a grid cluster whose comprehensive ecological state attribute values ​​are within a stable threshold range and are spatially connected. The connectivity enhancement and structural optimization operations are performed on the ecological steady-state network to generate an ecological restoration optimization scheme that includes specific restoration engineering quantities, spatial layout and implementation sequence.

[0007] As a further aspect of the present invention, the construction of a dynamic three-dimensional ecological field characterizing the spatiotemporal evolution process of the ecological restoration area includes: The geographic space of the ecological restoration area is discretized into regular spatial grids, with each spatial grid assigned horizontal coordinates and vertical elevation, forming a three-dimensional spatial base. For each spatial grid, the vegetation index sequence and surface temperature sequence of the corresponding area of ​​the grid are extracted from the remote sensing image sequence. The soil moisture sequence, water nitrogen and phosphorus concentration sequence and pH value sequence of the grid location are matched from the continuous physicochemical parameter time series data recorded by the ground sensor network. The dominant plant species composition and soil animal abundance information of the grid area are extracted from the ecological sample data of periodic field surveys. The extracted vegetation index sequence, surface temperature sequence, soil moisture sequence, water nitrogen and phosphorus concentration sequence, pH value sequence, dominant plant species composition and soil animal abundance information are aligned and standardized according to a unified time axis to form a multi-dimensional attribute observation vector for each spatial grid at multiple time points. Spatiotemporal kriging interpolation is performed on the multidimensional attribute observation vectors of each spatial grid to generate a comprehensive ecological state attribute field that covers all spatial grids and is defined in the continuous time dimension. This comprehensive ecological state attribute field is the dynamic three-dimensional ecological field.

[0008] As a further aspect of the present invention, the improved ecological cellular automata algorithm, based on the difference in comprehensive ecological state attribute values ​​between the ecological restoration intervention measure library and adjacent grids in the dynamic three-dimensional ecological field, synchronously calculates the comprehensive ecological state attribute value of each grid at the next time step. Its working principle includes: Define a set of evolutionary rules, which are derived from a library of ecological restoration intervention measures. Each rule is associated with a specific restoration intervention measure and its corresponding comprehensive ecological state attribute value transformation pattern. Define the neighborhood range for each spatial grid in the dynamic three-dimensional ecological field, and calculate the degree of difference between the comprehensive ecological state attribute value of the grid and the mean of the comprehensive ecological state attribute values ​​of all grids in its neighborhood range; Based on the degree of difference, applicable evolutionary rules are adaptively selected from the set of evolutionary rules, and the selection probability is positively correlated with the magnitude of the degree of difference. Based on the transformation mode of the comprehensive ecological state attribute value associated with the selected evolution rule, the current comprehensive ecological state attribute value of the grid is transformed, and the result of the transformation is used as the comprehensive ecological state attribute value of the grid at the next moment. The above calculation process is executed in parallel on all grids of the dynamic three-dimensional ecological field to complete one synchronous evolution. After multiple iterations, the final generated grid attribute distribution constitutes the ecological evolution field.

[0009] As a further aspect of the present invention, extracting an ecological homeostatic network from the ecological evolution field includes: A stable threshold range for the comprehensive ecological state attribute values ​​is set, and the stable threshold range is determined based on the distribution of attribute values ​​characterizing ecosystem health in historical observation data; In the ecological evolution field that has completed iterative evolution, all spatial grids are traversed, and grids whose comprehensive ecological state attribute values ​​are within the range of the stability threshold are selected and marked as steady-state grids. Spatial clustering analysis is performed on all marked steady-state grids to merge spatially adjacent steady-state grids into the same grid cluster, generating one or more spatially independent steady-state grid clusters; Using steady-state grid clusters as nodes, if the shortest distance between two steady-state grid clusters in space is less than the connectivity threshold, then a connection edge is established between the two nodes, and all nodes and connection edges together constitute an ecological steady-state network.

[0010] As a further aspect of the present invention, connectivity enhancement and structural optimization operations are performed on the ecological steady-state network, including: Identify the connection edges between all nodes in the ecological steady-state network. If there are connection edges between nodes, plan an ecological corridor in the geographic space between the corresponding two steady-state grid clusters. The specific direction of the ecological corridor is determined based on topographic data and land use data. The global topological efficiency of the ecological steady-state network is calculated by simulating the removal of connecting edges or nodes in the network, and evaluating the contribution of each connecting edge and each node to the global topological efficiency. For connection edges whose contribution is lower than a preset threshold, assess the construction and maintenance costs of their corresponding ecological corridors. If the cost is higher than the benefit threshold, mark them as removable connection edges in the network optimization scheme. For potential connections between nodes whose contribution exceeds a preset threshold but do not currently exist, assess the cost of establishing new ecological corridors between the corresponding steady-state grid clusters and the expected increase in global topology efficiency. If the incremental benefits are significant, mark them as recommended new connections in the network optimization scheme.

[0011] As a further aspect of the present invention, the generation of the ecological restoration optimization scheme, which includes specific restoration engineering quantities, spatial layout, and implementation sequence, includes: By integrating the results of ecological corridor planning, the evaluation results of removable connecting edges and suggested new connections, a network structure optimization map is formed. The elements in the network structure optimization map are transformed into specific engineering measures, which include the species and density of vegetation planting, the earthwork volume of micro-topography modification, the excavation size of waterways connecting water bodies, and the construction specifications of artificial wetland units. Based on the interdependencies of engineering measures and seasonal construction requirements, an implementation time window is allocated for each engineering measure, forming a phased implementation sequence table; The network structure optimization map, specific engineering measures, and phased implementation schedule are integrated to form the final ecological restoration and optimization plan.

[0012] As a further aspect of the present invention, ecological sample data from periodic field surveys of the target micro-wetland ecological restoration area are obtained, including: In the ecological restoration area, fixed monitoring plots were pre-designated, and surveys were conducted during the plant growing season and the non-growing season. During the survey, the species names, abundance, cover and height of all vascular plants in the sample plot were recorded, and mixed samples of topsoil were collected for laboratory determination of organic matter content. Trapping was used to collect surface arthropod samples, which were then classified and counted. The plant data, soil measurement data, and animal data recorded in each survey are stored together with the geographical coordinates of the sample plot and the survey date to form a cycle of ecological sample data record.

[0013] As a further aspect of the present invention, the step of performing spatiotemporal kriging interpolation on the multi-dimensional attribute observation vectors of each spatial grid to generate a comprehensive ecological state attribute value field covering all spatial grids and defined in a continuous time dimension includes: The multi-dimensional attribute observation vectors of each spatial grid are reduced in dimensionality through principal component analysis to obtain the comprehensive ecological state index of each spatial grid at each time point; The comprehensive ecological state index is regarded as a sample point value in a four-dimensional domain consisting of three-dimensional space and one-dimensional time. A four-dimensional variogram model combining spatial and temporal distances is constructed to quantify the autocorrelation of the comprehensive ecological state index in the spatiotemporal domain. Using the aforementioned four-dimensional variogram model, based on the comprehensive ecological state index at known sample points, the optimal unbiased estimate of the comprehensive ecological state index at any unsampled location point in the four-dimensional domain is performed. The estimated comprehensive ecological state index is assigned back to the corresponding spatial grid and time point to form a comprehensive ecological state attribute value field that is continuously distributed in space and time.

[0014] As a further aspect of the present invention, the construction of the ecological restoration intervention measures library includes: Collect historical wetland ecological restoration project cases, and extract the restoration measures implemented in each case and the corresponding ecological and environmental parameters before and after restoration; By mapping ecological and environmental parameters to comprehensive ecological state attribute values, a mapping relationship between "restoration measures" and "changes in comprehensive ecological state attribute values" is established. Cluster analysis is performed on all mapping pairs. The K-means clustering algorithm is used to divide all mapping pairs into k clusters. The mapping pairs in each cluster have changes in the comprehensive ecological state attribute value with Euclidean distance less than the preset clustering threshold. The remediation measures that cause the Euclidean distance between the changes in the comprehensive ecological state attribute value to be less than the preset clustering threshold are grouped into the same category of rules. Define the triggering threshold and specific attribute value transformation function for each type of rule, and the collection of all rules constitutes an ecological restoration intervention measure library.

[0015] As a further aspect of the present invention, the allocation of an implementation time window for each engineering measure based on the interdependence of engineering measures and seasonal construction requirements includes: Establish a dependency graph of engineering measures. The nodes in the graph represent engineering measures, and the directed edges represent the dependencies between measures. An upstream measure must be completed before the downstream measures that depend on it begin. Identify all critical paths in the dependency graph; engineering measures on critical paths will receive the highest scheduling priority. Based on local climate conditions, species phenology, and construction prohibition periods, define the permissible seasonal time intervals for each engineering measure; Under the premise of satisfying the constraints of dependency relationships and seasonal time intervals, with the optimization goal of minimizing the total project duration or leveling resources, a heuristic scheduling algorithm is used to assign a fixed start date and end date to each engineering measure, forming a phased implementation sequence table.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: A dynamic three-dimensional ecological field is constructed using a spatial grid as the basic carrier. Remote sensing image sequences, time-series data of physical parameters, and field ecological sample data are assimilated and fused to unify the spatial scale and temporal dimension representation of multimodal data, generating comprehensive ecological state attribute values ​​with timestamps. Multi-source heterogeneous ecological data are integrated and summarized within a unified spatiotemporal framework, eliminating structural differences and dimensional fragmentation between different monitoring data. This improves the continuous representation of the ecological state of small wetland areas, expands the quantitative expression dimensions of ecological environment characteristics, and maintains the continuity and integrity of data records of ecological spatiotemporal evolution.

[0017] The operational logic of the ecological cellular automata algorithm was optimized, incorporating content from an ecological restoration intervention measures library. Evolution rules were set based on the differences in comprehensive ecological state attribute values ​​between adjacent grids, and the state evolution calculations of all grid units were completed simultaneously. The adjusted operation mode can better align with the spatial heterogeneity characteristics within wetlands, refine the interaction relationships between units, improve the deductive logic of regional ecological dynamic changes, and form an ecological evolution field with spatial distribution characteristics.

[0018] Based on the ecological evolution field, connected grid clusters with attribute values ​​within stable ranges are delineated. Connectivity enhancement and structural optimization adjustments are carried out for the ecologically stable network, and the internal combination forms of ecological space are analyzed. The spatial layout corresponding to restoration work is gradually improved, the pace of various governance tasks is standardized, the scope and implementation content of different stages are defined, and standardized restoration implementation configurations are formed to maintain the coordinated development of the internal structure of the wetland ecosystem and standardize the implementation methods of various ecological restoration tasks. Attached Figure Description

[0019] Figure 1 This is a state diagram of the optimization method for ecological restoration of small wetlands based on multimodal data as described in this invention. Figure 2 A flowchart for constructing a dynamic three-dimensional ecological field characterizing the spatiotemporal evolution of an ecological restoration area; Figure 3 A flowchart for improving the synchronous evolution calculation of the ecological cellular automata algorithm. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0021] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0022] See Figure 1 This invention provides a method for optimizing the ecological restoration of small wetlands based on multimodal data, the specific method including: Multimodal monitoring data of the target small wetland ecological restoration area is acquired. This data includes regularly collected remote sensing image sequences, continuous time-series data of physical and chemical parameters recorded by ground sensor networks, and ecological sample data from periodic field surveys. A dynamic three-dimensional ecological field characterizing the spatiotemporal evolution of the ecological restoration area is constructed. In this field, based on a spatial grid, the multimodal monitoring data is assimilated and fused into a comprehensive ecological state attribute value with timestamps. The comprehensive ecological state attribute value is input into an improved ecological cellular automata algorithm. This algorithm performs synchronous evolution calculations on the comprehensive ecological state attribute value of each grid at the next time step based on the difference between the comprehensive ecological state attribute values ​​of the ecological restoration intervention library and the adjacent grids in the dynamic three-dimensional ecological field, generating an ecological evolution field containing spatial explicit restoration schemes. An ecological steady-state network is extracted from the ecological evolution field. This network consists of spatially connected grid clusters whose comprehensive ecological state attribute values ​​are within a stable threshold range. Connectivity enhancement and structural optimization operations are performed on the ecological steady-state network to generate an ecological restoration optimization scheme that includes specific restoration engineering quantities, spatial layout, and implementation timing.

[0023] In one embodiment of the present invention, ecological sample data from periodic field surveys of the target micro-wetland ecological restoration area are obtained. Surveys are conducted in pre-defined fixed monitoring plots within the ecological restoration area during both the growing and non-growing seasons. During the surveys, the species names, abundance, cover, and height of all vascular plants within the plots are recorded. Mixed samples of surface soil are collected for laboratory determination of organic matter content. Surface arthropod samples are collected using a trapping method and then classified and counted. The plant data, soil measurement data, and animal data recorded from each survey, along with the geographical coordinates of the plot and the survey date, are stored together to form a periodic ecological sample data record. A dynamic three-dimensional ecological field characterizing the spatiotemporal evolution of the ecological restoration area is constructed. (See [reference needed]). Figure 2The geospatial area of ​​the ecological restoration zone is discretized into regular spatial grids, each with horizontal coordinates and vertical elevation, forming a three-dimensional spatial base. For each spatial grid, vegetation index sequences and land surface temperature sequences are extracted from remote sensing image sequences. Soil moisture sequences, water nitrogen and phosphorus concentration sequences, and pH sequences are matched from continuous physicochemical parameter time-series data recorded by ground sensor networks. Dominant plant species composition and soil animal abundance information for the grid area are extracted from ecological sample data from periodic field surveys. The extracted vegetation index sequences, land surface temperature sequences, soil moisture sequences, water nitrogen and phosphorus concentration sequences, pH sequences, dominant plant species composition, and soil animal abundance information are aligned and standardized according to a unified time axis to form multi-dimensional attribute observation vectors for each spatial grid at multiple time points. Spatiotemporal kriging interpolation is performed on the multi-dimensional attribute observation vectors of each spatial grid, and principal component analysis is used to reduce the dimensionality of the multi-dimensional attribute observation vectors of each spatial grid to obtain the comprehensive ecological state index of each spatial grid at each time point. The comprehensive ecological state index is considered as sample point values ​​in a four-dimensional domain consisting of three-dimensional space and one-dimensional time. A four-dimensional variogram model combining spatial and temporal distances is constructed to quantify the autocorrelation of the comprehensive ecological state index in the spatiotemporal domain. Using the four-dimensional variogram model, based on the comprehensive ecological state index at known sample points, the optimal unbiased estimate of the comprehensive ecological state index at any unsampled location point in the four-dimensional domain is performed. The estimated comprehensive ecological state index is assigned back to the corresponding spatial grid and time point, generating a comprehensive ecological state attribute value field covering all spatial grids and defined in the continuous time dimension. This comprehensive ecological state attribute value field is the dynamic three-dimensional ecological field.

[0024] In the specific implementation, taking the "Donghu Wetland Restoration Area," a degraded small wetland patch of approximately 5 hectares located in the suburbs of a city, as an example, ecological sample data from periodic field surveys of the target small wetland ecological restoration area were obtained. Within the ecological restoration area, three fixed monitoring plots were pre-set based on habitat type differences, with coordinates A (longitude X1, latitude Y1), B (longitude X2, latitude Y2), and C (longitude X3, latitude Y3). Surveys were conducted in June during the plant growing season and December during the non-growing season. During the surveys, investigators recorded the species names of all vascular plants in the plots and used visual estimation to record the abundance, cover, and average height of each species. Topsoil samples from a depth of 0-20 cm were collected using a ring cutter and sent to the laboratory for determination of soil organic matter content using the potassium dichromate external heating method. Five Parshall traps were set up in the center and four corners of the plots to continuously collect surface arthropod samples for 48 hours. Morphological classification and counting of dominant taxa were performed in the laboratory. The plant species list, abundance and height data, soil organic matter content measurement data, and surface arthropod group and quantity data recorded in each survey are stored together with the geographical coordinates of the corresponding sample plot and the survey date information in a relational database to form an independent cycle of ecological sample data record.

[0025] A dynamic three-dimensional ecological field representing the spatiotemporal evolution of the ecological restoration area was constructed. The geographical space of the "Donghu Wetland Restoration Area" was discretized into a square grid with a side length of 10 meters on the horizontal plane, forming a total of 500 spatial grids. The center point of each spatial grid was assigned a plane rectangular coordinate, and the elevation value of the corresponding point was extracted from the high-precision digital elevation model. The horizontal coordinates and vertical elevation together constitute the three-dimensional spatial base. For each spatial grid, the normalized vegetation index (NDI) and land surface temperature (LTV) sequences for the corresponding region of each grid were extracted from the Landsat-8 remote sensing image sequence (one scene per month over the past three years) based on the grid boundary range. From the continuous physicochemical parameter time series data recorded by 15 soil moisture sensors and 8 water quality multi-parameter monitoring buoys deployed in the restoration area, the soil volumetric water content sequence, total nitrogen and total phosphorus concentration sequence, and pH value sequence at the center point of each grid were obtained by matching and interpolation according to the nearest neighbor principle. From the ecological sample data of the above periodic field surveys, the dominant plant species composition and soil animal abundance information of the sample plots covered by each grid or the nearest sample plot were extracted according to spatial attribution. The dominant plant species composition is represented by the names of the top three species and their relative abundance, and the soil animal abundance is represented by the number of individuals captured per trap per day.

[0026] In the specific implementation, the vegetation index sequence, surface temperature sequence, soil moisture sequence, water nitrogen and phosphorus concentration sequence, pH value sequence, dominant plant species composition and soil animal abundance information extracted for each spatial grid will be aligned and standardized according to a unified time axis. The unified time axis is based on days. For non-daily collected data, such as monthly remote sensing data and quarterly survey data, linear interpolation will be performed between their collection days to generate daily sequences. All numerical sequence data will be mapped to the [0,1] interval using the minimum-maximum standardization method. Categorical data such as dominant plant species composition will be vectorized using one-hot encoding. Finally, a multi-dimensional attribute observation vector of each spatial grid at multiple time points in a continuous time dimension will be formed. Taking the observation vector of a typical grid G101 located in the middle of the remediation area on June 1, 2025 as an example, it may contain the following values: NDVI value 0.65, surface temperature 22.5℃, soil moisture 35.2%, total nitrogen concentration 1.2mg / L, total phosphorus concentration 0.15mg / L, pH value 6.8, and encoded plant species vector and animal abundance value 0.8.

[0027] Spatiotemporal kriging interpolation is performed on the multi-dimensional attribute observation vectors of each spatial grid. At each standardized time point, the multi-dimensional attribute observation vectors of all spatial grids within the restoration area at that time point, totaling p attributes, are organized into an n×p observation data matrix. Principal component analysis is performed on this data matrix to extract the first principal component. The score of each spatial grid on the first principal component is defined as the comprehensive ecological state index of that grid at that time point. The formula for calculating the comprehensive ecological state index is:

[0028] in: This represents the comprehensive ecological state index of the i-th spatial grid at time t. This represents the weight coefficient of the k-th standardized attribute obtained through principal component analysis. Let p represent the k-th normalized attribute value of the i-th grid at time t, and p represent the total number of attributes.

[0029] In practical implementation, the comprehensive ecological state index of each spatial grid at multiple time points is regarded as a sample point value in a four-dimensional domain consisting of three-dimensional space and one-dimensional time. The spatial dimension is the (X,Y,Z) coordinates of the grid center, and the time dimension is the observation date. A four-dimensional variogram model combining spatial Euclidean distance and time interval is constructed to quantify the autocorrelation of the comprehensive ecological state index in the spatiotemporal domain. This model can be expressed as:

[0030] in: For spatial distance, For time distance, Value of a nugget. For sill values, For spatial range, For time-varying distances, The model is a spherical function. Using the fitted four-dimensional variogram model, and based on the comprehensive ecological state index at known sample point locations, the ordinary kriging method is employed to perform optimal unbiased estimation of the comprehensive ecological state index at any unsampled location in the four-dimensional domain. The estimated comprehensive ecological state index is then assigned back to the corresponding spatial grid and time point, ultimately generating a comprehensive ecological state attribute field that covers all 500 spatial grids and is continuously defined in the time dimension. This continuous attribute field with a spatiotemporal grid structure is the dynamic three-dimensional ecological field characterizing the ecological state evolution process of the "East Lake Wetland Restoration Area".

[0031] In one embodiment of the present invention, the improved ecological cellular automata algorithm performs synchronous evolution calculations on the comprehensive ecological state attribute values ​​of each grid at the next time step based on the difference between the comprehensive ecological state attribute values ​​of the ecological restoration intervention measure library and the adjacent grids in the dynamic three-dimensional ecological field. See also Figure 3 An evolutionary rule set is defined, derived from an ecological restoration intervention library. Each rule is associated with a specific restoration intervention and its corresponding transformation mode for the comprehensive ecological state attribute value. A neighborhood range is defined for each spatial grid in the dynamic three-dimensional ecological field, and the difference between the comprehensive ecological state attribute value of that grid and the mean of the comprehensive ecological state attribute values ​​of all grids within its neighborhood is calculated. Based on this difference, an applicable evolutionary rule is adaptively selected from the rule set, with the selection probability positively correlated with the magnitude of the difference. According to the transformation mode of the comprehensive ecological state attribute value associated with the selected evolutionary rule, a transformation is applied to the current comprehensive ecological state attribute value of that grid, and the result of the transformation is used as the comprehensive ecological state attribute value of that grid at the next moment. The above calculation process is executed in parallel on all grids in the dynamic three-dimensional ecological field, completing one synchronous evolution. After multiple iterations, the final generated grid attribute distribution constitutes the ecological evolution field.

[0032] In practical implementation, the improved ecological cellular automata algorithm is used for iterative calculations on the dynamic three-dimensional ecological field of the "East Lake Wetland Restoration Area" to simulate the ecological spatial evolution process under different restoration intervention measures. An evolutionary rule set is defined, derived from a pre-built library of ecological restoration intervention measures. Each rule is associated with a specific restoration intervention measure and its corresponding comprehensive ecological state attribute value transformation mode. For example, rule R1 is associated with the restoration measure "planting submerged plants (such as water spinach)," and its corresponding comprehensive ecological state attribute value transformation mode is an incremental function that increases the current grid's comprehensive ecological state attribute value by a function related to water depth and transparency; rule R2 is associated with the restoration measure "constructing an ecological infiltration dam," and its corresponding transformation mode is a function that can improve the comprehensive ecological state attribute value and positively influence the attribute values ​​of its downstream adjacent grids; rule R3 is associated with the restoration measure "dredging," and its corresponding transformation mode is a function that rapidly improves the grid's comprehensive ecological state attribute value but may cause short-term disturbances to neighboring grids.

[0033] For each spatial grid in the dynamic three-dimensional ecological field, its neighborhood is defined. In practice, the Moore neighborhood definition is adopted, meaning that the neighborhood of a central grid is the immediate neighboring grids in the eight directions: east, south, west, north, northeast, northwest, southeast, and southwest. The difference between the comprehensive ecological state attribute value of the central grid and the mean of the comprehensive ecological state attribute values ​​of all grids within its neighborhood is calculated. The formula for calculating the difference is:

[0034] in: Indicates the degree of difference in the central grid. This represents the comprehensive ecological status attribute value of the central grid. This represents the comprehensive ecological state attribute value of the nth neighboring grid. This represents the total number of neighboring grid cells, in the Moore neighborhood. For example, if the comprehensive ecological state attribute value of the central grid G101 at time t is 0.65, and the average attribute value of its eight neighboring grids is 0.58, then the degree of difference is... .

[0035] In some embodiments, applicable evolutionary rules are adaptively selected from the set of evolutionary rules based on the calculated dissimilarity level. The selection probability is positively correlated with the magnitude of the dissimilarity level; the greater the dissimilarity level, the more likely the system is to select evolutionary rules that can cause greater attribute changes. This relationship can be understood to be achieved through a probability allocation function, for example, by setting a list of probability thresholds: when the dissimilarity level... When the probability of selecting R1 (slightly improved rule) is 80%, the probability of selecting R2 (moderately improved rule) is 20%; when When the probability of selecting R2 is 70%, the probability of selecting R3 (strength improvement rule) is 30%; when When the probability of selecting R3 is 90%, the probability of selecting R2 is 10%. By generating a random number, the interval in which the random number falls determines which evolution rule is ultimately selected for the current grid.

[0036] Based on the transformation mode of the comprehensive ecological state attribute value associated with the selected evolution rule, a transformation is applied to the current comprehensive ecological state attribute value of the grid. For example, if rule R2 is selected for grid G101, the transformation function of rule R2 is: ,in If the mean of the neighborhood attributes is true, then the comprehensive ecological state attribute value of grid G101 at time t+1 will be calculated as follows: Optionally, some rules may include more complex conditional functions or random terms to simulate uncertainties in the ecological restoration process.

[0037] The above calculation process is executed in parallel across all grids in the dynamic three-dimensional ecological field. This means that for all 500 spatial grids, rules are independently selected and attribute values ​​for the next time step are calculated based on their current attribute values, the average attribute values ​​of their neighborhoods, and the calculated differences, completing one synchronous evolution. In one iteration, the update of grid G101 may be based on the state of its neighborhood at time t, while the update of grid G102 is also based on the state of its own neighborhood at time t. The new states of all grids are calculated and take effect simultaneously. After multiple iterations, for example, 50 iterations simulating the evolution over the next five years, the resulting grid attribute distribution map after the 50th iteration constitutes the ecological evolution field. This ecological evolution field demonstrates the spatial configuration of restoration measures and the final pattern of the ecosystem state under the drive of preset rules.

[0038] In one embodiment of the present invention, an ecological steady-state network is extracted from the ecological evolution field, and a stable threshold range for the comprehensive ecological state attribute values ​​is set. This stable threshold range is determined based on the distribution of attribute values ​​characterizing ecosystem health in historical observation data. In the iteratively evolved ecological evolution field, all spatial grids are traversed, and grids whose comprehensive ecological state attribute values ​​fall within the stable threshold range are selected and marked as steady-state grids. Spatial clustering analysis is performed on all marked steady-state grids, merging spatially adjacent steady-state grids into the same grid cluster, generating one or more spatially independent steady-state grid clusters. Using steady-state grid clusters as nodes, if the closest spatial distance between two steady-state grid clusters is less than a connectivity threshold, a connection edge is established between the two nodes. All nodes and connection edges together constitute the ecological steady-state network. Connectivity enhancement and structure optimization operations are performed on the ecological steady-state network to identify connection edges between all nodes in the ecological steady-state network. If a connection edge exists between nodes, an ecological corridor is planned in the geographic space between the corresponding two steady-state grid clusters. The specific direction of the ecological corridor is determined based on topographic data and land use data. The global topological efficiency of the ecological steady-state network is calculated by simulating the removal of connecting edges or nodes from the network, and evaluating the contribution of each connecting edge and each node to the global topological efficiency. For connecting edges with a contribution below a preset threshold, the construction and maintenance costs of their corresponding ecological corridors are evaluated. If the cost exceeds the benefit threshold, they are marked as removable connecting edges in the network optimization scheme. For potential connections between nodes with a contribution above the preset threshold but not currently existing, the cost of constructing new ecological corridors between the corresponding steady-state grid clusters and the expected increase in global topological efficiency are evaluated. If the incremental benefit is significant, they are marked as recommended new connections in the network optimization scheme.

[0039] In the specific implementation, an ecological steady-state network was extracted from the ecological evolution field generated after 50 iterations of simulation in the "East Lake Wetland Restoration Area". A stable threshold range for the comprehensive ecological state attribute values ​​was set. This range was determined based on the distribution of comprehensive ecological state attribute values ​​obtained from long-term historical observations of undisturbed natural wetland patches in the surrounding area. The range between the 25th and 75th percentiles was defined as the stable threshold range, and in this implementation, it was set to [0.60, 0.90]. In the iteratively evolved ecological evolution field, all 500 spatial grids were traversed, and grids with comprehensive ecological state attribute values ​​within the stable threshold range [0.60, 0.90] were selected using a program script and marked as stable grids. After selection, a total of 320 grids were marked as stable grids.

[0040] In the specific implementation, spatial clustering analysis was performed on all marked steady-state grids. An eight-neighborhood connectivity-based clustering algorithm was used to merge spatially adjacent steady-state grids into the same grid cluster. After merging, 320 steady-state grids formed four spatially independent steady-state grid clusters. Information for each steady-state grid cluster is shown in Table 1.

[0041]

[0042] Using steady-state grid clusters as nodes, the shortest spatial distance between any two steady-state grid clusters is calculated. The shortest spatial distance is defined as the minimum Euclidean distance between all grid pairs in the two clusters. A connectivity threshold of 200 meters is set. Understandably, if the shortest spatial distance between two steady-state grid clusters is less than the connectivity threshold of 200 meters, an undirected connection edge is established between the two nodes. Calculations show that the shortest distance between nodes N1 and N2 is 150 meters, less than 200 meters, so connection edge E12 is established; the shortest distance between nodes N2 and N3 is 185 meters, so connection edge E23 is established; the shortest distance between nodes N1 and N4 is 320 meters, greater than 200 meters, so no connection is established; the shortest distance between nodes N3 and N4 is 410 meters, so no connection is established. All nodes and connection edges together constitute an ecological steady-state network containing 4 nodes and 2 connection edges.

[0043] In some embodiments, connectivity enhancement and structural optimization operations are performed on the ecological steady-state network. All connecting edges between nodes in the ecological steady-state network are identified; currently, connecting edges E12 and E23 exist in the network. Since connecting edge E12 exists between nodes N1 and N2, an ecological corridor is planned in the geographic space between the steady-state grid clusters corresponding to node N1 and node N2. The specific direction of the ecological corridor is determined based on high-precision digital terrain model data and land use type map data. The planned path must avoid existing buildings and roads and prioritize the use of existing green belts or water systems. Similarly, another ecological corridor is planned between nodes N2 and N3.

[0044] The global topological efficiency of an ecologically stable network is calculated, which is a holistic measure of the efficiency of the network's topological structure. By simulating the removal of individual edges or nodes from the network, the contribution of each edge and node to the global topological efficiency is evaluated. The contribution of edge E12 to the global topological efficiency is understandable. Through the formula:

[0045] in: It is the global topology efficiency of the original network. This represents the global topology efficiency of the network after removing the connecting edge E12. The contribution of each node is calculated similarly. The preset contribution threshold is 0.05.

[0046] In some embodiments, for connecting edges with a contribution value below a preset threshold of 0.05, the construction and maintenance costs of the corresponding ecological corridor are evaluated. The evaluation includes costs for land reclamation, vegetation establishment, and conservation facilities within the corridor area. If the cost exceeds the benefit threshold calculated based on the ecological benefit assessment model, this connecting edge is marked as a removable connecting edge in the network optimization scheme. Optionally, if the contribution value of connecting edge E12... If the calculated value is 0.03, which is lower than 0.05, and the estimated cost of its planned ecological corridor is 500,000 yuan, which is higher than the benefit threshold of 400,000 yuan, then the connecting edge E12 is marked as a removable connecting edge.

[0047] For potential connections between nodes whose contribution exceeds a preset threshold but do not currently exist (e.g., there is no connection between nodes N1 and N4), simulations show that adding a virtual connection between them can significantly improve the global topology efficiency of the network. The value is 0.08, which is higher than the critical value of 0.05. Further evaluation is needed on the cost and expected increase in global topology efficiency of constructing a new ecological corridor between the steady-state grid clusters corresponding to nodes N1 and N4. The estimated cost of constructing the new corridor is 600,000 yuan, and the expected efficiency increase translates to an ecological benefit of 800,000 yuan. Given the significant incremental benefit, the connection between nodes N1 and N4 is marked as a recommended new connection in the network optimization scheme.

[0048] In one embodiment of the present invention, an ecological restoration optimization scheme is generated, including specific restoration engineering quantities, spatial layout, and implementation sequence. This scheme integrates ecological corridor planning results, evaluation results of removable connecting edges, and suggested new connections to form a network structure optimization map. The elements in the network structure optimization map are transformed into specific engineering measures, including the species and density of vegetation planting, the earthwork volume for micro-topography modification, the excavation dimensions of waterways connecting water bodies, and the construction specifications of artificial wetland units. Based on the interdependencies of the engineering measures and seasonal construction requirements, appropriate implementation time windows are allocated to each engineering measure, forming a phased implementation sequence table. The network structure optimization map, the specific engineering measures, and the phased implementation sequence table are integrated to form the final ecological restoration optimization scheme.

[0049] In practice, an ecological restoration optimization plan is generated, which includes specific restoration work volume, spatial layout, and implementation sequence. The plan integrates the ecological corridor planning results, the evaluation results of removable connecting edges and suggested new connections to form a network structure optimization map. The network structure optimization map is presented in the form of a geographic information system layer, which includes a node layer and a connecting edge layer. The node layer marks the spatial location and attributes of the four steady-state grid cluster nodes N1, N2, N3, and N4. The connecting edge layer includes the planned ecological corridor spatial path corresponding to the planned retainable connecting edge E23, the original path marked as the removable connecting edge E12, and the proposed corridor spatial path between nodes N1 and N4 marked as suggested new connections. The network structure optimization map also includes evaluation attribute fields corresponding to each connecting edge, such as "type" (planned retainable, removable, suggested new), "contribution", "cost estimate", and "expected benefits".

[0050] In some embodiments, the elements in the network structure optimization graph are transformed into specific engineering measures, which are designed based on the type and spatial location of the connecting edges. For the ecological corridor corresponding to the planned retained connecting edge E23, the specific engineering measures include planting wetland vegetation within the corridor area, selecting reeds and cattails as species, setting the planting density at 4 clumps per square meter, carrying out micro-topographical modifications in the low-lying areas of the corridor to create small water purification ponds, with an estimated earthwork excavation volume of 150 cubic meters, and excavating an ecological ditch 0.8 meters wide, 0.5 meters deep, and 30 meters long at the junction of the corridor and the existing water system to achieve water connectivity. For the proposed connection between nodes N1 and N4, the engineering measures include constructing an artificial wetland unit with a surface area of ​​200 square meters and a depth of 0.6-1.2 meters. The construction specifications of the artificial wetland unit include a filler layer structure (from top to bottom: 0.2 meters of planting soil, 0.3 meters of gravel, and 0.1 meters of pebbles), a water distribution system design, and specific aquatic plant configurations. A list of specific engineering measures is provided in Table 2.

[0051]

[0052] Based on the interdependencies of the engineering measures and seasonal construction requirements, appropriate implementation time windows are allocated for each engineering measure. It is understood that engineering measure M2 (micro-topography modification) is the foundation for engineering measures M1 (vegetation planting) and M3 (waterway excavation) and must be completed before engineering measures M1 and M3; engineering measure M4 (artificial wetland construction) is an independent project. Taking into account local climate conditions, species phenological periods, and construction restrictions, the permissible seasonal time intervals for each engineering measure are defined. Vegetation planting for engineering measure M1 is preferably carried out in spring (March to May) or autumn (September to October); earthwork for engineering measures M2 and M3 should avoid the rainy season (June to August); and artificial wetland construction for engineering measure M4 is preferably carried out during the dry season (November to February of the following year). Under the premise of satisfying dependency and seasonal time interval constraints, and with the shortest total project duration as the optimization objective, a heuristic scheduling algorithm is used to assign a defined start date and end date to each engineering measure. The scheduling algorithm must ensure that the actual construction window of each engineering measure falls entirely within its allowed seasonal time interval, and that all prerequisite dependencies are satisfied. Optionally, the start time of an engineering measure... It can be done through the formula:

[0053] in: This represents the set of latest completion dates for all its immediate predecessor engineering measures. This means taking the latest date among them. This indicates the necessary process intervals; for example, the start time of engineering measure M1 must be at least 7 days after the completion of engineering measure M2. The final result is a phased implementation sequence table, which clearly specifies the planned start and end dates for each engineering measure corresponding to each project number.

[0054] The network structure optimization map, the summary table of specific engineering measures, and the phased implementation sequence table are integrated to form the final ecological restoration optimization plan document. The document is presented in a graphic and textual format. The first part is a general map of the restoration spatial layout based on geographic information, the second part is a detailed quantitative list of engineering measures, and the third part is a Gantt chart of the phased implementation plan. This ecological restoration optimization plan is directly used to guide the preparation of construction budgets and on-site operations.

[0055] In one embodiment of the present invention, the construction of the ecological restoration intervention measure library includes collecting historical wetland ecological restoration project cases, extracting the restoration measures implemented in each case and their corresponding pre- and post-restoration ecological environment parameters. The ecological environment parameters are mapped to comprehensive ecological state attribute values, thereby establishing a mapping relationship between "restoration measures" and "changes in comprehensive ecological state attribute values." Cluster analysis is performed on all mapping relationship pairs, grouping restoration measures that lead to similar changes in comprehensive ecological state attribute values ​​into the same rule category. For each rule category, a triggering condition threshold and a specific attribute value transformation function are defined, and the set of all rules constitutes the ecological restoration intervention measure library. When allocating a suitable implementation time window for each engineering measure based on the interdependence of engineering measures and seasonal construction requirements, an engineering measure dependency graph is established. Nodes in the graph represent engineering measures, and directed edges represent the dependencies between measures. Upstream measures must be completed before the downstream measures that depend on them begin. All critical paths in the dependency graph are identified, and engineering measures on critical paths receive the highest scheduling priority. Combining local climate conditions, species phenological periods, and construction taboo periods, a seasonal time interval for the permitted implementation of each engineering measure is defined. Under the premise of satisfying the constraints of dependency relationships and seasonal time intervals, with the optimization goal of minimizing the total project duration or leveling resources, a heuristic scheduling algorithm is used to assign a fixed start date and end date to each engineering measure, forming a phased implementation sequence table.

[0056] In its implementation, the construction of the ecological restoration intervention measures database involves collecting historical wetland ecological restoration project cases. These cases come from multiple completed small-scale wetland restoration project reports, engineering case records in academic literature, and historical project archives from relevant ecological management departments. The restoration measures implemented in each case document, along with their corresponding pre- and post-restoration ecological and environmental parameters, are extracted. Restoration measures are recorded in a structured list format, for example, "Measure type: vegetation restoration; Specific content: planting reeds, density 6 clumps / square meter; Implementation area: nearshore shallow water area." Ecological and environmental parameters correspond to observation data from one year before the restoration project and during the stabilization period (usually the third year). This data includes water transparency, dissolved oxygen, total phosphorus, total nitrogen, chlorophyll a concentration, emergent plant cover, and zooplankton density. In practice, multiple sets of pre- and post-restoration ecological and environmental parameters for each case are input into a pre-trained comprehensive evaluation model. This model maps multidimensional ecological and environmental parameters into a single scalar value, namely, a comprehensive ecological state attribute value. Through this process, a clear mapping relationship was established for each historical case. The mapping relationship clearly records the combination of restoration measures adopted and the corresponding change in the comprehensive ecological state attribute value ΔV. For example, a mapping relationship is recorded as "Measure combination {planting reeds, releasing benthic shellfish} -> comprehensive ecological state attribute value increased from 0.42 to 0.71, ΔV=+0.29".

[0057] Cluster analysis is performed on all mapping pairs, primarily based on the change in the comprehensive ecological state attribute value ΔV within each pair, supplemented by a similarity metric derived from vectorized encoding of the technical types of the combined measures. The K-means clustering algorithm divides all mapping pairs into k clusters. Each cluster contains mapping pairs with similar combinations of remediation measures resulting from changes in the comprehensive ecological state attribute value ΔV, thus grouping remediation measures leading to similar changes in the comprehensive ecological state attribute value into the same category. Specifically, a preset clustering threshold is set, determined based on the distribution range of changes in historical data, to constrain the maximum Euclidean distance between data points within a cluster, ensuring high similarity in changes within the cluster. During clustering, k cluster centers are randomly initialized. The Euclidean distance from each change vector to each cluster center is calculated, and the vector is assigned to the nearest cluster. The center point of each cluster is then recalculated as the mean of all vectors within that cluster. This process is iterated until the cluster centers no longer change significantly or the maximum number of iterations is reached, resulting in k stable clusters. Based on the final clustering results, mapping pairs belonging to the same cluster are grouped together. The restoration measures in these mapping pairs are grouped into the same category because they cause the Euclidean distance between the changes in the comprehensive ecological state attribute values ​​to be less than the preset clustering threshold, i.e., they produce similar ecological effects.

[0058] For each rule category, a triggering threshold and a specific attribute value transformation function are defined. The triggering threshold is typically related to the current grid's overall ecological state attribute value level and its difference from the neighborhood mean. The attribute value transformation function quantifies the expected change in the overall ecological state attribute value after applying this type of rule. For example, for a rule cluster classified as "moderate nutrient reduction and habitat improvement," the triggering condition is defined as: the current grid's overall ecological state attribute value... The attribute value is located in the interval [0.40, 0.60] and its difference from the mean of its eight neighbors is greater than 0.05; its attribute value transformation function is defined as:

[0059] in: This indicates the attribute value at the next moment. Indicates the current attribute value. This represents the theoretically achievable maximum local attribute value reference value for this type of rule. It is the effect coefficient. It is the process rate constant. This refers to the time step in which the rule takes effect. All defined and parameterized rules are stored in a queryable and scalable digital knowledge base, which constitutes the ecological restoration intervention measure library, providing a set of evolutionary rules for the improved ecological cellular automata algorithm.

[0060] In practice, based on the interdependencies of engineering measures and seasonal construction requirements, a suitable implementation time window is allocated to each engineering measure. An engineering measure dependency graph is established, represented as a directed graph. Nodes in the graph represent specific engineering measures; for example, node M1 represents "vegetation planting," and node M2 ​​represents "micro-topography modification." Directed edges represent the dependencies between measures. A directed edge from node M2 ​​to node M1 indicates that engineering measure M2 must be completed before its dependent engineering measure M1 begins; that is, micro-topography modification work must be completed before vegetation planting. All critical paths in the dependency graph are identified. The critical path is the longest path from the project start node to the project end node. Any delay in the start or end time of an engineering measure on the critical path will directly lead to a delay in the overall project duration; therefore, engineering measures on the critical path will receive the highest scheduling priority. Taking into account local climate conditions, species phenology, and construction restrictions, the permitted seasonal time intervals for each engineering measure are defined. For example, the permitted seasonal time interval for the engineering measure "planting aquatic plants" is set from March 15 to May 15 in spring, and the permitted seasonal time interval for the engineering measure "earthwork excavation and terrain shaping" is set from October 1 to February 28 of the following year in autumn and winter (avoiding the rainy season and freezing period).

[0061] Under the premise of satisfying dependency and seasonal time interval constraints, and with the shortest total project duration as the optimization objective, a heuristic scheduling algorithm is used to assign a fixed start and end date to each engineering measure. The heuristic scheduling algorithm iteratively calculates the earliest possible start date for each engineering measure, starting from the project start date. Understandably, for an engineering measure... Its earliest possible start date The calculation requires that all its immediate predecessor works have been completed, and that the date falls within the period of the works. Its own permitted seasonal time intervals Inside. The calculation formula is:

[0062] in: Indicates engineering measures The set of all immediate engineering measures, Indicates immediate engineering measures The earliest completion date, Indicates engineering measures The start date of the permitted seasonal time interval for implementation. Engineering measures. Earliest completion date Then through calculate, Engineering measures The estimated construction period in days. During the calculation process, if the calculated... Exceeded Then the engineering measures need to be implemented. The schedule is postponed to the start date of the next permitted seasonal time window. In some embodiments, when multiple non-critical path engineering measures compete for construction resources in the same time period, the algorithm will sort and adjust the measures according to their free float, prioritizing those with smaller free floats. Finally, the algorithm outputs a specific planned start date and planned end date for each engineering measure, and the planned times of all engineering measures are summarized to form a phased implementation sequence table.

[0063] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A multi-modal data-based small and micro wetland ecological restoration optimization method, characterized in that, include: Acquire multimodal monitoring data of the target small wetland ecological restoration area. The multimodal monitoring data includes regularly collected remote sensing image sequences, continuous time-series data of physical and chemical parameters recorded by ground sensor networks, and ecological sample data from periodic field surveys. A dynamic three-dimensional ecological field is constructed to characterize the spatiotemporal evolution of the ecological restoration area. In the dynamic three-dimensional ecological field, multimodal monitoring data is assimilated and fused into a comprehensive ecological state attribute value with a timestamp, based on a spatial grid. The comprehensive ecological state attribute value is input into the improved ecological cellular automaton algorithm. The improved ecological cellular automaton algorithm performs synchronous evolution calculation on the comprehensive ecological state attribute value of each grid at the next moment based on the difference between the comprehensive ecological state attribute values ​​of the ecological restoration intervention measures library and the adjacent grids in the dynamic three-dimensional ecological field, and generates an ecological evolution field containing spatial explicit restoration schemes. An ecological steady-state network is extracted from the ecological evolution field. The ecological steady-state network consists of a grid cluster whose comprehensive ecological state attribute values ​​are within a stable threshold range and are spatially connected. The connectivity enhancement and structural optimization operations are performed on the ecological steady-state network to generate an ecological restoration optimization scheme that includes specific restoration engineering quantities, spatial layout and implementation sequence.

2. The method of claim 1, wherein, The construction of a dynamic three-dimensional ecological field characterizing the spatiotemporal evolution of the ecological restoration area includes: The geographic space of the ecological restoration area is discretized into regular spatial grids, with each spatial grid assigned horizontal coordinates and vertical elevation, forming a three-dimensional spatial base. For each spatial grid, the vegetation index sequence and surface temperature sequence of the corresponding area of ​​the grid are extracted from the remote sensing image sequence. The soil moisture sequence, water nitrogen and phosphorus concentration sequence and pH value sequence of the grid location are matched from the continuous physicochemical parameter time series data recorded by the ground sensor network. The dominant plant species composition and soil animal abundance information of the grid area are extracted from the ecological sample data of periodic field surveys. The extracted vegetation index sequence, surface temperature sequence, soil moisture sequence, water nitrogen and phosphorus concentration sequence, pH value sequence, dominant plant species composition and soil animal abundance information are aligned and standardized according to a unified time axis to form a multi-dimensional attribute observation vector for each spatial grid at multiple time points. Spatiotemporal kriging interpolation is performed on the multidimensional attribute observation vectors of each spatial grid to generate a comprehensive ecological state attribute field that covers all spatial grids and is defined in the continuous time dimension. This comprehensive ecological state attribute field is the dynamic three-dimensional ecological field.

3. The method of claim 1, wherein, The improved ecological cellular automata algorithm is based on the difference in comprehensive ecological state attribute values ​​between the ecological restoration intervention measure library and adjacent grids in the dynamic three-dimensional ecological field. It performs synchronous evolution calculation on the comprehensive ecological state attribute value of each grid at the next time step. Its working principle includes: Define a set of evolutionary rules, which are derived from a library of ecological restoration intervention measures. Each rule is associated with a specific restoration intervention measure and its corresponding comprehensive ecological state attribute value transformation pattern. Define the neighborhood range for each spatial grid in the dynamic three-dimensional ecological field, and calculate the degree of difference between the comprehensive ecological state attribute value of the grid and the mean of the comprehensive ecological state attribute values ​​of all grids in its neighborhood range; Based on the degree of difference, applicable evolutionary rules are adaptively selected from the set of evolutionary rules, and the selection probability is positively correlated with the magnitude of the degree of difference. Based on the transformation mode of the comprehensive ecological state attribute value associated with the selected evolution rule, the current comprehensive ecological state attribute value of the grid is transformed, and the result of the transformation is used as the comprehensive ecological state attribute value of the grid at the next moment. The above calculation process is executed in parallel on all grids of the dynamic three-dimensional ecological field to complete one synchronous evolution. After multiple iterations, the final generated grid attribute distribution constitutes the ecological evolution field.

4. The method of claim 3, wherein, Extracting the ecological homeostasis network from the aforementioned ecological evolution field includes: A stable threshold range for the comprehensive ecological state attribute values ​​is set, and the stable threshold range is determined based on the distribution of attribute values ​​characterizing ecosystem health in historical observation data; In the ecological evolution field that has completed iterative evolution, all spatial grids are traversed, and grids whose comprehensive ecological state attribute values ​​are within the range of the stability threshold are selected and marked as steady-state grids. Spatial clustering analysis is performed on all marked steady-state grids to merge spatially adjacent steady-state grids into the same grid cluster, generating one or more spatially independent steady-state grid clusters; Using steady-state grid clusters as nodes, if the shortest distance between two steady-state grid clusters in space is less than the connectivity threshold, then a connection edge is established between the two nodes, and all nodes and connection edges together constitute an ecological steady-state network.

5. The method of claim 4, wherein, Perform connectivity enhancement and structure optimization operations on the aforementioned ecological steady-state network, including: Identify the connection edges between all nodes in the ecological steady-state network. If there are connection edges between nodes, plan an ecological corridor in the geographic space between the corresponding two steady-state grid clusters. The specific direction of the ecological corridor is determined based on topographic data and land use data. The global topological efficiency of the ecological steady-state network is calculated by simulating the removal of connecting edges or nodes in the network, and evaluating the contribution of each connecting edge and each node to the global topological efficiency. For connection edges whose contribution is lower than a preset threshold, assess the construction and maintenance costs of their corresponding ecological corridors. If the cost is higher than the benefit threshold, mark them as removable connection edges in the network optimization scheme. For potential connections between nodes whose contribution exceeds a preset threshold but do not currently exist, assess the cost of creating new ecological corridors between the corresponding steady-state grid clusters and the expected increase in global topology efficiency. If the incremental benefits are significant, mark them as recommended new connections in the network optimization scheme.

6. The method for optimizing the ecological restoration of small wetlands based on multimodal data according to claim 5, characterized in that, The generated ecological restoration optimization plan, which includes specific restoration work quantities, spatial layout, and implementation sequence, includes: By integrating the results of ecological corridor planning, the evaluation results of removable connecting edges and suggested new connections, a network structure optimization map is formed. The elements in the network structure optimization map are transformed into specific engineering measures, which include the species and density of vegetation planting, the earthwork volume of micro-topography modification, the excavation size of waterways connecting water bodies, and the construction specifications of artificial wetland units. Based on the interdependencies of engineering measures and seasonal construction requirements, an implementation time window is allocated for each engineering measure, forming a phased implementation sequence table; The network structure optimization map, specific engineering measures, and phased implementation schedule are integrated to form the final ecological restoration and optimization plan.

7. The method for optimizing the ecological restoration of small wetlands based on multimodal data according to claim 1, characterized in that, Obtain ecological sample data from periodic field surveys of the target small wetland ecological restoration area, including: In the ecological restoration area, fixed monitoring plots were pre-designated, and surveys were conducted during the plant growing season and the non-growing season. During the survey, the species names, abundance, cover and height of all vascular plants in the sample plot were recorded, and mixed samples of topsoil were collected for laboratory determination of organic matter content. Trapping was used to collect surface arthropod samples, which were then classified and counted. The plant data, soil measurement data, and animal data recorded in each survey are stored together with the geographical coordinates of the sample plot and the survey date to form a cycle of ecological sample data record.

8. The method for optimizing the ecological restoration of small wetlands based on multimodal data according to claim 2, characterized in that, The process of performing spatiotemporal kriging interpolation on the multi-dimensional attribute observation vectors of each spatial grid to generate a comprehensive ecological state attribute field covering all spatial grids and defined in a continuous time dimension includes: The multi-dimensional attribute observation vectors of each spatial grid are reduced in dimensionality through principal component analysis to obtain the comprehensive ecological state index of each spatial grid at each time point; The comprehensive ecological state index is regarded as a sample point value in a four-dimensional domain consisting of three-dimensional space and one-dimensional time. A four-dimensional variogram model combining spatial and temporal distances is constructed to quantify the autocorrelation of the comprehensive ecological state index in the spatiotemporal domain. Using the aforementioned four-dimensional variogram model, based on the comprehensive ecological state index at known sample points, the optimal unbiased estimate of the comprehensive ecological state index at any unsampled location point in the four-dimensional domain is performed. The estimated comprehensive ecological state index is assigned back to the corresponding spatial grid and time point to form a comprehensive ecological state attribute value field that is continuously distributed in space and time.

9. The method for optimizing the ecological restoration of small wetlands based on multimodal data according to claim 1, characterized in that, The construction of the ecological restoration intervention measures library includes: Collect historical wetland ecological restoration project cases, and extract the restoration measures implemented in each case and the corresponding ecological and environmental parameters before and after restoration; By mapping ecological and environmental parameters to comprehensive ecological state attribute values, a mapping relationship between "restoration measures" and "changes in comprehensive ecological state attribute values" is established. Cluster analysis is performed on all mapping pairs. The K-means clustering algorithm is used to divide all mapping pairs into k clusters. The mapping pairs in each cluster have changes in the comprehensive ecological state attribute value with Euclidean distance less than the preset clustering threshold. The remediation measures that cause the Euclidean distance between the changes in the comprehensive ecological state attribute value to be less than the preset clustering threshold are grouped into the same category of rules. Define the triggering threshold and specific attribute value transformation function for each type of rule, and the collection of all rules constitutes an ecological restoration intervention measure library.

10. The method for optimizing the ecological restoration of small wetlands based on multimodal data according to claim 6, characterized in that, Based on the interdependencies of engineering measures and seasonal construction requirements, an implementation time window is allocated to each engineering measure, including: Establish a dependency graph of engineering measures. The nodes in the graph represent engineering measures, and the directed edges represent the dependencies between measures. An upstream measure must be completed before the downstream measures that depend on it begin. Identify all critical paths in the dependency graph; engineering measures on critical paths will receive the highest scheduling priority. Based on local climate conditions, species phenology, and construction prohibition periods, define the permissible seasonal time intervals for each engineering measure; Under the premise of satisfying the constraints of dependency relationships and seasonal time intervals, with the optimization goal of minimizing the total project duration or leveling resources, a heuristic scheduling algorithm is used to assign a fixed start date and end date to each engineering measure, forming a phased implementation sequence table.

Images