Assessment Method for Watershed Ecological Health Threshold Based on Landscape Fragmentation and Microbial Network Modularity

Abstract

The present invention discloses a method for evaluating the ecological health threshold of a watershed based on landscape fragmentation and microbial network modularity, which relates to the cross - technical field of watershed ecology and microbiomics, and solves the problems in the prior art that the correlation mechanism between landscape fragmentation and microbial network stability is unclear and the restoration strategy lacks threshold warning. The method includes sampling point layout and sample collection, calculation of landscape fragmentation index, clustering and grouping of sampling points, construction of microbial symbiotic network, calculation of modularity index, modeling of the relationship between landscape fragmentation and modularity, and identification of critical thresholds; dynamically correlating landscape fragmentation with microbial network modularity, quantifying the relationship between the two through a scientific and rigorous method, opening up new perspectives and methods for watershed ecological research, and providing a clear quantitative basis for watershed ecological health early warning and land use management based on the identification of critical thresholds of landscape fragmentation.

Description

Technical Field

[0001] The present invention relates to the field of the intersection of watershed ecology and microbiome, and in particular to a watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity. Background Art

[0002] The health of watershed ecosystems is increasingly disturbed by human activities. Urban expansion, agricultural development and road network construction have led to an increase in the density of natural landscape patches, increased edge complexity and reduced connectivity, forming a highly fragmented spatial pattern. This change in landscape heterogeneity not only hinders material circulation and energy flow, but also directly affects the habitat suitability of microbial communities by changing microhabitat conditions, thereby threatening the functional stability of watershed ecosystems. As the core driver of biogeochemical cycles, the community structure and interaction network of microorganisms have highly sensitive response characteristics to ecosystem health and are important indicators for ecosystem health assessment.

[0003] Existing watershed health assessment methods only use microbial diversity index or metabolic function indicators, which have certain limitations and are difficult to fully reflect the status of ecosystems affected by external interference. Traditional landscape ecological models mostly focus on the impact of landscape pattern on water quality or macro-biodiversity, while ignoring its role in the stability of microbial networks. Microorganisms are the core executors of ecosystem functions. The modularity of their symbiotic networks can characterize the anti-interference ability and functional coordination of the community, but existing technologies have not yet established a quantitative correlation between landscape fragmentation index and modularity dynamics. In watershed ecological health management, there is a lack of effective methods to assess the impact of landscape fragmentation on microbial communities and even the entire watershed ecosystem, and it is difficult to accurately determine the critical threshold for significant changes in ecosystem health. Therefore, developing a watershed ecological health threshold assessment method that can comprehensively consider landscape fragmentation and microbial network modularity is of great significance for scientifically guiding watershed ecological protection and land use planning. Summary of the invention

[0004] The purpose of the present invention is to provide a watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity, so as to solve the problems in the prior art that the association mechanism between landscape fragmentation and microbial network stability is unclear and the restoration strategy lacks threshold warning.

[0005] The present invention provides a watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity, comprising:

[0006] Step 1: Arrange sampling points according to the land use type, topographic division and river hydrological characteristics in the basin, record the longitude and latitude coordinates and collect sediment samples;

[0007] Step 2: Process the satellite remote sensing images of the basin, extract patch density, edge density and aggregation degree in combination with the positions of sampling points, and construct the comprehensive landscape fragmentation index for each sampling point;

[0008] Step 3: Conduct standardized preprocessing on the comprehensive landscape fragmentation indices of all sampling points, determine the optimal number of clusters using the elbow method, and execute the K-means clustering algorithm to divide the sampling points into several fragmentation level groups;

[0009] Step 4: Measure the microbial species composition of all sediment samples, calculate the Spearman correlation between operational taxonomic units based on the species abundance matrix, screen out significant correlation relationships, and construct a microbial co-occurrence network using the igraph package in R language;

[0010] Step 5: Conduct modularity analysis on each microbial co-occurrence network using the Louvain algorithm, and calculate the modularity index of each fragmentation level group;

[0011] Step 6: Apply the locally weighted regression fitting model to the mean value of the comprehensive landscape fragmentation index of all fragmentation level groups and the corresponding modularity index of each fragmentation level group to construct a non-linear response curve of landscape fragmentation and modularity index at the regional scale;

[0012] Step 7: Determine the critical threshold of the comprehensive landscape fragmentation index at which the modularity significantly decreases by minimizing the Akaike information criterion value.

[0013] Furthermore, Step 1 includes:

[0014] Reasonably arrange sampling points comprehensively considering the land use type, topographic zoning and river hydrological characteristics within the basin;

[0015] Use a GPS positioning device to record the longitude and latitude coordinates of each sampling point;

[0016] Collect riverbed sediment samples at each sampling point. Collect the surface sediment with a depth of 0 - 25 cm. Each sampling point should be repeated at least 3 times, and the amount of bottom sediment samples collected each time should be no less than 100 g. After fully mixing the surface sediment samples, store them in an environment of -20 °C until they are transported back to the laboratory.

[0017] Furthermore, Step 2 includes:

[0018] Obtain and process the satellite remote sensing images of the basin;

[0019] Perform supervised classification on the remote sensing images with the aid of a remote sensing image processing platform;

[0020] Use the ArcGIS platform to divide the river from the river source to the sampling point, set a buffer at a preset distance to extract the land use data within the buffer, import the processed remote sensing image into Fragstats, and calculate the landscape pattern index;

[0021] , where PD is the patch density, N is the number of patches, and A is the landscape area;

[0022] , where ED is the edge density and E is the total length of the patch edges;

[0023] , where AI is the aggregation index, is the number of adjacent patches of the same type;

[0024] Construct a comprehensive landscape fragmentation index:

[0025]

[0026] where FI is the comprehensive landscape fragmentation index.

[0027] Furthermore, in the step of supervised classification of the remote sensing image by means of a remote sensing image processing platform, the maximum likelihood method is used for land use classification, and the classification result is verified by the Kappa coefficient, Kappa > 0.75; the deletion threshold of isolated nodes in network construction is set to Degree = 0, that is, the nodes with Degree = 0 will be deleted from the network.

[0028] Furthermore, the step three includes:

[0029] Standardize the comprehensive landscape fragmentation index calculated for all sampling points to eliminate the influence of different dimensions on the clustering result; use the normalization formula:

[0030]

[0031] where, is the original FI value, is the standardized comprehensive landscape fragmentation index, and are the minimum and maximum values among the FI values of all sampling points respectively;

[0032] Use the elbow method to determine the optimal number of clusters, perform the K-means clustering algorithm within a certain range, calculate the clustering error at each k value, draw the relationship curve between the number of clusters k and the within-group sum of squares, and observe the inflection point of the curve. The k value corresponding to the inflection point is the optimal number of clusters;

[0033] According to the determined optimal number of clusters, the K-means clustering algorithm is executed, using the standardized comprehensive landscape fragmentation index as input data. Through iterative calculations, each sampling point is assigned to the cluster that is most similar to its comprehensive landscape fragmentation index. Finally, all sampling points are divided into several fragmentation level groups, and the sampling points within each group have similar landscape fragmentation characteristics.

[0034] Further, step four includes:

[0035] The 16S rRNA gene V4 region is sequenced using the Illumina NovaSeq platform, and an operational taxonomic unit abundance table is generated through the QIIME2 process;

[0036] The Spearman correlation between operational taxonomic units in each group is calculated using the Hmisc package in R language, with the correlation coefficient threshold being | |>0.6 and the significance level p<0.01 to screen significant edges; an undirected weighted network is constructed using the igraph package, and isolated nodes are removed.

[0037] Further, step five includes:

[0038] The Louvain algorithm is used to partition modules and calculate the modularity index. The formula is:

[0039]

[0040] where Q is the modularity index, is the proportion of edges within module i, is the proportion of edges of module i.

[0041] Further, step five evaluates the significance of the modularity index through a null model test, with the significance level p<0.05 to verify the reliability of the module partitioning results.

[0042] Further, step six includes:

[0043] Based on the mean of the comprehensive landscape fragmentation index of all fragmentation level groups and the dataset of the modularity index of the corresponding fragmentation level groups, the local weighted regression function in the stats package of R language is used to fit the non-linear relationship. The model formula is:

[0044]

[0045] where Q is the modularity index, FI is the comprehensive landscape fragmentation index, is the local weighted polynomial function, is the error term;

[0046] By iteratively weighting and adjusting the local weight function, an optimal fitting curve is obtained, which describes the response of the modularity index when the degree of landscape fragmentation changes.

[0047] Further, the seventh step includes:

[0048] In the R language environment, call the segmented package to perform piecewise regression analysis on the comprehensive landscape fragmentation index and modularity index data;

[0049] During the piecewise regression process, continuously adjust the segmentation points of the model to find the segmentation scheme that minimizes the Akaike information criterion value; the segmentation point determined by minimizing the Akaike information criterion value is the critical threshold of the comprehensive landscape fragmentation index at which the modularity significantly decreases.

[0050] The present invention has the following beneficial effects: The method for evaluating the watershed ecological health threshold based on landscape fragmentation and microbial network modularity of the present invention dynamically correlates landscape fragmentation and microbial network modularity, quantifies the relationship between the two through a scientific and rigorous method, opens up new perspectives and methods for watershed ecological research, and provides a clear quantitative basis for watershed ecological health early warning and land use management based on the identification of the critical threshold of landscape fragmentation. When the landscape fragmentation index approaches or exceeds this threshold, timely measures should be taken to prevent the degradation of the watershed ecosystem, which has important practical guiding significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0052] Figure 1 It is a flowchart of the method for evaluating the watershed ecological health threshold based on landscape fragmentation and microbial network modularity of the present invention.

[0053] Figure 2 It is a schematic diagram of the locally weighted regression fitting curve of the landscape fragmentation index and the microbial network modularity index. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments and corresponding drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention will be described in detail below with reference to the drawings.

[0055] See also Figure 1 , which is a watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity provided by the present invention, comprising:

[0056] Step 1: Arrange sampling points according to the land use type, terrain division and river hydrological characteristics in the basin, record the longitude and latitude coordinates and collect sediment samples.

[0057] Specifically, according to the land use type, terrain division and river hydrological characteristics in the basin, the sampling points are reasonably arranged with comprehensive consideration, and the various information layers are superimposed and analyzed using Geographic Information System (GIS) technology to determine the location of the sampling points. Land use type refers to the distribution of different land use methods such as cultivated land, forest land, construction land, and water areas. Terrain division refers to the division of different terrain areas such as mountains, plains, and hills. River hydrological characteristics refer to characteristics such as river velocity, flow, water depth, and river width. The sampling points are reasonably arranged with comprehensive consideration so that the sampling points can reflect the characteristics of different areas in the basin as comprehensively as possible.

[0058] A high-precision GPS locator was used to measure and record the longitude and latitude coordinates at each sampling point, accurate to six decimal places, to ensure the accuracy of the sampling point location information.

[0059] At the selected sampling points, use sterile sampling tools to collect riverbed sediment samples. Insert the sampling tool vertically into the riverbed and collect surface sediments at a depth of 0-25 cm. Repeat the collection three times at each sampling point, with each collection amount of no less than 100g. Place the sample in a sterile bag and mix it thoroughly. After marking, quickly place it in a portable refrigerator, maintain a low temperature environment of about 4°C, transport it back to the laboratory as soon as possible, and transfer it to a -20°C refrigerator to prevent changes in the microbial community structure.

[0060] Step 2: Process the satellite remote sensing images of the watershed, extract the patch density, edge density and aggregation degree based on the sampling point locations, and construct a comprehensive landscape fragmentation index for each sampling point.

[0061] Specifically, high-resolution satellite remote sensing data, such as Landsat series satellite images or Sentinel satellite images, are used to obtain remote sensing image data of the study basin. The image acquisition time should be selected during the vegetation growing season and when the weather is clear and cloud-free to ensure image quality. Preprocessing operations such as radiometric calibration, atmospheric correction, and geometric correction are performed on the acquired images in sequence. Radiometric calibration converts the digital quantization values recorded by the sensor into the actual radiance values on the ground, correcting the errors of the sensor itself; atmospheric correction is carried out through the FLAASH module in the remote sensing image processing platform (The Environment for Visualizing Images, ENVI) to remove the influence of atmospheric scattering and absorption of light, so that the image can truly reflect the reflection characteristics of surface features. 45 evenly distributed ground control points are carefully selected on the topographic map, and these points cover different terrain and feature types within the basin, including landmark features such as mountain peaks, river confluences, and road intersections. The quadratic polynomial and cubic convolution interpolation methods are used to geometrically correct the image to ensure the spatial position accuracy of the image and make the subsequent analysis results more reliable.

[0062] With the help of the remote sensing image processing platform, the maximum likelihood method is used to perform supervised classification on the preprocessed remote sensing images. Referring to the "Classification of Current Land Use Status" (GB / T 21010-2017), the land use types are carefully divided into 9 categories, including cultivated land, forest land, construction land, water area, grassland, wetland, unused land, transportation land, and special land.

[0063] After classification, 500 ground truth sample points are randomly selected, and the accurate land use types of these sample points are determined through on-site inspections, historical data inquiries, etc. The Kappa coefficient is calculated using the accuracy verification tool of ENVI software. When the Kappa coefficient is greater than 0.75, it indicates that the classification result is reliable and can accurately reflect the land use status of the basin. The deletion threshold for isolated nodes in network construction is set to Degree = 0, that is, nodes with Degree = 0 will be deleted from the network to ensure the connectivity and effectiveness of the network.

[0064] The ArcGIS platform is used to divide the river from the river source to the sampling point, and buffer zones are set at a preset distance to extract the land use data within the buffer zones. The processed remote sensing images are imported into Fragstats to calculate the landscape pattern indices.

[0065] , where PD is the patch density, N is the number of patches, A is the landscape area, and the unit is hm²;

[0066] , where ED is the edge density, E is the total length of the patch edges, and the unit is m;

[0067] , where AI is the aggregation index, is the number of adjacent patches of the same type;

[0068] Construct a comprehensive landscape fragmentation index:

[0069]

[0070] Among them, FI is the comprehensive landscape fragmentation index. Substitute the patch density (Patch Density, PD), edge density (Edge density, ED), and aggregation index (Aggregation Index, AI) calculated above into the formula to obtain the comprehensive landscape fragmentation index (Fragmentation Index, FI). This index can comprehensively and quantitatively measure the fragmentation degree of the watershed landscape and provide key data support for subsequent analysis.

[0071] Step 3: Perform standardized preprocessing on the comprehensive landscape fragmentation index of all sampling points, use the elbow method to determine the optimal number of clusters, and execute the K-means clustering algorithm to divide the sampling points into several fragmentation level groups.

[0072] Specifically, perform standardized processing on the comprehensive landscape fragmentation index calculated for all sampling points to eliminate the influence of different dimensions on the clustering results; use the normalization formula:

[0073]

[0074] Among them, is the original FI value, is the standardized comprehensive landscape fragmentation index, and are the minimum and maximum values among the FI values of all sampling points respectively;

[0075] Use the elbow method (Elbow Method) to determine the optimal number of clusters. Execute the K-means clustering algorithm within a certain range, such as k = 2 - 10, and calculate the clustering error for each k value. Usually, the within-cluster sum of squares (Within-Cluster Sum of Squares, WCSS) is used to measure it, that is, the sum of the squares of the distances from each sampling point to its affiliated cluster center. Plot the relationship curve between the number of clusters k and the within-cluster sum of squares, and observe the inflection point of the curve. The k value corresponding to the inflection point is the optimal number of clusters.

[0076] According to the determined optimal number of clusters, the K-means clustering algorithm is executed. The standardized comprehensive landscape fragmentation index is used as the input data. Through iterative calculations, each sampling point is assigned to the cluster that is most similar to its comprehensive landscape fragmentation index. Finally, all sampling points are divided into several fragmentation level groups, and the sampling points within each group have similar landscape fragmentation characteristics.

[0077] Step 4: Determine the microbial species composition of all sediment samples. Calculate the Spearman correlation between Operational Taxonomic Units (OTUs) based on the species abundance matrix, screen for significant correlation relationships, and construct a microbial co-occurrence network using the igraph package in R language.

[0078] Specifically, the Illumina NovaSeq platform is used for 16S rRNA gene V4 region sequencing, and an OTU abundance table is generated through the QIIME2 pipeline. First, a professional DNA extraction kit is used to extract microbial DNA from sediment samples. The extraction process is strictly operated according to the kit instructions to ensure that the quality and purity of the extracted DNA meet the sequencing requirements. For example, the QIAGEN PowerSoil DNA Isolation Kit is used to extract microbial DNA from sediment samples, and the extraction process is strictly operated according to the kit instructions to ensure that the quality and purity of the extracted DNA meet the sequencing requirements. After extraction, the quality and concentration of DNA are detected by agarose gel electrophoresis and a nucleic acid concentration detector.

[0079] Then, the extracted DNA is subjected to PCR amplification. The amplification primers are selected as universal primers targeting the 16S rRNA gene V4 region. The amplification reaction system and conditions are optimized to ensure the specificity and efficiency of amplification. After purification of the amplification products, a sequencing library is constructed and high-throughput sequencing is performed on the Illumina NovaSeq platform. The sequencing data is processed through the QIIME2 pipeline, including steps such as quality control, sequence splicing, denoising, and species annotation, and finally an OTU abundance table is generated.

[0080] For example, the extracted DNA is amplified by PCR. The amplification primers are selected as the universal primers 515F (5’-GTGCCAGCMGCCGCGGTAA-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’) targeting the V4 region of the 16S rRNA gene. The amplification reaction system and conditions are optimized. For example, the reaction system is 25 μL, containing 12.5 μL of 2×Taq PCR MasterMix, 1 μL of forward primer, 1 μL of reverse primer, 2 μL of template DNA, and 8.5 μL of sterile water; the reaction conditions are pre-denaturation at 95 °C for 3 min, then denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, extension at 72 °C for 30 s, for a total of 30 cycles, and finally extension at 72 °C for 10 min to ensure the specificity and efficiency of amplification. After the amplification products are purified, a sequencing library is constructed and high-throughput sequencing is performed on the Illumina NovaSeq platform. The sequencing data is processed through the QIIME2 pipeline, including quality control (removing low-quality sequences, primer sequences, etc.), sequence splicing, denoising, species annotation, etc., and finally an OTU abundance table is generated.

[0081] The Spearman correlation between OTUs is calculated using the Hmisc package in R language. During the calculation, the correlation coefficient threshold is set to | |>0.6, and the significance level p<0.01. OTU pairs with significant correlations are selected as the edges in the microbial co-occurrence network. An undirected weighted network is constructed using the igraph package. The significant edges and the corresponding OTU node information obtained by screening are input into the igraph package to construct the microbial co-occurrence network. During the construction process, the deletion threshold for isolated nodes with Degree = 0 is set, that is, nodes with Degree = 0 will be deleted from the network to ensure the connectivity and effectiveness of the network.

[0082] Step five, the Louvain algorithm is used to perform modularity analysis on each microbial co-occurrence network, and the modularity index of each fragmentation level group is calculated.

[0083] Specifically, the constructed microbial co-occurrence network is partitioned using the Louvain algorithm. The Louvain algorithm is a community discovery algorithm based on a greedy strategy that can quickly and effectively partition the network into different modules. During the partitioning process, through multiple iterations of optimization, the module partitioning result reaches the optimal. The Louvain algorithm is used to partition the modules and calculate the modularity index. The formula is:

[0084]

[0085] Among them, Q is the modularity index, It is the proportion of the inner edges of module i, obtained by calculating the ratio of the number of edges within module i to the total number of edges in the network; It is the edge proportion of module i, obtained by calculating the ratio of the number of edges in module i to the sum of the degrees of all nodes in the network. The larger the Q value, the stronger the modularity of the microbial network and the more stable the community structure.

[0086] The significance of the modularity index Q is evaluated through null model testing. The network is randomly reconnected 100 times while maintaining the degree distribution. Calculate the modularity Q value of the network after each reconnection. Compare these random Q values with the actually calculated Q value to calculate the significance level p value. The significance of the modularity Q value is tested by the null model. If the significance level p < 0.05, it is used to verify the reliability of the module division result.

[0087] Step six, for the mean value of the comprehensive landscape fragmentation index of all fragmentation level groups and the modularity index corresponding to each fragmentation level group, apply the locally weighted regression fitting model to construct a non - linear response curve of landscape fragmentation and modularity index at the global scale.

[0088] Specifically, based on the dataset of the mean value of the comprehensive landscape fragmentation index of all fragmentation level groups and the modularity index corresponding to the fragmentation level groups, use the locally weighted regression function in the stats package of R language to fit the non - linear relationship. The model formula is:

[0089]

[0090] where Q is the modularity index, FI is the comprehensive landscape fragmentation index, is the locally weighted polynomial function, is the error term. The initial value of the model span parameter is set to 0.75. The model span value corresponding to the minimum mean square error is optimized and selected through cross - validation. The polynomial order is fixed at the second order to better capture the non - linear relationship and possible inflection point characteristics between the landscape fragmentation index and the microbial network modularity index.

[0091] Execute the iteratively reweighted least squares method, gradually adjust the local weight function, such as the Tricube kernel function, to obtain the optimal fitting curve, which describes the response of the modularity index when the landscape fragmentation degree changes. During the model construction process, by iteratively reweighting and adjusting the local weight function, the model is continuously optimized, and finally the optimal fitting curve between landscape fragmentation and modularity index is obtained. This curve intuitively reflects the response trend of the microbial network modularity as the landscape fragmentation degree changes.

[0092] Step seven, determine the critical threshold of the comprehensive landscape fragmentation index at which the modularity significantly decreases by minimizing the Akaike information criterion value.

[0093] Specifically, in the R language environment, the segmented package is called to perform segmented regression analysis on the comprehensive landscape fragmentation index and modularity index data.

[0094] During the segmented regression analysis process, different segmentation points are continuously tried, and the Akaike Information Criterion (AIC) value of the model under each segmentation scheme is calculated. The AIC value comprehensively considers the goodness of fit and complexity of the model, and the segmentation scheme that minimizes the AIC value is selected. The segmentation point determined by this scheme is the critical threshold (FIcrit) of the comprehensive landscape fragmentation index (FI) corresponding to the significant decrease in modularity. When the landscape fragmentation index reaches or exceeds the critical threshold, the modularity of the microbial network will change significantly, indicating that the watershed ecosystem may face the risk of structural and functional changes, providing a key quantitative indicator for watershed ecological health early warning and land use management.

[0095] As Figure 2 shown, the abscissa is the degree of landscape fragmentation; the ordinate is the modularity index of the microbial network. The curve shows a decreasing trend that is first slow and then rapid, clearly demonstrating the non-linear relationship between the two. The critical threshold determined by minimizing the Akaike Information Criterion value is clearly marked in the figure, providing a key quantitative basis for watershed ecological early warning and management.

[0096] The embodiment of the present invention also provides a storage medium, in which a computer program is stored. When the computer program is executed by a processor, it implements some or all of the steps in the embodiments of the watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity provided by the present invention. The storage medium can be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), etc.

[0097] The above-described embodiments of the present invention do not constitute a limitation on the protection scope of the present invention.

Claims

1. A watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity, characterized in that: include: Step 1: Arrange sampling points according to the land use type, topographic division and river hydrological characteristics in the basin, record the longitude and latitude coordinates and collect sediment samples; Step 2: Process the satellite remote sensing images of the watershed, extract the patch density, edge density and aggregation degree based on the sampling point locations, and construct a comprehensive landscape fragmentation index for each sampling point; Constructing a comprehensive landscape fragmentation index: Among them, FI is the comprehensive landscape fragmentation index, PD is the patch density, ED is the edge density, and AI is the aggregation degree; Step 3: Standardize and preprocess the comprehensive landscape fragmentation index of all sampling points, use the elbow rule to determine the optimal number of clusters, and execute the K-means clustering algorithm to divide the sampling points into several fragmentation level groups; Step 4: The microbial species composition of all sediment samples was determined, the Spearman correlation between operational taxonomic units was calculated based on the species abundance matrix, significant correlations were screened, and the microbial symbiosis network was constructed using the R language igraph package; Step 5: The Louvain algorithm was used to perform modular analysis on each microbial symbiotic network and calculate the modularity index of each fragmentation level group; Step 6: For the mean of the comprehensive landscape fragmentation index of all fragmentation level groups and the modularity index of each fragmentation level group, a local weighted regression fitting model is applied to construct a nonlinear response curve of landscape fragmentation and modularity index at the global scale; Step seven: Determine the critical threshold of the comprehensive landscape fragmentation index at which modularity decreases significantly by minimizing the Akaike information criterion value.

2. The watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity according to claim 1, characterized in that: The step one comprises: The sampling points are rationally arranged according to the land use type, topographic division and river hydrological characteristics in the basin; Use GPS positioning equipment to record the latitude and longitude coordinates of each sampling point; Riverbed sediment samples were collected at each sampling point, and surface sediments with a depth of 0-25 cm were collected. Each sampling point was repeated at least 3 times, and the amount of sediment sample collected each time was not less than 100 g. After the surface sediment samples were fully mixed, they were stored in an environment of -20°C until transported back to the laboratory.

3. The watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity according to claim 1, characterized in that: The second step comprises: Acquire and process satellite remote sensing images of the watershed; Supervised classification of remote sensing images is performed with the help of remote sensing image processing platform; The ArcGIS platform was used to divide the rivers from the source to the sampling point, and the buffer was set at a preset distance to extract the land use data within the buffer. The processed remote sensing images were imported into Fragstats to calculate the landscape pattern index. , where PD is the patch density, N is the number of patches, and A is the landscape area; , where ED is the edge density and E is the total length of the patch edge; , where AI is the degree of aggregation, is the number of adjacent patches of the same type.

4. The watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity according to claim 3 is characterized in that: In the step of supervised classification of remote sensing images with the help of remote sensing image processing platform, the maximum likelihood method is used for land use classification, and the classification result is verified by Kappa coefficient, Kappa>0.75; the isolated node deletion threshold of network construction is set to Degree=0, that is, nodes with Degree=0 will be deleted from the network.

5. The watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity according to claim 1, characterized in that: The step three comprises: The comprehensive landscape fragmentation index calculated from all sampling points is standardized to eliminate the influence of different dimensions on the clustering results; the normalization formula is used: in, is the original FI value, is the standardized comprehensive landscape fragmentation index, and are the maximum and minimum FI values ​​of all sampling points respectively; Use the elbow rule to determine the optimal number of clusters, execute the K-means clustering algorithm within a certain range, calculate the clustering error under each k value, draw a curve of the relationship between the number of clusters k and the sum of squares within the group, and observe the inflection point of the curve. The k value corresponding to the inflection point is the optimal number of clusters; According to the determined optimal number of clusters, the K-means clustering algorithm is executed, and the standardized comprehensive landscape fragmentation index is used as input data. Through iterative calculation, each sampling point is assigned to the cluster that is most similar to its comprehensive landscape fragmentation index. Finally, all sampling points are divided into several fragmentation level groups, and the sampling points in each group have similar landscape fragmentation characteristics.

6. The watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity according to claim 1, characterized in that: The fourth step comprises: The 16S rRNA gene V4 region was sequenced using the Illumina NovaSeq platform, and the operational taxonomic unit abundance table was generated using the QIIME2 pipeline; The Hmisc package of R language was used to calculate the Spearman correlation between operational classification units of each group, and the correlation coefficient threshold was | |>0.6, significance level p<0.01, significant edges were screened; the igraph package was used to construct an undirected weighted network and isolated nodes were deleted.

7. The watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity according to claim 1, characterized in that: The step five comprises: The Leuven algorithm is used to divide the modules and calculate the modularity index. The formula is: Where Q is the modularity index, is the ratio of the inner edge of module i, is the edge ratio of module i.

8. The watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity according to claim 7, characterized in that: The step 5 evaluates the significance of the modularity index through a null model test, with a significance level of p<0.05, to verify the reliability of the module division result.

9. The watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity according to claim 1, characterized in that: The step six comprises: Based on the mean of the comprehensive landscape fragmentation index of all fragmentation level groups and the modularity index data set of the corresponding fragmentation level groups, the local weighted regression function of the stats package of R language was used to fit the nonlinear relationship. The model formula is: Among them, Q is the modularity index, FI is the comprehensive landscape fragmentation index, is a local weighted polynomial function, is the error term; The optimal fitting curve is obtained by iteratively weighting and adjusting the local weight function, which describes the response of the modularity index to changes in the degree of landscape fragmentation.

10. The watershed ecological health threshold assessment method based on landscape fragmentation and microbial network modularity according to claim 1, characterized in that: The step seven comprises: In the R language environment, the segmented package was called to perform segmented regression analysis on the comprehensive landscape fragmentation index and modularity index data; During the segmented regression process, the segmentation points of the model are continuously adjusted to find the segmentation scheme that minimizes the Akaike information criterion value; the segmentation point determined by minimizing the Akaike information criterion value is the critical threshold of the comprehensive landscape fragmentation index where the modularity decreases significantly.

Images