Methods and systems for analyzing population distribution in mangrove ecosystems

Abstract

This application relates to the field of data processing technology and discloses an analysis method and system for mangrove ecosystem population distribution. The method includes: acquiring high-resolution remote sensing images and field data; establishing a multidimensional database using image enhancement algorithms; identifying community types using random forest algorithms; extracting plant location information through deep learning; calculating indicators such as relative abundance to obtain characteristic parameters; using landscape index analysis to obtain key patches; establishing a prediction model based on environmental parameters; and constructing an evaluation model based on hierarchical analysis to generate management decision recommendations. This application realizes an analytical method for predicting mangrove population distribution and making intelligent management decisions, thereby improving the scientific rigor of mangrove ecosystem research and the effectiveness of management.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to an analytical method and system for the population distribution of mangrove ecosystems. Background Technology

[0002] Mangroves are a type of woody plant community distributed in the intertidal zone of tropical and subtropical coasts, playing a vital role in maintaining coastal ecosystem stability and protecting coastal biodiversity. Existing research on mangrove ecosystems mainly focuses on two aspects: remote sensing monitoring and quadrat surveys. Remote sensing monitoring uses satellite image interpretation to obtain spatial distribution information of mangroves, while quadrat surveys collect data on species composition, growth status, and other information through field investigations. Currently commonly used research methods include using remote sensing to monitor changes in mangrove area and landscape, studying community structure and species diversity through quadrat surveys, and assessing ecosystem health based on ecological indicators.

[0003] However, existing research methods have the following shortcomings: First, remote sensing monitoring and quadrat surveys are often conducted separately, lacking an effective data integration mechanism, which prevents the full utilization of the advantages of multi-source data; second, in the data analysis process, most methods rely on single statistical methods or ecological indicators, making it difficult to comprehensively reflect the complexity of mangrove ecosystems; third, existing assessment methods are mostly at the level of descriptive analysis, lacking in-depth exploration and prediction capabilities of mangrove population distribution patterns; and finally, existing management decisions are mostly based on experience-based judgments, lacking systematic and intelligent decision support tools. Summary of the Invention

[0004] This application provides an analytical method and system for mangrove ecosystem population distribution, which enables the prediction of mangrove population distribution and intelligent management decision-making, thereby improving the scientific nature of mangrove ecosystem research and the effectiveness of management.

[0005] Firstly, this application provides a method for analyzing the population distribution of mangrove ecosystems. The method includes: acquiring high-resolution remote sensing images using a remote sensing image acquisition device, collecting field survey data and environmental parameters within the study area, preprocessing the high-resolution remote sensing images using an image enhancement algorithm to establish a multidimensional database containing spatial information, species information, and environmental parameters; identifying community types in the multidimensional database using a random forest classification algorithm; extracting the location information of individual mangrove plants from the high-resolution remote sensing images using a deep learning object detection algorithm to generate a community distribution vector data layer; and calculating the relative abundance, relative frequency, and relative significance of mangrove plants in the woody layer based on the community distribution vector data layer to obtain importance indexes, and further analyzing the herbaceous layer of mangrove plants... The relative cover and relative height of the species are calculated to obtain the overall dominance index. Spatial statistical algorithms are used to calculate the spatial distribution characteristic parameters of the population. Based on the spatial distribution characteristic parameters, a landscape pattern analysis is performed using a landscape index calculation model to obtain the maximum patch index, clumpiness index, similarity adjacency percentage, and patch aggregation index. A connectivity analysis model is established based on graph theory algorithms to obtain key patch identification results. The key patch identification results are correlated with environmental parameters, and an environmental factor influence model is established using a gradient boosting algorithm. Uncertainty analysis is performed using a Bayesian network algorithm to obtain population distribution prediction results. Based on the population distribution prediction results, combined with the importance index, overall dominance index, and landscape index, a health assessment model is constructed using the analytic hierarchy process (AHP), a knowledge graph reasoning system is established, and management decision-making suggestions for mangrove ecosystems are generated.

[0006] Secondly, this application provides an analysis system for the population distribution of mangrove ecosystems, the analysis system for the population distribution of mangrove ecosystems comprising:

[0007] The acquisition module is used to acquire high-resolution remote sensing images through a remote sensing image acquisition device, collect field survey data and environmental parameters in the study area, preprocess the high-resolution remote sensing images through image enhancement algorithms, and establish a multidimensional database containing spatial information, species information and environmental parameters.

[0008] The identification module is used to identify community types in the data in the multidimensional database using a random forest classification algorithm, and to extract the location information of individual mangrove plants from the high-resolution remote sensing image using a deep learning target detection algorithm, thereby generating a community distribution vector data layer.

[0009] The calculation module is used to calculate the relative abundance, relative frequency, and relative significance of mangrove plants in the woody layer based on the community distribution vector data layer, to obtain the importance value index; to calculate the relative cover and relative height of mangrove plants in the herbaceous layer, to obtain the total dominance index; and to calculate the spatial distribution characteristic parameters of the population through spatial statistical algorithms.

[0010] The analysis module is used to perform landscape pattern analysis using a landscape index calculation model based on the spatial distribution characteristic parameters of the population, to obtain the maximum patch index, clustering index, similarity adjacency percentage and patch aggregation index, and to establish a connectivity analysis model based on graph theory algorithm to obtain key patch identification results.

[0011] The correlation module is used to correlate the key patch identification results with environmental parameters, establish an environmental factor influence model through gradient boosting algorithm, perform uncertainty analysis using Bayesian network algorithm, and obtain population distribution prediction results.

[0012] The generation module is used to construct a health assessment model based on the population distribution prediction results, combined with the importance value index, total dominance index and landscape index, through the hierarchical analysis method, establish a knowledge graph reasoning system, and generate management decision-making suggestions for mangrove ecosystems.

[0013] The technical solution provided in this application acquires high-resolution remote sensing images through a remote sensing image acquisition device, and combines this with field survey data and environmental parameter collection to achieve effective fusion of multi-source data, improving data integrity and accuracy. Image enhancement algorithms are used to preprocess the remote sensing images, establishing a multi-dimensional database containing spatial information, species information, and environmental parameters, providing comprehensive data support for subsequent analysis. Random forest classification algorithms and deep learning object detection algorithms are used to process the data, achieving not only accurate identification of community types but also extraction of location information of individual mangrove plants, significantly improving the accuracy of spatial distribution analysis. By calculating various indicators of mangrove plants in the woody and herbaceous layers and combining them with spatial statistical algorithms, quantitative expression of population distribution characteristics is achieved. A landscape index calculation model is used for landscape pattern analysis, and a connectivity analysis model is established using graph theory algorithms, making the identification of key patches more accurate and objective. Gradient boosting algorithms and Bayesian network algorithms are used for environmental factor impact analysis, significantly improving the accuracy of population distribution prediction. Finally, through hierarchical analysis and a knowledge graph reasoning system, intelligent and scientific decision-making for mangrove ecosystem management is achieved. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of an embodiment of the method for analyzing population distribution in mangrove ecosystems in this application.

[0016] Figure 2 This is a schematic diagram of an embodiment of the analysis system for mangrove ecosystem population distribution in this application. Detailed Implementation

[0017] This application provides a method and system for analyzing the population distribution of mangrove ecosystems. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the method for analyzing population distribution in mangrove ecosystems in this application includes:

[0019] Step S101: Acquire high-resolution remote sensing images through a remote sensing image acquisition device, collect field survey data and environmental parameters in the study area, preprocess the high-resolution remote sensing images through image enhancement algorithms, and establish a multidimensional database containing spatial information, species information and environmental parameters.

[0020] Step S102: Use the random forest classification algorithm to identify the community type of the data in the multidimensional database, and use the deep learning target detection algorithm to extract the location information of individual mangrove plants from the high-resolution remote sensing image to generate a community distribution vector data layer.

[0021] Step S103: Based on the community distribution vector data layer, calculate the relative abundance, relative frequency and relative significance of mangrove plants in the woody layer to obtain the importance value index; calculate the relative cover and relative height of mangrove plants in the herbaceous layer to obtain the total dominance index; and calculate the spatial distribution characteristic parameters of the population through spatial statistical algorithms.

[0022] Step S104: Based on the spatial distribution characteristic parameters of the population, the landscape pattern analysis is carried out using the landscape index calculation model to obtain the maximum patch index, clustering index, similar adjacency percentage and patch aggregation index. A connectivity analysis model is established based on graph theory algorithm to obtain the key patch identification results.

[0023] Step S105: Correlate the key patch identification results with environmental parameters, establish an environmental factor influence model through gradient boosting algorithm, perform uncertainty analysis using Bayesian network algorithm, and obtain population distribution prediction results.

[0024] Step S106: Based on the population distribution prediction results, and combined with the importance value index, total dominance index and landscape index, construct a health assessment model through the hierarchical analysis method, establish a knowledge graph reasoning system, and generate management decision-making suggestions for mangrove ecosystems.

[0025] It is understood that the executing entity of this application can be an analysis system for mangrove ecosystem population distribution, or it can be a terminal or a server; no specific limitation is made here. This application's embodiment uses a server as an example for illustration.

[0026] Specifically, high-resolution remote sensing images were acquired using a remote sensing image acquisition device. The acquisition device primarily employed the GeoEye-1 high-resolution satellite to perform multispectral imaging of the study area, acquiring remote sensing data encompassing both visible and near-infrared bands. Simultaneously, multiple transects were established within the study area, with quadrats evenly distributed across each transect. Quadrats included two sizes: 10m×10m and 20m×5m. Within each quadrat, data on mangrove tree height, diameter at breast height (DBH), crown width, species type, quantity, and cover were recorded as field survey data. Environmental parameters, including tidal level, salinity, and soil pH, were also collected. The acquired high-resolution remote sensing images underwent preprocessing including geometric correction, radiometric calibration, atmospheric correction, and image fusion to construct a multidimensional database containing spatial information, species information, and environmental parameters. Subsequently, a random forest classification algorithm was used to identify community types from the data in the multidimensional database. The random forest algorithm classifies data by constructing multiple decision trees, each trained using a randomly selected subset of features. The final classification result was determined through a voting process. In the specific processing, spectral and texture features are used as input variables. A classification model is established using training samples to identify different mangrove community types. Simultaneously, a deep learning object detection algorithm is employed to extract the location information of individual mangrove plants from high-resolution remote sensing images. This algorithm uses a convolutional neural network structure, extracting image features through multi-layer convolution and pooling operations, and finally locating individual plants using bounding box regression. The community type identification results are integrated with the individual plant location information to generate a community distribution vector data layer.

[0027] Based on the generated community distribution vector data layer, population dominance analysis was performed on mangrove plants in the woody layer. Relative abundance (the percentage of individuals of a certain plant species out of the total number of individuals of all species), relative frequency (the percentage of the frequency of a certain plant species appearing in each quadrat out of the sum of the frequencies of all species), and relative significance (the percentage of the basal area at breast height of a certain plant out of the sum of the basal areas of all species) were calculated, and the sum of these three values ​​yielded the importance index. For mangrove plants in the herbaceous layer, relative cover (the ratio of the area covered by a certain plant species to the total area) and relative height (the ratio of the height of a certain plant species to the average height of all species) were calculated to obtain the overall dominance index. Spatial statistical algorithms were used to calculate spatial distribution characteristic parameters of the population, including the variance / mean ratio, negative binomial distribution parameter, clustering index, and average crowding index, quantifying the spatial distribution pattern of the population. Based on the calculated spatial distribution characteristic parameters of the population, a landscape pattern analysis was performed using a landscape index calculation model. First, the largest patch index is calculated, which is the proportion of the largest patch area to the total area. Second, the clustering index is calculated, reflecting the degree of aggregation of patches of the same type. Then, the similarity adjacency percentage is calculated, representing the degree of adjacency of patches of the same type. Finally, the patch aggregation degree index is calculated, characterizing the aggregation status of landscape types. A connectivity analysis model is established based on graph theory algorithms, treating mangrove patches as nodes and the connections between patches as edges. Key patches are identified by calculating indicators such as node degree and centrality.

[0028] The identified key patches were correlated with environmental parameters, and an environmental factor impact model was established using the gradient boosting algorithm. The gradient boosting algorithm iteratively constructs multiple weak learners, training on the residuals of the previous prediction in each iteration, ultimately combining the results of all weak learners to obtain a strong learner. During modeling, environmental parameters were used as feature variables, and mangrove distribution characteristics as target variables, training to obtain the influence relationship between environmental factors and mangrove distribution. Uncertainty analysis was performed using a Bayesian network algorithm. The Bayesian network represents the conditional dependencies between variables using a directed acyclic graph, calculating the conditional probability distribution of each node, and finally obtaining the population distribution prediction results. Based on the prediction results, a health assessment model was constructed using the analytic hierarchy process (AHP) in conjunction with importance, total dominance, and landscape indices. The AHP method constructs a hierarchical structure of assessment indicators, builds a judgment matrix through pairwise comparisons, calculates the eigenvectors to obtain the weights of each indicator, and finally comprehensively assesses the health status of the mangrove ecosystem. A knowledge graph reasoning system was established, constructing a semantic network of the various elements and relationships of the mangrove ecosystem. Knowledge reasoning was performed through reasoning rules to provide decision-making suggestions for mangrove ecosystem management.

[0029] For example, image data of the area was acquired through remote sensing and, after preprocessing, four main community types were identified: *Avicennia marina*, *Kandelia candel*, *Haloxylon ammodendron*, and *Gnaphalium affine*. The dominance of woody plants was calculated, with *Avicennia marina* showing the highest importance value at 105.39%, indicating it is the dominant species in the area. Spatial distribution analysis showed that *Lagerstroemia indica* had the highest aggregation degree, and landscape pattern analysis revealed that the *Avicennia marina* community accounted for the largest area, reaching 82.01%. Environmental factor analysis indicated that this species is most concentrated in the mid- and low-tidal flat areas. Based on these analysis results, the management policy recommendations include: strengthening the planting of native mangrove plants, appropriately controlling the introduction scale of *Avicennia marina*, and maintaining the species diversity and ecological balance of the mangrove ecosystem.

[0030] In this embodiment, high-resolution remote sensing images are acquired through a remote sensing image acquisition device, and combined with field survey data and environmental parameter collection, effective fusion of multi-source data is achieved, improving the integrity and accuracy of the data. Image enhancement algorithms are used to preprocess the remote sensing images, establishing a multi-dimensional database containing spatial information, species information, and environmental parameters, providing comprehensive data support for subsequent analysis. Random forest classification algorithms and deep learning object detection algorithms are used to process the data, achieving not only accurate identification of community types but also extraction of location information of individual mangrove plants, significantly improving the accuracy of spatial distribution analysis. By calculating various indicators of mangrove plants in the woody and herbaceous layers and combining them with spatial statistical algorithms, a quantitative expression of population distribution characteristics is achieved. A landscape index calculation model is used for landscape pattern analysis, and a connectivity analysis model is established using graph theory algorithms, making the identification of key patches more accurate and objective. Gradient boosting algorithms and Bayesian network algorithms are used for environmental factor impact analysis, significantly improving the accuracy of population distribution prediction. Finally, through hierarchical analysis and a knowledge graph reasoning system, intelligent and scientific decision-making for mangrove ecosystem management is achieved.

[0031] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0032] (1) The study area is imaged by a remote sensing image acquisition device to obtain high-resolution remote sensing images;

[0033] (2) Set up transects in the study area by random sampling, set up quadrats in each transect, and collect field survey data and environmental parameters;

[0034] (3) Spatial registration of high-resolution remote sensing images is performed by geometric correction algorithm, spectral correction is performed by radiometric calibration algorithm, and atmospheric correction algorithm is used to eliminate atmospheric effects, thus obtaining corrected high-resolution remote sensing images.

[0035] (4) The corrected high-resolution remote sensing image is enhanced by the histogram equalization algorithm to obtain the enhanced high-resolution remote sensing image.

[0036] (5) Based on the enhanced high-resolution remote sensing images, field survey data and environmental parameters, establish a data standardization table and determine the data storage format;

[0037] (6) Transform the data standardization table into a multidimensional database containing spatial information, species information and environmental parameters through data structuring processing.

[0038] Specifically, multispectral remote sensing images of the study area were acquired using a remote sensing image acquisition device carried by the GeoEye-1 satellite. The image acquisition device includes multiple sensor bands, collecting reflectance information in the visible and near-infrared bands. The visible light band reflects the chlorophyll content of vegetation, while the near-infrared band is more sensitive to vegetation biomass and canopy structure. High-resolution remote sensing images were obtained by combining multiple bands. Multiple transects were established within the study area using a random sampling method, with the transects designed to take into account changes in tidal gradients. Fixed-size quadrats were laid out in each transect, including 10m×10m and 20m×5m sizes. Various growth parameters of mangrove plants, such as tree height, diameter at breast height (DBH), and crown width, were recorded within the quadrats, along with community parameters such as species type, quantity, and cover. Environmental parameter data, including tidal elevation, soil salinity, and pH, were collected concurrently with quadrat sampling.

[0039] The acquired high-resolution remote sensing imagery is first geometrically corrected by establishing a transformation relationship between image coordinates and geographic coordinates through the selection of ground control points, thus eliminating geometric distortions. Then, a radiometric calibration algorithm is used to convert the image's numerical identifiers into actual spectral reflectance, correcting for differences in spectral response. Next, an atmospheric correction algorithm is employed to remove the effects of atmospheric scattering and absorption on image quality, resulting in a corrected image that reflects the true spectral characteristics of ground features. Image enhancement processing is then applied to the corrected high-resolution remote sensing image, using a histogram equalization algorithm to adjust the image's grayscale distribution. Histogram equalization redistributes pixel grayscale values, fully utilizing the entire grayscale range, enhancing image contrast, highlighting the texture features and boundary information of the mangrove canopy, and thus obtaining a visually superior enhanced image.

[0040] Enhanced high-resolution remote sensing imagery was integrated with field survey data and environmental parameters to establish a unified data standardization table. Data standardization included three aspects: format standardization, unit standardization, and scale standardization. Format standardization converted data from different sources into the same data format; unit standardization ensured that all measurement data used the same units of measurement; and scale standardization normalized data with different dimensions. Through data structuring, the standardized table was transformed into a multidimensional database. Data structuring established the relationships between spatial information, species information, and environmental parameters, forming a hierarchical data storage system. The multidimensional database was stored in matrix form, with each data record containing information from multiple dimensions, including spatial coordinates, species attributes, and environmental variables.

[0041] For example, multispectral images of the area were acquired using a remote sensing image acquisition device, containing data in four bands: blue, green, red, and near-infrared. Three transects perpendicular to the coastline were established within the study area based on the tidal gradient. Quadrats were laid out at 25-meter intervals along each transect, recording the species composition and growth status within each quadrat. Coordinate transformation relationships were established using spatial control points to complete the geometric correction of the images. Histogram equalization was performed on the corrected images to make the spectral characteristics of different community types more distinct. Finally, all data were integrated into a multidimensional database, establishing a complete data record including spatial location, community characteristics, and environmental factors, providing data support for subsequent analysis. This database not only records the specific location and species composition of each quadrat but also includes environmental parameter information for that location, achieving effective integration of multi-source data.

[0042] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0043] (1) Use feature extraction algorithms to filter spectral and texture features in a multidimensional database to obtain community feature data;

[0044] (2) Select training samples based on community feature data, and substitute the training samples into the random forest classification algorithm to obtain community type identification parameters;

[0045] (3) Based on the community type identification parameters, filter the community distribution information in the multidimensional database to form community type distribution data;

[0046] (4) Input high-resolution remote sensing images into a deep learning target detection algorithm, and generate target detection feature maps after processing by convolutional layers;

[0047] (5) Calculate the boundary range of mangrove plants from the target detection feature map, extract the location information of individual mangrove plants, and establish coordinate data of individual plants;

[0048] (6) Combine community type distribution data and individual plant coordinate data for geospatial processing to generate a community distribution vector data layer.

[0049] Specifically, effective features are extracted from a multidimensional database using feature extraction algorithms. Feature extraction includes two aspects: spectral features and texture features. Spectral features refer to the reflectance characteristics of different mangrove plants in various wavelengths, including reflectance values ​​in the visible and near-infrared bands. Texture features describe the spatial structural characteristics of the mangrove canopy, such as statistical quantities like mean, variance, and entropy. These features are then filtered using feature extraction algorithms to obtain feature combinations that can effectively distinguish different community types, forming community feature data. Based on the obtained community feature data, representative samples are selected as training data. The selection of training samples needs to cover various community types within the study area, including Avicennia marina, Kandelia candel, and Lepidium apetalum communities. These training samples are then input into a random forest classification algorithm. The random forest algorithm constructs multiple decision trees, each trained using a randomly selected subset of features. Finally, the classification result is determined through voting, yielding community type identification parameters.

[0050] The obtained community type identification parameters are used to filter community distribution information in a multidimensional database. During the filtering process, the feature values ​​of each spatial location are matched with a trained classifier to determine the community type at that location, thus forming complete community type distribution data. This data includes the community type identifier and spatial coordinate information for each location.

[0051] High-resolution remote sensing images are input into a deep learning-based object detection algorithm for processing. This algorithm employs a convolutional neural network structure, extracting hierarchical features from the image through multiple convolutional operations. During the convolutional layer processing, local features are first extracted using convolutional kernels of different scales, and then feature dimensionality reduction is performed through pooling layers, ultimately generating a feature map containing target location and category information.

[0052] Based on the target detection feature map, the boundary range of mangrove plants is calculated. Boundary calculation employs edge detection and contour extraction algorithms to accurately locate the boundary position of each mangrove plant. The spatial position of individual mangrove plants is extracted using boundary information, determining their center point coordinates and establishing a complete database of individual plant coordinates. This coordinate data accurately records the spatial distribution of each mangrove plant within the study area. Geospatial processing is performed on the community type distribution data and individual plant coordinate data. Through spatial overlay analysis, the community type to which each plant belongs is determined, and a hierarchical spatial data structure is established, ultimately generating a community distribution vector data layer. This layer contains complete spatial distribution information of mangroves, including both community-scale distribution patterns and the precise location of individual plants.

[0053] For example, feature extraction was first performed on the acquired multispectral remote sensing images. In the near-infrared band, the *Avicennia marina* community exhibited high reflectance values, while in the visible red band, it showed low reflectance values, a characteristic that clearly distinguished it from other community types. Simultaneously, by calculating the gray-level co-occurrence matrix of the images and extracting texture features, it was found that the *Avicennia marina* community possessed high homogeneity and low entropy. These features were combined to form training samples, which were then input into a random forest classifier for training. After training, the classifier was used to identify community types throughout the study area, generating community distribution data. Simultaneously, a deep learning object detection algorithm was used to process the high-resolution images. This algorithm can accurately identify the location of individual mangrove plants, especially in areas with high growth density, and can accurately distinguish the boundaries between adjacent plants. A boundary extraction algorithm was used to determine the spatial extent of each plant, establishing a coordinate database. Finally, the community distribution data was combined with the individual plant location data to generate a complete spatial distribution layer. This layer clearly shows the spatial distribution patterns of different community types, as well as the individual plant distribution characteristics within each community.

[0054] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0055] (1) Extract the number of individual mangrove plants in the woody layer from the community distribution vector data layer, calculate the frequency of species occurrence for each quadrat, and form the relative abundance;

[0056] (2) Count the number of times mangrove plants in the woody layer appear in the quadrats, divide the number of occurrences by the total number of quadrats, and generate the relative frequency.

[0057] (3) Read the diameter at breast height (DBH) values ​​of the woody mangrove plants, calculate the total cross-sectional area, normalize the results, and output the relative significance.

[0058] (4) The relative abundance, relative frequency and relative significance are combined and weighted to obtain the importance value index;

[0059] (5) Extract the coverage area and height information of herbaceous mangrove plants in the community distribution vector data layer, calculate the ratio and then calculate the average value to obtain the total dominance index;

[0060] (6) Select species distribution points in the community distribution vector data layer, and perform spatial statistical operations using the variance-mean ratio method, negative binomial distribution parameter method and clustering index method to calculate the spatial distribution characteristic parameters of the population.

[0061] Specifically, mangrove plant data in the woody layer was extracted from the community distribution vector data layer. This data records the specific attributes of each mangrove plant in the woody layer, including species type, location coordinates, and growth parameters. Individual plants within each quadrat were statistically analyzed, recording the number of individuals of each species within the quadrat to obtain the species' frequency of occurrence. Relative abundance was calculated by dividing the number of individuals of a particular species by the total number of individuals of all species, reflecting the species' numerical dominance in the community. After obtaining the relative abundance data, the distribution of mangrove plants in the woody layer across all quadrats was further analyzed. By recording the number of times each species appeared in different quadrats, this frequency was divided by the total number of quadrats surveyed to obtain the relative frequency data. Relative frequency reflects the spatial prevalence of a species; a higher value indicates a more widespread distribution.

[0062] For mangrove plants in the woody layer, their diameter at breast height (DBH) is also required for analysis. DBH refers to the diameter of a tree measured at 1.3 meters above the ground, and its cross-sectional area can be calculated from the DBH. The cross-sectional area of ​​all individuals of each species is calculated, and the sum of the cross-sectional areas of a particular species is divided by the sum of the cross-sectional areas of all species. After normalization, the relative significance is obtained. Relative significance reflects the dominant position of a species in the community. The weighted sum of the calculated relative abundance, relative frequency, and relative significance yields the importance index. The importance index comprehensively reflects the species' position in the community and is an important parameter for evaluating species importance. For different ecosystems, the weights of these three parameters can be adjusted according to the specific circumstances.

[0063] For herbaceous mangrove plants, cover area and height information were extracted from the community distribution vector data layer. Cover area refers to the projected area of ​​herbaceous plants on the ground, obtained through field measurements or remote sensing image interpretation. Height information is obtained from plant height data recorded through field measurements. The ratio of cover area to height information was calculated, and the average value was calculated for multiple samples to obtain the overall dominance index. The overall dominance index reflects the degree of dominance of herbaceous plants in the community. Finally, species distribution point data from the community distribution vector data layer were selected, and spatial distribution characteristics were analyzed using three different statistical methods. The variance-to-mean ratio method determines the distribution pattern by calculating the ratio of the variance to the mean of the number of individuals of a species; the negative binomial distribution parameter method assesses the spatial distribution pattern based on the frequency distribution characteristics of individual data; and the clustering index method quantifies the distribution characteristics by calculating the spatial clustering degree of individual species. The combined application of these three methods can comprehensively reflect the spatial distribution characteristics of the population.

[0064] For example, firstly, the distribution data of *Avicennia marina* was extracted from the community distribution vector data layer, and the number of individuals was counted in 20 quadrats. Statistical analysis revealed that *Avicennia marina* accounted for the largest proportion of individuals in all quadrats, resulting in a high relative abundance. Further analysis showed that *Avicennia marina* was distributed in 15 quadrats, and its relative frequency was calculated. Measuring the diameter at breast height (DBH) of each *Avicennia marina* plant revealed a large average DBH and a significant proportion of the total cross-sectional area, resulting in a high relative significance. Weighted summation of these three indicators confirmed that *Avicennia marina* has the highest importance value in the woody layer. For herbaceous plants such as *Hydrangea spp.*, their coverage area and plant height were measured to calculate the overall dominance index. Finally, the spatial distribution of *Avicennia marina* was analyzed. The results of calculations using three spatial statistical methods showed that *Avicennia marina* exhibits a significant aggregated distribution characteristic in the mid-tidal zone, and this distribution characteristic is closely related to environmental factors such as tide level and salinity.

[0065] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0066] (1) Extract patch area data from the spatial distribution characteristic parameters of the population, and divide the maximum patch area by the total area of ​​the study area to obtain the maximum patch index;

[0067] (2) Extract the adjacency information of each patch from the spatial distribution characteristic parameters of the population, and generate the clustering index by calculating the spatial adjacency probability;

[0068] (3) Statistical analysis is performed on the patch adjacency information in the spatial distribution characteristic parameters of the population. The similarity adjacency ratio is calculated based on the total boundary length, and the similarity adjacency percentage is output.

[0069] (4) Screen patch pair data from the population spatial distribution characteristic parameters, calculate the ratio of the number of adjacent patches to the maximum possible number of adjacent patches, and obtain the patch aggregation index.

[0070] (5) Construct an adjacency matrix based on the maximum patch index, clustering index, similarity adjacency percentage, and patch aggregation index, and calculate node connectivity using graph theory algorithms;

[0071] (6) Perform network analysis on node connectivity and patch spatial distribution data to obtain key patch identification results.

[0072] Specifically, patch area data is extracted from the spatial distribution characteristics of the population. A patch refers to a continuous spatial unit with the same landscape type. The largest continuous patch area is identified using spatial analysis tools, and this area is divided by the total area of ​​the entire study area to calculate the maximum patch index. The maximum patch index reflects the relative size of dominant patches in the landscape; a larger value indicates the presence of a dominant patch type in the landscape.

[0073] For adjacency analysis among patches, it is necessary to extract the adjacency information of each patch from the spatial distribution characteristic parameters of the population. Adjacency information includes the spatial relationship data between each patch and its neighboring patches. A clumpiness index is generated by calculating the spatial adjacency probability; the calculation formula is as follows:

[0074]

[0075] Where, α ij β represents the adjacency length between patch i and patch j. ij The similarity coefficient for patch types is represented by γ, the total boundary length by δ, the number of landscape types by λ, the normalization coefficient by n, the total number of patches by n, and the number of patches adjacent to patch i. The clustering index reflects the degree of spatial aggregation of patches of the same type.

[0076] Statistical analysis was further performed on patch adjacency information in the spatial distribution characteristic parameters of the population. The ratio of the boundary length within each landscape type to the total boundary length of that type was calculated to obtain the percentage of similar adjacency. In the data processing, the boundaries of all patches were first identified, and the common boundary length between each patch and its neighbors was calculated. Then, the boundary lengths between patches of the same type were summed, and finally divided by the total boundary length of that type of patch. When filtering patch pair data from the spatial distribution characteristic parameters of the population, the adjacency relationship of each patch needs to be identified. A patch pair refers to two spatially adjacent patch units. The ratio of the current number of adjacent patches to the theoretically maximum number of possible adjacent patches was calculated to obtain the patch aggregation index. The patch aggregation index reflects the spatial clustering degree of the landscape type.

[0077] Based on the calculated maximum patch index, clustering index, similarity adjacency percentage, and patch aggregation index, an adjacency matrix is ​​constructed. The adjacency matrix is ​​a two-dimensional array representing the connectivity relationships between patches, with each element reflecting the connection strength. Node connectivity is calculated using graph theory algorithms, treating each patch as a node in the network and the connections between patches as edges. Network characteristic parameters such as degree and centrality are calculated for each node. Finally, network analysis is performed using node connectivity and patch spatial distribution data to identify key patches by assessing node importance. Key patches are patch units that play a crucial connecting role in the entire ecological network, and these patches are essential for maintaining the connectivity and stability of the ecosystem.

[0078] For example, the mangrove communities in the study area were first divided into patches, and different types of patches, such as *Avicennia marina*, *Kandelia candel*, and *Haloxylon ammodendron*, were identified through remote sensing image interpretation. Area statistics revealed that the *Avicennia marina* community constituted the largest continuous patch. Analysis of the spatial relationships between patches showed that *Avicennia marina* patches had more boundary contacts with other types of patches, exhibiting a high clumping index. Statistical analysis of the similarity adjacency percentage showed that patches of the same type tended to be adjacent to each other, especially in the mid-tidal zone, where *Avicennia marina* patches exhibited significant spatial aggregation characteristics. Analysis of patch aggregation degree showed that patches of different community types exhibited different degrees of aggregation in spatial distribution, with the *Avicennia marina* community showing the highest aggregation degree. By constructing an adjacency matrix and conducting network analysis, patches that play a key role in the connectivity of the entire mangrove ecosystem were identified. These key patches are mainly distributed in the central intertidal zone and play an important role in maintaining the spatial continuity of the mangrove ecosystem.

[0079] In one specific embodiment, the process of executing step S105 may specifically include the following steps:

[0080] (1) Match and align the key patch identification results with environmental parameters to generate patch environment association data;

[0081] (2) Input patch environment association data into gradient boosting algorithm, and distinguish training data from validation data by randomly splitting the dataset;

[0082] (3) Use the gradient boosting algorithm to calculate feature weights for the training data and establish an environmental factor influence model;

[0083] (4) Validate the environmental factor impact model based on the validation data and output the environmental factor impact weight data;

[0084] (5) Construct a probability distribution network of the influence of environmental factors on weight data using Bayesian network algorithm, and calculate the conditional probability of each node;

[0085] (6) The conditional probability is transformed into the population distribution probability, and combined with the environmental factor influence model, the population distribution prediction results are obtained.

[0086] Specifically, the spatial location information of each key patch is spatially overlaid with corresponding environmental parameter data, including tidal level, salinity, and soil pH. Environmental feature descriptions for each patch are established through spatial correspondence, forming patch environmental association data. This generated patch environmental association data is then input into a gradient boosting algorithm for processing. Gradient boosting is an iterative decision tree algorithm that improves prediction accuracy by constructing multiple decision trees. First, the dataset is randomly split, with 70% used as training data and 30% as validation data to ensure the model's generalization ability.

[0087] For the training data, the gradient boosting algorithm is used to calculate the feature weights, and the calculation formula is as follows:

[0088]

[0089] Among them, W ij ω represents the weight of the influence of the i-th environmental factor on the j-th patch. k This represents the weight coefficient of the k-th positive influence factor. θ represents the intensity of the effect of the k-th positive factor on the plaque. m ψ represents the weighting coefficient of the m-th negative influence factor. ijm ξ represents the intensity of the effect of the m-th negative factor on the plaque. ij represents the normalized adjustment coefficient, p represents the number of positive factors, and q represents the number of negative factors. This calculation is used to establish an environmental factor impact model.

[0090] The environmental factor impact model was validated using validation data. During validation, the differences between the model's predicted values ​​and actual observations were compared. The model parameters were iteratively optimized, ultimately outputting the environmental factor impact weights. These weights reflect the degree of influence of different environmental factors on mangrove distribution. Based on the environmental factor impact weights, a probability distribution network was constructed using a Bayesian network algorithm. A Bayesian network is a probabilistic graphical model that represents the conditional dependencies between variables using a directed acyclic graph. In the network, each node represents an environmental factor, and the connections between nodes represent the mutual influence relationships between factors. The conditional probability of each node was calculated.

[0091] Finally, the conditional probability was transformed into population distribution probability, and a comprehensive analysis was conducted in conjunction with an environmental factor influence model to obtain the predicted results of mangrove population distribution. The predicted results include the probability of mangrove distribution in different regions and the main influencing factors.

[0092] For example, analyzing a mangrove reserve, the first step is to spatially match identified key patches with environmental monitoring data. In the intertidal zone, each patch corresponds to a set of environmental parameters, including tidal level, soil salinity, and pH. Data matching revealed that the *Avicennia marina* community is mainly distributed in the mid-tidal zone, which exhibits a specific combination of environmental characteristics. Inputting this matching data into a gradient boosting algorithm, after training, it was found that tidal level has the greatest influence on community distribution, followed by soil salinity. In the validation phase, the model's predicted community distribution showed a high degree of consistency with actual observations. Bayesian network analysis indicated that the *Avicennia marina* community has the highest distribution probability when the tidal level is within a certain range and the soil salinity is moderate.

[0093] In one specific embodiment, the process of executing step S106 may specifically include the following steps:

[0094] (1) Integrate the population distribution prediction results, importance value indicators, total dominance indicators and landscape indices into a set of evaluation indicators, and determine the weights of the indicators through principal component analysis;

[0095] (2) Establish a discriminant matrix for the set of evaluation indicators, calculate the eigenvectors through the analytic hierarchy process, and output the relative importance of the indicators;

[0096] (3) Substitute the relative importance of the indicators into the analytic hierarchy process to calculate the consistency ratio and generate the indicator weight correction value;

[0097] (4) Construct a health assessment model based on the indicator weight correction value and calculate the health score of the mangrove ecosystem;

[0098] (5) Input the mangrove ecosystem health score into the knowledge graph reasoning system and extract management rules through semantic analysis;

[0099] (6) Combine the diagnostic results of the joint management rules and health assessment model to generate management decision recommendations for mangrove ecosystems.

[0100] Specifically, multidimensional evaluation indicators, including population distribution predictions, importance values, overall dominance, and landscape indices, are integrated. These indicators reflect different aspects of the mangrove ecosystem. Principal component analysis (PCA) is used to reduce the dimensionality of the indicators and determine their weights. PCA calculates the correlations between indicators, combining correlated indicators into new comprehensive indicators to avoid information overlap. A discriminant matrix is ​​established for the integrated set of evaluation indicators, and the analytic hierarchy process (AHP) is used to calculate the indicator weights. The elements in the discriminant matrix represent the relative importance of different indicators. By calculating the eigenvectors of the discriminant matrix, the relative importance of each indicator is obtained. The relative importance reflects the degree of influence of each indicator in the ecosystem health assessment.

[0101] Substituting the relative importance of the indicators into the analytic hierarchy process (AHP), the consistency ratio is calculated using the following formula:

[0102]

[0103] Where, ρ ij η represents the relative importance value between the i-th indicator and the j-th indicator. ij σ represents the weighting adjustment factor. ij The correlation coefficient between indicators is represented by ∈, the largest eigenvalue is represented by τ, the matrix dimension is represented by υ, the random consistency index is represented by ζ, the correction coefficient is represented by n, and the total number of indicators is represented by n. By calculating the consistency ratio, correction values ​​for the indicator weights are generated to ensure the scientific validity of the evaluation results.

[0104] Based on the corrected index weights, a health assessment model for mangrove ecosystems was constructed. This model comprehensively considers multiple aspects such as population distribution, community structure, and landscape pattern, and calculates a health score for the mangrove ecosystem through weighted calculation. The health score reflects the overall condition of the mangrove ecosystem.

[0105] The calculated ecosystem health score is input into a knowledge graph reasoning system, and management rules are extracted using semantic analysis. The knowledge graph contains the relationships between various components of the mangrove ecosystem, as well as management experiences under different health conditions. Through semantic analysis, the health assessment results are matched with existing management knowledge to extract applicable management rules. Finally, the extracted management rules are integrated with the diagnostic results of the health assessment model to generate targeted management decision-making recommendations for the mangrove ecosystem. These recommendations include specific measures such as species conservation, habitat restoration, and community optimization.

[0106] For example, data on various evaluation indicators for the region were first collected. Population distribution predictions showed multiple species communities, including *Avicennia marina*, *Kandelia candel*, and *Gnaphalium affine*. Importance value analysis revealed that *Avicennia marina* was dominant in the community. The overall dominance index indicated that the herbaceous layer was dominated by *Avicennia marina*, and landscape index analysis showed that the mangroves in the region exhibited a distinct patchy distribution. After determining the weights of these indicators through principal component analysis, a discriminant matrix was established for hierarchical analysis. After consistency testing and weight correction, a relatively accurate indicator weight system was obtained. Based on the health assessment model, the ecosystem in the region was in a sub-healthy state, with the main problems being low species diversity and an excessively high proportion of dominant species. Management rules extracted through knowledge graph analysis indicated the need to strengthen the planting of native mangrove plants and moderately control the expansion of *Avicennia marina*. Combining these specific diagnostic results, management decision recommendations were ultimately formulated, including increasing species diversity, optimizing community structure, and protecting key habitats.

[0107] The analysis method for mangrove ecosystem population distribution in the embodiments of this application has been described above. The analysis system for mangrove ecosystem population distribution in the embodiments of this application is described below. Please refer to [link / reference]. Figure 2 One embodiment of the analysis system for mangrove ecosystem population distribution in this application includes:

[0108] The acquisition module is used to acquire high-resolution remote sensing images through a remote sensing image acquisition device, collect field survey data and environmental parameters in the study area, preprocess the high-resolution remote sensing images through image enhancement algorithms, and establish a multidimensional database containing spatial information, species information and environmental parameters.

[0109] The identification module is used to identify community types in the data in the multidimensional database using a random forest classification algorithm, and to extract the location information of individual mangrove plants from the high-resolution remote sensing image using a deep learning target detection algorithm, thereby generating a community distribution vector data layer.

[0110] The calculation module is used to calculate the relative abundance, relative frequency, and relative significance of mangrove plants in the woody layer based on the community distribution vector data layer, to obtain the importance value index; to calculate the relative cover and relative height of mangrove plants in the herbaceous layer, to obtain the total dominance index; and to calculate the spatial distribution characteristic parameters of the population through spatial statistical algorithms.

[0111] The analysis module is used to perform landscape pattern analysis using a landscape index calculation model based on the spatial distribution characteristic parameters of the population, to obtain the maximum patch index, clustering index, similarity adjacency percentage and patch aggregation index, and to establish a connectivity analysis model based on graph theory algorithm to obtain key patch identification results.

[0112] The correlation module is used to correlate the key patch identification results with environmental parameters, establish an environmental factor influence model through gradient boosting algorithm, perform uncertainty analysis using Bayesian network algorithm, and obtain population distribution prediction results.

[0113] The generation module is used to construct a health assessment model based on the population distribution prediction results, combined with the importance value index, total dominance index and landscape index, through the hierarchical analysis method, establish a knowledge graph reasoning system, and generate management decision-making suggestions for mangrove ecosystems.

[0114] Through the collaborative efforts of the aforementioned components, high-resolution remote sensing images were acquired using a remote sensing image acquisition device. Combined with field survey data and environmental parameter collection, effective fusion of multi-source data was achieved, improving data integrity and accuracy. Image enhancement algorithms were used to preprocess the remote sensing images, establishing a multi-dimensional database containing spatial information, species information, and environmental parameters, providing comprehensive data support for subsequent analysis. Random forest classification algorithms and deep learning object detection algorithms were used to process the data, enabling not only accurate identification of community types but also extraction of location information of individual mangrove plants, significantly improving the accuracy of spatial distribution analysis. By calculating various indicators of mangrove plants in the woody and herbaceous layers and combining them with spatial statistical algorithms, quantitative expression of population distribution characteristics was achieved. Landscape pattern analysis was conducted using a landscape index calculation model, and a connectivity analysis model was established using graph theory algorithms, making the identification of key patches more accurate and objective. Gradient boosting algorithms and Bayesian network algorithms were used for environmental factor impact analysis, significantly improving the accuracy of population distribution prediction. Finally, through hierarchical analysis methods and a knowledge graph reasoning system, intelligent and scientific decision-making for mangrove ecosystem management was realized.

[0115] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for analyzing population distribution in mangrove ecosystems, characterized in that, include: High-resolution remote sensing images are acquired using a remote sensing image acquisition device, and field survey data and environmental parameters are collected in the study area. The high-resolution remote sensing images are preprocessed using image enhancement algorithms to establish a multidimensional database containing spatial information, species information and environmental parameters. Random forest classification algorithm was used to identify community types in a multidimensional database. A deep learning object detection algorithm was used to extract the location information of individual mangrove plants from high-resolution remote sensing images, generating a community distribution vector data layer. Based on this layer, relative abundance, relative frequency, and relative salience of mangrove plants in the woody layer were calculated to obtain importance indices. Relative cover and relative height of mangrove plants in the herbaceous layer were calculated to obtain the overall dominance index. Spatial statistical algorithms were used to calculate population spatial distribution characteristic parameters. Based on these parameters, a landscape index calculation model was used to analyze the landscape pattern, obtaining the maximum patch index, clumpiness index, similarity adjacency percentage, and patch aggregation index. A connectivity analysis model was established based on graph theory algorithms to obtain key patch identification results. This paper analyzes the correlation between key patch identification results and environmental parameters, establishes an environmental factor influence model using the gradient boosting algorithm, and performs uncertainty analysis using a Bayesian network algorithm to obtain population distribution prediction results. The process includes: matching and aligning key patch identification results with environmental parameters to generate patch-environment correlation data; inputting the patch-environment correlation data into the gradient boosting algorithm, distinguishing training and validation data through random dataset splitting; calculating feature weights on the training data using the gradient boosting algorithm to establish an environmental factor influence model; validating the environmental factor influence model based on validation data and outputting environmental factor influence weight data; constructing a probability distribution network for the environmental factor influence weight data using a Bayesian network algorithm and calculating the conditional probability of each node; converting the conditional probability into population distribution probability and combining it with the environmental factor influence model to obtain population distribution prediction results. Based on population distribution predictions, and combining importance, total dominance, and landscape indices, a health assessment model is constructed using the analytic hierarchy process (AHP). A knowledge graph reasoning system is then established to generate management decision-making recommendations for mangrove ecosystems. This includes integrating population distribution predictions, importance, total dominance, and landscape indices into a set of evaluation indicators, determining indicator weights through principal component analysis, establishing a discriminant matrix for the evaluation indicator set, calculating eigenvectors using AHP, and outputting the relative importance of the indicators. Finally, the relative importance of the indicators is substituted into AHP to calculate the consistency ratio, generating indicator weight correction values. A health assessment model is constructed based on the indicator weight correction value. The model considers population distribution, community structure, and landscape pattern, and calculates the health score of the mangrove ecosystem by weighting. The health score of the mangrove ecosystem is input into the knowledge graph reasoning system, and management rules are extracted through semantic analysis. The management rules are combined with the diagnostic results of the health assessment model to generate management decision suggestions for the mangrove ecosystem.

2. The method for analyzing population distribution in mangrove ecosystems according to claim 1, characterized in that, High-resolution remote sensing images were acquired using a remote sensing image acquisition device, and field survey data and environmental parameters were collected within the study area. The high-resolution remote sensing images were preprocessed using image enhancement algorithms to establish a multidimensional database containing spatial information, species information, and environmental parameters. This process included: imaging the study area using a remote sensing image acquisition device to obtain high-resolution remote sensing images; establishing transects within the study area using a random sampling method, setting up quadrats in each transect, and collecting field survey data and environmental parameters; spatially registering the high-resolution remote sensing images using a geometric correction algorithm, performing spectral correction using a radiometric calibration algorithm, and eliminating atmospheric effects using an atmospheric correction algorithm to obtain corrected high-resolution remote sensing images; enhancing the corrected high-resolution remote sensing images using a histogram equalization algorithm to obtain enhanced high-resolution remote sensing images; establishing a data standardization table based on the enhanced high-resolution remote sensing images, field survey data, and environmental parameters, and determining the data storage format; and converting the data standardization table into a multidimensional database containing spatial information, species information, and environmental parameters through data structuring processing.

3. The method for analyzing population distribution in mangrove ecosystems according to claim 1, characterized in that, This study utilizes a random forest classification algorithm to identify community types from data in a multidimensional database. A deep learning target detection algorithm extracts the location information of individual mangrove plants from high-resolution remote sensing images, generating a community distribution vector data layer. The process includes: filtering spectral and textural features from the multidimensional database using a feature extraction algorithm to obtain community feature data; selecting training samples based on the community feature data and inputting them into the random forest classification algorithm to obtain community type identification parameters; filtering community distribution information from the multidimensional database based on the community type identification parameters to form community type distribution data; inputting high-resolution remote sensing images into a deep learning target detection algorithm, which generates a target detection feature map after convolutional layer processing; calculating the boundary range of mangrove plants from the target detection feature map, extracting the location information of individual mangrove plants, and establishing individual plant coordinate data; and performing geospatial processing on the community type distribution data and individual plant coordinate data to generate a community distribution vector data layer.

4. The method for analyzing population distribution in mangrove ecosystems according to claim 1, characterized in that, Based on the community distribution vector data layer, relative abundance, relative frequency, and relative significance of mangrove plants in the woody layer were calculated to obtain importance indicators. Relative cover and relative height of mangrove plants in the herbaceous layer were calculated to obtain the overall dominance index. Spatial statistical algorithms were used to calculate the spatial distribution characteristic parameters of the population, including: extracting the number of individual mangrove plants in the woody layer from the community distribution vector data layer; calculating the frequency of species occurrence for each quadrat to form relative abundance; counting the number of times mangrove plants in the woody layer appeared in the quadrats; dividing the number of occurrences by the total number of quadrats to generate relative frequency; and reading... The diameter at breast height (DBH) values ​​of mangrove plants in the woody layer were collected, and the sum of their cross-sectional areas was calculated and normalized to output relative significance. A weighted summation of relative abundance, relative frequency, and relative significance was performed to obtain the importance index. The coverage area and height information of mangrove plants in the herbaceous layer were extracted from the community distribution vector data layer, and the ratios were calculated and averaged to obtain the overall dominance index. Species distribution points in the community distribution vector data layer were selected, and spatial statistical operations were performed using the variance-to-mean ratio method, the negative binomial distribution parameter method, and the clustering index method to calculate the spatial distribution characteristic parameters of the population.

5. The method for analyzing population distribution in mangrove ecosystems according to claim 1, characterized in that, Based on the spatial distribution characteristics of the population, a landscape pattern analysis was conducted using a landscape index calculation model to obtain the maximum patch index, clustering index, similarity adjacency percentage, and patch aggregation degree index. A connectivity analysis model was established based on graph theory algorithms to obtain key patch identification results, including: extracting patch area data from the spatial distribution characteristics of the population, dividing the maximum patch area by the total area of ​​the study area to obtain the maximum patch index; extracting the adjacency information of each patch from the spatial distribution characteristics of the population, and generating a clustering index through spatial adjacency probability calculation; statistically analyzing the patch adjacency information from the spatial distribution characteristics of the population, calculating the similarity adjacency ratio based on the total boundary length, and outputting the similarity adjacency percentage; filtering patch pair data from the spatial distribution characteristics of the population, calculating the ratio of the number of adjacent patches to the maximum possible number of adjacent patches to obtain the patch aggregation degree index; and constructing an adjacency matrix based on the maximum patch index, clustering index, similarity adjacency percentage, and patch aggregation degree index, and calculating node connectivity using graph theory algorithms. By performing network analysis on node connectivity and patch spatial distribution data, the key patch identification results are obtained.

6. An analysis system for mangrove ecosystem population distribution, used to implement the analysis method for mangrove ecosystem population distribution as described in any one of claims 1-5, characterized in that, include: The acquisition module is used to acquire high-resolution remote sensing images through a remote sensing image acquisition device, collect field survey data and environmental parameters in the study area, preprocess the high-resolution remote sensing images through image enhancement algorithms, and establish a multidimensional database containing spatial information, species information and environmental parameters. The identification module is used to identify community types in data from a multidimensional database using a random forest classification algorithm, and to extract the location information of individual mangrove plants from high-resolution remote sensing images using a deep learning target detection algorithm, thereby generating a community distribution vector data layer. The calculation module is used to calculate the relative abundance, relative frequency, and relative significance of mangrove plants in the woody layer based on the community distribution vector data layer, to obtain the importance value index; to calculate the relative cover and relative height of mangrove plants in the herbaceous layer, to obtain the total dominance index; and to calculate the spatial distribution characteristic parameters of the population through spatial statistical algorithms. The analysis module is used to perform landscape pattern analysis based on the spatial distribution characteristics of the population and the landscape index calculation model to obtain the maximum patch index, clustering index, similarity adjacency percentage and patch aggregation index. Based on graph theory algorithm, a connectivity analysis model is established to obtain the key patch identification results. The correlation module is used to correlate the key patch identification results with environmental parameters, establish an environmental factor impact model through gradient boosting algorithm, perform uncertainty analysis using Bayesian network algorithm, and obtain population distribution prediction results. The generation module is used to construct a health assessment model based on the population distribution prediction results, combined with importance value indicators, total dominance indicators and landscape indices, and to establish a knowledge graph reasoning system to generate management decision-making suggestions for mangrove ecosystems.

Images