Blue space degradation diagnosis method and system, computer equipment and medium

Abstract

The invention provides a blue space degradation diagnosis method and system, computer equipment and a medium, and belongs to the field of ecological environment monitoring, and the method comprises the steps: obtaining a remote sensing image, climate data and human activity data of a to-be-diagnosed blue space, and extracting water body images of the to-be-diagnosed blue space at a plurality of time points through the remote sensing image; calculating a multi-class landscape form index of each grid in the water body image; clustering the water body image based on the multi-class landscape form index of each grid to obtain a blue space combination mode corresponding to each grid; inputting the climate data, the human activity data and the landscape form index into a random forest model to obtain a contribution degree; and based on the contribution degree, screening from the climate data, the human activity data and the landscape form index to obtain a water body change driving factor, and determining a degradation diagnosis result of the blue space to be diagnosed. The blue space combination mode constructed by the method can reflect the structural difference of the blue space, and the accuracy of degradation diagnosis of the blue space is guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of ecological environment monitoring, and specifically relates to a method, system, computer equipment and medium for diagnosing the degradation of blue space. Background Technology

[0002] Blue spaces refer to natural or semi-natural ecosystems centered on water bodies, such as rivers, lakes, and wetlands. As key carriers of the water cycle and important habitats for biodiversity, their morphology, structure, and dynamic changes directly affect water resource sustainability, water quality safety, and regional ecological resilience. However, under the dual pressures of climate change and human activities, the morphological evolution of blue spaces is becoming increasingly complex, and the driving mechanisms of their advancement and retreat remain unclear. Precise diagnosis is urgently needed to reveal their dynamic patterns in order to support the formulation of adaptive water resource management strategies.

[0003] Existing methods for diagnosing blue space evolution and degradation rely on remote sensing data combined with morphological index analysis, such as calculating landscape pattern indices based on remote sensing data to quantify the morphology of blue spaces. However, water bodies themselves possess highly similar spectral and textural characteristics, making it difficult to classify water cover types based on single water body indices. Furthermore, different water cover types exhibit significantly different responses to the impacts of climate and human activities, thus limiting the diagnosis of blue space evolution and degradation. Summary of the Invention

[0004] To address the existing problems, this invention provides a blue space degradation diagnosis method, system, computer equipment, and medium.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for diagnosing blue space evolution and degradation includes: Acquire remote sensing images, climate data, and human activity data of the blue space to be diagnosed, and extract water body images at multiple time points from the time series of remote sensing images; The water body image is divided into multiple grids, and the multi-class landscape morphology index of each grid in the water body image is calculated. Based on the multi-class landscape morphology index of each grid, the water body image is clustered, and the clustering category of each grid is used as the blue space combination pattern corresponding to that grid. Climate data, human activity data, landscape morphology index, and area time series data corresponding to blue space combination patterns are input into a pre-trained random forest model to obtain the contribution of climate data, human activity data, and landscape morphology index to area time series data. The area time series data is determined based on the area of ​​blue space combination patterns in water images at different time points. Based on contribution, water body change driving factors are screened from climate data, human activity data, and landscape morphology indices; the degradation diagnosis results of the blue space to be diagnosed are determined by the changing trend of area time series data and the relationship between area time series data and critical threshold, wherein the critical threshold is determined based on the relationship between water body change driving factors and area time series data.

[0006] The present invention also provides a blue space evolution / degradation diagnostic system, comprising: The image acquisition module is used to acquire remote sensing images, climate data, and human activity data of the blue space to be diagnosed, and to extract water body images at multiple time points from the time series of remote sensing images. The water body clustering module is used to divide water body images into multiple grids, calculate multiple landscape morphology indices for each grid in the water body image, and cluster the water body image based on the multiple landscape morphology indices for each grid, and use the clustering category of each grid as the blue space combination pattern corresponding to that grid. The contribution calculation module is used to input climate data, human activity data, landscape morphology index, and area time series data corresponding to blue space combination patterns into a pre-trained random forest model to obtain the contribution of climate data, human activity data, and landscape morphology index to area time series data. The area time series data is determined based on the area of ​​blue space combination patterns in water images at different time points. The degradation diagnosis module is used to screen water body change drivers from climate data, human activity data, and landscape morphology indices based on contribution. The degradation diagnosis results of the blue space to be diagnosed are determined by the trend of area time series data and the relationship between area time series data and critical threshold. The critical threshold is determined based on the relationship between water body change drivers and area time series data.

[0007] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any of the steps in a blue space evolution degradation diagnostic method.

[0008] The present invention also provides a computer-readable storage medium storing a computer program that, when loaded by a processor, can execute any of the steps in a blue space evolution / degeneration diagnostic method.

[0009] The blue space evolution / degeneration diagnostic method provided by this invention has the following beneficial effects: The blue space evolution and degradation diagnosis method provided by this invention acquires remote sensing, climate, and human activity data of blue spaces, extracts multi-time-point water body images, and calculates landscape morphology indices for grids, thereby clustering blue space combination patterns. Subsequently, a random forest model is used to analyze the contribution of various data to the area change of the combination patterns, screen key driving factors, and finally diagnose the evolution and degradation status of blue spaces by comparing the trend of area time series changes with critical thresholds set based on driving factors. Because this invention can comprehensively calculate multiple types of landscape morphology indices and perform cluster analysis to obtain various different blue space combination patterns, it can accurately reflect the structural differences within blue spaces in terms of fragmentation, connectivity, and shape complexity, thereby effectively distinguishing highly similar water cover types. Based on this, contribution prediction and critical threshold determination are performed, and the final diagnostic results fully consider the differences in the influence of environment and human activities on different blue space combination patterns, ensuring the accuracy of blue space evolution and degradation diagnosis. Attached Figure Description

[0010] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of a blue space evolution / degeneration diagnosis method provided in an embodiment of the present invention; Figure 2 This is one example of a blue space evolution / degeneration diagnostic process provided in an embodiment of the present invention; Figure 3 This is a second example of a blue space evolution / degeneration diagnostic process provided in an embodiment of the present invention; Figure 4 This is the third example of a blue space evolution / degeneration diagnostic process provided in this embodiment of the invention. Figure 5 This is the fourth example of a blue space evolution / degeneration diagnostic process provided in this embodiment of the invention. Figure 6 This is an example of a blue spatial distribution map over a long time series of a freshwater lake provided in an embodiment of the present invention; Figure 7 This is an example of the probability density distribution of a freshwater lake landscape morphology index provided in an embodiment of the present invention; Figure 8 This is an example of the ranking results of the contribution values ​​of a certain freshwater lake landscape morphology index to the classification of blue space provided in an embodiment of the present invention; Figure 9 This is an example of five types of blue space and their index characteristics in a freshwater lake provided in an embodiment of the present invention; Figure 10 This is an example of blue space morphology type clustering results provided in an embodiment of the present invention; Figure 11 Examples of different spatiotemporal distribution characteristics of blue space from 1986 to 2024 provided for embodiments of the present invention; Figure 12 This is an example of ranking the importance of evolutionary periods as provided in an embodiment of the present invention; Figure 13 This is an example of ranking the importance of degradation period in an embodiment of the present invention; Figure 14 Examples of the differences in driving forces due to factors influencing evolutionary stages provided in embodiments of the present invention; Figure 15 This is an example of the difference in driving forces caused by factors affecting the degradation period, provided in an embodiment of the present invention. Figure 16 This is an example of a partial dependence analysis of the influence of driving indicators at different times on the evolution and degradation of blue space, provided in an embodiment of the present invention. Figure 17 Examples of distribution characteristics and morphological zoning for different blue space clusters provided in embodiments of the present invention. Detailed Implementation

[0012] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0013] In response to the need for diagnostics on the evolution or degradation of blue spaces, existing research on blue space morphology has failed to consider that water bodies themselves have similar spectral and textural characteristics. Furthermore, it lacks quantitative assessment of the driving factors in the long-term series changes of blue spaces and dynamic analysis of the contribution of driving factors such as climate warming and human intervention. This makes it difficult to reveal the differentiated mechanisms by which morphological remodeling affects water degradation and management thresholds.

[0014] Specifically, methods for extracting blue space using water body indices based on remote sensing are relatively limited. The blue space encompasses a complex array of water body types, and selecting more appropriate methods and morphological indices for extraction is a key challenge. Currently, most studies employ single water body indices such as MNDVI and NDVI, along with edge and shape features, through deep learning to extract water bodies. However, due to the highly similar spectral and textural characteristics of water bodies, these methods struggle to directly classify water body cover types, resulting in insufficient quantification of blue space morphology. Furthermore, while blue space exhibits significant seasonal and interannual dynamic variations, existing research, both domestically and internationally, largely focuses on short-term cross-sections, such as analyses of a year or a specific season, limited by the availability and processing costs of long-term, high-resolution imagery. It lacks systematic monitoring of long-term evolutionary processes exceeding 20 years, making it difficult to reveal the long-term patterns and trends of its morphological evolution.

[0015] Furthermore, existing blue space classification systems suffer from deficiencies such as insufficient consideration of morphology and fragmented ecological functions, hindering the accurate analysis of morphology-effect relationships. Numerous studies indicate that many classifications originate from the overall blue-green space framework, such as treating the entire blue space as a single part of the green space for holistic research, or classifying it based on its inundation frequency or fragmentation type. The blue space classification system lacks a systematic consideration of morphological dimensions such as spatial configuration and landscape pattern, making it difficult to effectively support morphological evolution analysis.

[0016] To address the aforementioned issues, some studies have applied K-means to morphological classification. This method uses morphological features to cluster and can classify morphologies into multiple categories. However, the relationship between morphological features and influencing factors is usually non-linear, and multicollinearity among variables can complicate the interpretation of the results.

[0017] The blue space evolution and degradation diagnosis method provided by this invention combines GEE remote sensing observation, selects multiple landscape spatial morphology indices, and uses Kmeans and XGBoost machine learning methods for aggregation, classification and identification of blue space types. It analyzes blue space information over long time series, thereby improving the accuracy of blue space evolution and degradation diagnosis.

[0018] Example 1 This invention provides a method for diagnosing blue space evolution and degradation, specifically as follows: Figure 1 As shown, it includes the following steps: Step 1: Acquire remote sensing images, climate data, and human activity data of the blue space to be diagnosed, and extract water body images at multiple time points from the time series of remote sensing images. Specifically, water bodies are identified pixel by pixel from the time series remote sensing images to obtain water body images of the blue space to be diagnosed at different time points based on the MNDWI index.

[0019] Specifically, taking a freshwater lake area as an example, this invention introduces the method for diagnosing the evolution and degradation of blue space. This freshwater lake exhibits rich blue space morphology and has experienced dramatic water changes in recent years, making the diagnosis of blue space evolution and degradation typical. The core area of ​​this region covers 2579.2 km². 2 It has important functions such as flood regulation and storage. The blue spatial morphology of this freshwater lake area has changed from being concentrated to gradually dispersed into various watersheds in recent years. With the evolution of the lake's hydrological and morphological characteristics, the core lake has gradually become fragmented. Simultaneously, due to human activities and climate change, the lake's water resources have experienced multifaceted ecological degradation.

[0020] First, such as Figure 2 As shown, the Google Earth Engine (GEE) platform was used to extract blue space information to provide blue space data for evolution analysis. Specifically, multi-source remote sensing data covering the study area with less than 10% cloud cover from May 1, 1986 to October 30, 2024 were first acquired as the data source, with a resolution of 30m. The multi-source remote sensing data underwent radiometric correction, geometric correction, and cloud removal to select image maps for diagnosis. Water body images were extracted from the impact map using the MNDWI index, and data were extracted monthly, with data from months with more than 6 months of water as the base data. For some years, the remote sensing data was obscured by cloud cover and difficult to use for evolution / degradation diagnosis; in these cases, Landsat 5 imagery was used instead of remote sensing data.

[0021] Next, the accuracy of the aforementioned basic data was verified through both accuracy and visual comparison. For example, through visual interpretation of Landsat images, sample points were randomly selected and categorized into non-water bodies and water bodies; then, a confusion matrix was constructed using the water body extraction results from the corresponding time period. The final accuracy evaluation results were analyzed using overall accuracy and kappa coefficient. The overall accuracy of the freshwater lake's water body image was 96%, and the kappa coefficient was 0.95, making it suitable for subsequent degradation and evolution diagnosis. Furthermore, the extracted water body image could be compared with the JRC (Journal of the Monthly Water History, v1.4) global surface generation layer to select the clearer water body image.

[0022] Step 2: Divide the water image into multiple grids and calculate the various landscape morphology indices for each grid. Cluster the water image based on these indices, and use the cluster category of each grid as the corresponding blue space combination pattern. The landscape morphology indices include patch density, Shannon diversity index, fractal dimension index, maximum patch index, landscape shape index, patch assemblage index, spread index, and edge density. Specifically, K-means clustering is performed on each grid in the water image based on the Euclidean distance of the various landscape morphology indices to obtain the corresponding blue space combination pattern. The grid scale for K-means clustering is determined by the relative scale variation of the statistical characteristics of the landscape morphology indices, and the number of clusters in K-means clustering is determined by the elbow method based on the differences between grids.

[0023] Specifically, after extracting the water images of the freshwater lake area, such as Figure 3 As shown, based on the morphological and evolutionary characteristics of this freshwater lake area, structural parameter indicators or multiple landscape morphology indices are constructed to quantitatively describe the blue space morphology of this freshwater lake. Among them, the landscape morphology indices include eight categories: Patch Density (PD), Shannon's Diversity Index (SHDI), Fractal Dimension Index (FRAC), Largest Patch Index (LPI), Landscape Shape Index (LSI), Patch Cohesion Index (COHESION), Contagion Index (CONTAG), and Edge Density (ED). Their definitions are shown in Table 1.

[0024] Table 1 Definition and Calculation of Landscape Morphology Index Subsequently, the K-means clustering algorithm was used to cluster each index based on the morphological index characteristics of the blue space in the freshwater lake region, dividing the blue space into five blue spaces with significant differences in characteristics. K-means clustering uses the square of the Euclidean distance as the minimization objective. The clustering grid scale can be pre-determined by qualitatively comparing different numbers of clusters and grid resolutions, using the changes in the mean, standard deviation, and coefficient of variation relative to the grid scale to determine the minimum critical length of the grid cell. For example, for this freshwater lake region, the index variation is at its lowest and remains stable at a grid scale of 500m, while also including relevant statistical information from morphological analysis; therefore, a grid scale of 500m was chosen.

[0025] The choice of clustering data can be determined experimentally at different spatial scales by experimenting with varying numbers of clusters above the critical length. For this freshwater lake region, the elbow strike method or elbow method was used to determine the clustering data. When the K value reached 5 or 6, there were no significant differences between categories, and no major differences between grids. In summary, for this freshwater lake region, using 5 clusters on a 500m grid showed the most effective performance. This configuration strikes a balance between granularity and comprehensiveness, and can describe the combination patterns of different blue spatial morphologies.

[0026] Step 3: Using a pre-trained XGBoost model, determine the water type probability of each grid based on the multi-landscape morphology index of each grid; analyze the contribution of the multi-landscape morphology index to the water type probability using SHAP; re-divide the biased grid according to the contribution of the multi-landscape morphology index to the water type probability to obtain the corrected blue space combination pattern corresponding to each grid, wherein the grid includes the biased grid, and the difference between the water type probability of the biased grid and the blue space combination pattern is higher than the preset value.

[0027] Specifically, after the K-means clustering algorithm divides the space into five blue spaces, the XGBoost model can be used to predict the proportion of blue space grids based on multiple landscape morphology indices, identify the main influencing landscape indices for each grid, and thus construct a blue space classification framework.

[0028] For example, based on the grid proportion, the original blue space is first divided into a pure blue space and a mixed blue space. The mixed blue space is determined to be 1 in the XGBoost prediction value, and the pure blue space is determined to be 0. The contribution values ​​of different exponents during the evolution of the blue space are calculated. The XGBoost model processing principle is shown in formula (1):

[0029] (1) in, It is the loss function term. This represents the total number of training samples and defines the range of the summation. It is the index of the sample, used to refer to each training sample one by one. , It is its true value or target variable, which is the standard answer that the model needs to learn. This is the model's prediction for that sample. (Function) Calculate the predicted value Compared with the true value The error between them is calculated by the formula, which is the sum of the prediction errors of the model on all training samples. One of the core tasks of model training is to minimize this sum.

[0030] It is a regularization term. This represents the total number of decision trees in the model, which is also the number of iterations in the model construction. k It is the index of the tree, used to traverse each individual tree. The first in the reference sequence A decision tree or base learner, function It is the complexity function of a tree; it is applied to a single tree. The evaluation typically considers both the number of leaf nodes in the tree and the weight of each leaf node. This is done by evaluating all... The formula sums the complexity of each tree, providing a quantifiable penalty for the overall structural complexity of the model. Minimizing this term effectively constrains the model, preventing it from mechanically memorizing noise from the training data due to excessive structural complexity, thereby improving the model's robustness in predicting new data.

[0031] Furthermore, since hyperparameters are crucial to model performance, the XGBoost model hyperparameters in this embodiment mainly consider three factors: maximum depth, learning rate, and number of trees. Through experiments, the learning rate (learning_rate) was set to [0.01, 0.05], the number of trees (n_estimators) was set to [100, 200, 300], and the maximum tree depth (max_depth) was set to [3, 5, 7]. At this point, the model accuracy reached over 90%, providing optimal prediction performance under acceptable model complexity.

[0032] After the XGBoost model completes the type classification of each grid in the water image, the contribution of eight landscape morphology indices to the model prediction is measured by the SHAP value. The K-means clustering results are then corrected based on the difference between the XGBoost model prediction results and the K-means clustering results. The contribution is calculated as shown in formula (2):

[0033] (2) in, This represents the predicted output of the explanatory model and is an approximate function used to simulate the prediction behavior of the original complex model for a specific sample; parameters It is the baseline or expected value, representing the average level of the model's predictions for all samples, i.e., the model output "when there is no feature information". The total number of features determines the upper limit of the summation; It is the index of the feature, used to traverse from the 1st to the 2nd. For each feature; , It is its corresponding SHAP value, which quantifies the prediction result of this feature for the current sample relative to the baseline value. ， The specific magnitude and direction of the contribution, for example, a positive value indicates improved prediction, and a negative value indicates decreased prediction. It is the simplified input vector of the first This component is a binary indicator variable. A value of 1 indicates that the feature "exists" or "is observed," while a value of 0 indicates that the feature is "missing" or "masked." This variable ensures that its contribution is only recorded when the feature actually participates in the prediction. Only then will they be added to the final interpretation. The core idea of ​​the entire formula is to decompose a single predicted value of a complex model into a baseline value plus the sum of the linear contributions of each feature, thereby providing an intuitive, consistent, and additive interpretation of feature importance for model prediction.

[0034] Step 4: Input climate data, human activity data, landscape morphology index, and area time series data corresponding to the blue space combination pattern into a pre-trained random forest model to obtain the contribution of climate data, human activity data, and landscape morphology index to the area time series data. The area time series data is determined based on the area of ​​the blue space combination pattern in the water body image at different time points.

[0035] Specifically, after correcting the K-means clustering results using the XGBoost model, such as Figure 4 As shown, random forest is used to calculate the importance of different driving factors in the degradation and evolution phases of Dongting Lake, analyze the key factors affecting the changes in the blue space morphology in each period, and thus clarify its threshold effect.

[0036] For example, considering that the main influencing factors of changes in blue space area include both anthropogenic and natural factors, such as precipitation directly affecting blue space area and rising temperatures leading to water evaporation, while human activities such as production and daily life have an increasingly serious impact on the shrinkage of blue space area, this embodiment uses the above indicators as possible driving factors for further analysis, as shown in Table 2:

[0037] Table 2. Driving Factor Indicator System and Data Sources Subsequently, the contribution of driving factors and model validation were performed using Random Forest (RF). This study employed the Random Forest algorithm to model the relationship between the change in blue space area and the three-dimensional driving factors. 300 trees and a depth of 10 were selected for model training and testing. Model performance was evaluated by comparing the observed progression / degradation of the blue space with the predicted progression / degradation, achieving an accuracy of 78%.

[0038] Step 5: Based on contribution, water body change driving factors are screened from climate data, human activity data, and landscape morphology indices. The degradation / progression diagnosis results of the blue space to be diagnosed are determined by the changing trend of area time series data and the relationship between area time series data and critical thresholds. The critical thresholds are determined based on the relationship between water body change driving factors and area time series data. Furthermore, the blue space combination patterns include low fragmentation-high connectivity patterns and high fragmentation-low connectivity patterns, and the changing trends of area time series data for degradation / progression diagnosis include the area change trends of low fragmentation-high connectivity patterns and high fragmentation-low connectivity patterns.

[0039] Step 5 includes: Step 5.1: Construct a partial dependency graph based on the relationship between water body change driving factors and area time series data.

[0040] Step 5.2: Identify inflection points from the partial dependency graph using the second-order difference method based on curvature changes, and determine the critical threshold from the inflection points.

[0041] Specifically, after determining the contribution of each of the above driving factors, considering that the predicted value can be obtained by sequentially modifying the value of the target feature, thereby reflecting the marginal effect of one or two features, this embodiment can also generate a partial dependency graph (PDP) for each driving factor, thereby clarifying the threshold effect of these driving factors on the evolution and degradation of the blue space.

[0042] For example, the calculation of a partial dependency graph (PDP) is shown in Equation (3): (3) in, This represents a partial dependency function, which is a target feature. The function; The target feature is the specific value of the feature whose influence this invention aims to analyze. It's about other features. The expected operator, This indicates the expectation of all features except the target feature; It is the prediction function of the original machine learning model; This represents the input to the prediction function, containing the target feature values. Other features ; In a specific and The model predicts the value of the selected driver and the predicted evolution of the blue space by keeping other driving factors constant in the machine learning model and showing the relationship between the selected driver and the predicted evolution of the blue space through a partial dependency graph (PDP). This reveals the linear and nonlinear relationship between the driver and the area change of the blue space. For example, a straight line represents a linear relationship and a curve represents a nonlinear relationship.

[0043] Subsequently, a second-order difference method based on curvature changes was used to identify key inflection points in the partially dependent graph (PDP). Specifically, the intensity of the function curvature change was quantified by calculating the absolute value of the second-order difference of the PDP curve. Its extreme points correspond to significant inflection points in the influence of features on the stability of the blue space. This threshold, along with factors such as the current area change and trend of the blue region, was then used to determine the evolutionary or degenerate state of the blue space. Furthermore, as... Figure 5 As shown, based on the evolution and degradation states of the blue space, corresponding partitioning strategies are specified for the blue space of different partitions.

[0044] Example 2 Based on Example 1, further analysis and verification were conducted using remote sensing data of the freshwater lake.

[0045] Among them, the blue spatial morphology and landscape index probability density distribution of the freshwater lake imagery collected based on the GEE platform are as follows: Figure 3 As shown. Among them, Figure 6 This is a long-term time series blue space distribution map of the freshwater lake. The blue space morphology in the center of the freshwater lake changes significantly, showing a state of first decreasing and then increasing. At the same time, the overall blue space morphology shows a tendency to move from aggregation to fragmentation. Figure 7The probability density distribution of different landscape morphology indices for this freshwater lake shows that CONTAG, ED, FRAC, LSI, SHDI, and PD exhibit the same trend, while LPI and COHESION show the same trend. SHDI and PD values ​​are concentrated at small values, and the overall blue area shows a low degree of fragmentation. Figure 8 The ranking of the contribution values ​​of eight landscape morphology indices corresponding to this freshwater lake to the classification of blue space is as follows: Specifically, XGBoost-SHAP prediction was used to analyze the contribution values ​​of the above eight indices. PD had the largest SHAP value, accounting for 30.9%, indicating that it had the most significant impact on blue space types. This was followed by FRAC and LPI, accounting for 30.2% and 28.5% respectively. This indicates that these three landscape indices are the most important for the classification of mixed blue spaces.

[0046] The index characteristics of the five blue spatial domains of this freshwater lake based on K-means clustering are as follows: Figure 9 As shown, where, Figure 9 (a) shows the distribution and proportion of different types of blue space. Figure 9 (b) represents the blue space index features. The clustering results and definitions of blue space morphology types are as follows: Figure 10 As shown. In this embodiment, based on fragmentation, shape complexity, and connectivity, five categories are defined: Low Fragmentation - Low Shape Complexity - High Connectivity (LLH), Low Fragmentation - Medium Shape Complexity - High Connectivity (LMH), Medium Fragmentation - Medium Shape Complexity - Medium Connectivity (MMM), High Fragmentation - Medium Shape Complexity - Low Connectivity (HML), and High Fragmentation - High Shape Complexity - Low Connectivity (HHL). Furthermore, the overall service level for each blue space type is determined by measuring the integral values ​​of PD, FRAC, and LPI, i.e., the distance from the cluster centroid to the origin in the cluster coordinate system. For example... Figure 10As shown, the five blue spatial forms exhibit significant differences in landscape pattern and ecological function: the LLH type is dominated by large water bodies with high connectivity, mainly corresponding to regional lakes and main rivers, and undertakes the core functions of hydrological regulation and ecological corridors at the watershed scale; the LMH type has the characteristics of a combination of regular main body and complex edge, forming a network structure with large rivers as the skeleton and small lakes as nodes, which significantly enhances the water connectivity efficiency and ecological process integrity; the MMM type presents a balanced landscape pattern, as a composite system of wetlands and small streams, continuously playing a role in water purification and local climate regulation through multiple ecological interface processes; the HML type is a moderately fragmented swamp-pond system, maintaining microbial community diversity at the landscape scale, and realizing local water circulation and ecological regulation through micro-hydrological units; the HHL type has highly fragmented structural characteristics, as a typical representative of decentralized water systems, constructing a multifunctional system of hydrological regulation and ecological service synergy in the regional ecological pattern. The top three types contribute over 60% and are the dominant types. These five types of blue spaces together constitute a complementary and structurally coordinated aquatic ecological system, providing a scientific basis for implementing precise ecological management by zone and category.

[0047] Subsequently, the spatiotemporal distribution of these five types of blue spaces in the freshwater lake from 1986 to 2024 is as follows: Figure 11 As shown, where Figure 11 (a) shows the interannual variation of the largest water area from 1986 to 2024, including the temporal distribution of these five types of blue spaces, while Figure 11 (b) delineates the specific periods of evolution and regression of blue space, encompassing the spatial distribution of these five types of blue space. Figure 11 It can be seen that the different blue space morphology types are consistent with the overall water body change trend, with the area change showing an initial increase followed by a decrease. However, there are differences among the different types. Specifically, in terms of the annual water body area share, the trend is basically decreasing from LLH to HHL. LLH has undergone the greatest evolution, from 31% in 1986 to 24% in 2024. The share of LMH, MMM, and HML has changed relatively little, generally fluctuating around 20%, indicating that these three types are relatively stable. HHL has the smallest share, around 13% annually, and most of it is converted from LLH. The year with the largest LLH share was 1988, when the overall water body area was the second largest; while the shares of the other types have not changed much compared to other years, indicating that LLH is the main type of water body area increase; the periods when LMH and MMM had the largest shares actually occurred when the overall area was smaller, indicating that these two types play a major role when the water body area decreases. Figure 11As shown in (b), during the period of maximum water area, large areas of lakes interconnected, forming a large blue space each year that almost covered the entire Dongting Lake area, but with significant interannual variation. Gradually, it progressed from covering the entire lake surface to later covering the main shipping channels. It also shows that LLH, LMH, and MMM are its main evolutionary types, while HML exhibits a fragmented distribution.

[0048] After that, as Figure 12 and Figure 13 As shown, random forest was used to analyze the importance or contribution rate of driving factors to different blue space areas of the freshwater lake at different times, and the driving factors were ranked based on their importance or contribution rate. Figure 12 This is a ranking of different driving factors during the evolutionary period or the period of water volume increase of this freshwater lake. Figure 13 This presents the ranking results of different driving factors during the degradation or water reduction periods of the freshwater lake. Specifically, this embodiment categorizes driving factors into landscape pattern drivers, anthropogenic drivers, and natural drivers based on their effects. Per capita GDP, population size (POP), average annual temperature (T), average annual precipitation (P), and average annual evapotranspiration (ET) are selected as anthropogenic and natural driving factors; landscape indices PD, SHDI, ED, COHESION, CONTAG, LSI, LPI, and FRAC are used as landscape index driving factors. Random forest is used to calculate the contribution and driving force differences of each type of influencing factor.

[0049] like Figure 12 As shown, in the evolutionary stage, 47.7% of natural factors > 47.1% of landscape pattern drivers > 5.2% of anthropogenic factors. Among the main influencing factors, P and T are the primary natural factors, while SHDI, PD, and FRAC are the primary landscape pattern indices. The main influencing factors for different blue space types are generally consistent with the overall trend, and also include CONTAG. Besides HHL, its main influencing factors are GDP and POP. For example... Figure 13 As shown, during the degradation stage, 73.9% of natural factors > 17.7% of anthropogenic factors > 8.3% of landscape pattern drivers. Among the main influencing factors, P holds an absolute advantage, followed by T. The main factors for different blue spaces are P, T, ET, GDP, and the landscape indices CONTAG and LP.

[0050] The differences in the driving forces of different blue spaces due to influencing factors at different times are as follows: Figure 14 and Figure 15 As shown, where Figure 14 The differences in the driving forces of different blue spaces due to factors influencing evolutionary periods. Figure 15 The differences in the driving forces of different blue spaces due to factors influencing the degradation period. For example... Figure 14 As shown, during the evolutionary period, the increase in the total area of ​​the blue space is positively correlated with all five types. Among the main influencing factors, it is significantly positively correlated with P and significantly negatively correlated with T. In different blue spaces, it shows a significant positive correlation with P, with the strongest correlations observed in LMH, MMM, and HML. It shows a negative correlation with T, with the strongest correlation observed in LMH. Figure 15 As shown, during the degradation period, the decrease in the total area of ​​blue space is also positively correlated with all five types, while among the main influencing factors, it is negatively correlated with POP and GDP. Among the different blue space types, HHL shows a significant negative correlation with LPI and CONTAG.

[0051] Subsequently, considering the significant temporal heterogeneity of the impact of different landscape indices on the evolution of blue space, this embodiment employs a partial dependency graph method to reveal the nonlinear characteristics and marginal effects of key influencing factors on the evolution and degradation of blue space, such as... Figure 16 As shown, where Figure 16 (a) is a partial dependence analysis of the influence of evolutionary period driving indicators on the evolution and degradation of blue space. Figure 16 (b) shows the partial dependence analysis of the influence of driving indicators on the evolution and degradation of blue space during the degradation period. Specifically, during the evolution period, SHDI showed a stable ecological threshold effect in all five blue space types, with its optimal threshold concentrated around 0.04, which can be used as a unified ecological management benchmark; CONTAG only showed a significant nonlinear response in HHL, with a critical threshold of 7.63; FRAC is a key regulatory indicator of HML, and it can be reduced to 0 to optimize the spatial pattern; P mainly affects the evolutionary process of LLH, LMH, and MMM, with an optimal range of 30.55-296.55; PD, as the core influencing factor of LMH, has an ecologically optimal threshold of 0.35. During the degradation period, CONTAG exhibits a significant nonlinear relationship with all blue space types, with a risk control range of 71.22–78.33; ET has a threshold effect on MMM and HML, with a safe operating window of 36.67–830.27; LPI shows nonlinear characteristics in all types, with a risk threshold range of 43.55–46.03; P significantly affects the degradation process of LLH, LMH, and MMM, and the threshold needs to be controlled below 863.74. In contrast, T shows a linear relationship with all five blue space types, and this stable influence mechanism distinguishes it from the nonlinear response characteristics of other indices.

[0052] Based on the above analysis, it can be determined that different blue space types exhibit significant differences in characteristic indices, confirming the scientific validity of the classification system. The index distribution of the five blue space types shows a clear gradient change. This embodiment, building upon existing research, overcomes the limitations of single-index methods and uses K-means to classify blue spaces by integrating multi-dimensional indicators. The final results verify the effectiveness of this method.

[0053] Furthermore, in terms of diagnostic model construction, this invention achieves high-precision classification and strong interpretability. Dividing the period from 1986 to 2024 into two stages—area increase and decrease—facilitates a deeper identification of the driving factors of evolution. Water body extraction results based on the GEE platform achieved an accuracy of 96% after verification using a confusion matrix. The introduction of the XGBoost prediction model in the classification interpretation demonstrates a classification prediction accuracy exceeding 90%, reflecting the scientific rigor of this framework. Combined with the SHAP interpretation method, the contribution of each morphological factor is clearly presented, enhancing the understanding of spatial evolution mechanisms. This analytical framework integrating machine learning provides new insights for the zoning and planning of blue space.

[0054] Furthermore, the blue space evolution and degradation diagnosis method provided by this invention can assist in the zoning and management of blue space water resources, such as... Figure 17 The distribution characteristics and morphological zoning of different blue spatial clusters in this freshwater lake area are shown in Table 3. The zoning and threshold control measures are shown in Table 3. Table 3. Zoning strategy and threshold control measures for blue space based on morphological changes In the core water conservation area dominated by the LLH type, this zone constructs a core for watershed hydrological regulation by maintaining extremely low landscape diversity and a dominant pattern of large water bodies. Its macroscopic structure forms the basis for ensuring hydrological functions: the construction of ecological water replenishment hubs aims to stabilize the dynamic changes in actual inflow precipitation within a suitable threshold range of 30.55-863.74, achieving cross-temporal water allocation; the deployment of automatic water level monitoring stations provides support for dynamic scheduling based on real-time hydrological data, ensuring the stability of core functions.

[0055] For corridor connectivity transition zones dominated by the LMH type, maintaining extremely low patch density and high connectivity ensures efficient water system connectivity. The management logic centers on maintaining and optimizing connectivity: implementing shoreline naturalization and river meandering modifications fundamentally improve connectivity and land-water interaction efficiency; setting up connectivity monitoring sections and maintaining them regularly to directly monitor and safeguard connectivity indicators, ensuring the corridor's continuous and efficient function.

[0056] For the MMM-type resilient functional transition zone, maintaining system resilience is achieved by preserving a strict water baseline, a wide-ranging landscape structure, and strong climate regulation capabilities. Constructing a multi-functional reservoir system and ecological aquaculture facilities is key to the redistribution of water and energy; this system effectively stores precipitation, ensuring its utilization and purification, and maintaining total precipitation variations within a reasonable range of 30.55-863.74. Simultaneously, through surface evaporation, the system stabilizes evapotranspiration variations related to its climate regulation functions within the target range of 36.67-830.27. A distributed hydrological monitoring network is deployed to monitor P-ET balance, ensuring the system dynamics remain within a safe range.

[0057] For attenuation buffer remediation zones primarily of the HML type, an efficient engineering system is constructed by shaping patches with highly regular shapes and uniform scale. The remediation measures aim to precisely optimize the structure: constructing stormwater retention modules and connecting pipeline systems to the pond clusters; functionally transforming and integrating abandoned ponds to form a cluster of efficient functional units, thereby achieving precise control over the attenuation buffer function.

[0058] For fragmented and degradation-sensitive areas dominated by the HHL type, the measures aim to guide their evolution from highly fragmented to functionally networked and stable by increasing the spread from the current low value to the target range. The measures emphasize precision and networking: deploying intelligent water replenishment systems and constructing micro-constructed wetland clusters to improve the quality of key small patches and enhance their functional connectivity, thus building a distributed purification system; and establishing a space-ground collaborative monitoring system to provide dynamic sensing capabilities for identifying degradation risks and tracking system evolution.

[0059] This invention provides a diagnostic method for the evolution and degradation of blue space, integrating the GEE cloud platform, K-means clustering, and the XGBoost machine learning model to achieve accurate multi-dimensional, long-term time-series diagnosis of blue space morphological evolution. Specifically, through machine learning, five spatial types—LLH, LMH, MMM, HML, and HHL—were successfully identified, with the first three being the dominant types, contributing over 60% cumulatively. This classification framework effectively reveals the differences in the morphological structure of blue space. Furthermore, it clarifies that different types of blue space play different roles in watershed hydrological processes. For example, the LLH type, with a peak proportion of 31%, is the main contributor to water area expansion, while the LMH and MMM types dominate during the area shrinkage period, and the HHL type exhibits a continuous fragmentation trend. It also clarifies that the evolutionary stage is jointly dominated by 47.7% natural factors and 47.1% landscape patterns, while the degradation stage is mainly driven by 73.9% natural factors, with precipitation playing a particularly prominent role. More importantly, this invention is the first to quantify the critical threshold range of the impact of landscape indices such as SHDI, PD, and CONTAG on stability, providing a quantifiable scientific basis for the precise management and risk warning of blue spaces.

[0060] In summary, the blue space evolution and degradation diagnosis method provided by this invention not only deepens the understanding of the blue space evolution mechanism, but also provides important theoretical support and practical guidance for implementing integrated management of water resources SDGs and formulating water environment restoration strategies based on natural solutions (NbS).

[0061] Example 3 The present invention also provides a blue space evolution / degradation diagnostic system, comprising: The image acquisition module is used to acquire remote sensing images, climate data, and human activity data of the blue space to be diagnosed, and to extract water body images at multiple time points from the time series of the remote sensing images. The water body clustering module is used to divide the water body image into multiple grids, calculate the multi-class landscape morphology index of each grid in the water body image, and cluster the water body image based on the multi-class landscape morphology index of each grid, and use the clustering category of each grid as the blue space combination pattern corresponding to that grid. The contribution calculation module is used to input the climate data, human activity data, landscape morphology index, and area time series data corresponding to the blue space combination pattern into a pre-trained random forest model to obtain the contribution of the climate data, human activity data, and landscape morphology index to the area time series data. The area time series data is determined based on the area of ​​the blue space combination pattern in the water body image at different time points. The degradation diagnosis module is used to screen water body change driving factors from the climate data, human activity data, and landscape morphology index based on the contribution degree; and to determine the degradation diagnosis result of the blue space to be diagnosed based on the changing trend of the area time series data and the relationship between the area time series data and the critical threshold, wherein the critical threshold is determined based on the relationship between the water body change driving factors and the area time series data.

[0062] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps in an embodiment of a blue space degradation diagnosis method. Specific implementation methods can be found in the method embodiments, and will not be repeated here.

[0063] Furthermore, the present invention also provides a non-transitory computer-readable storage medium containing instructions on which a computer program is stored. For example, a memory containing instructions that can be executed by a processor of a computer device to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc. When the computer program is executed by the processor, it can implement the steps in an embodiment of a blue space degradation diagnosis method. Specific implementation methods can be found in the method embodiments, which will not be repeated here.

[0064] Those skilled in the art will understand that embodiments of the present invention can provide methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0065] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0066] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0067] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0068] It should be noted that the specific embodiments described above enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although the present invention has been described in detail in this specification and embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention; and all technical solutions and improvements that do not depart from the spirit and scope of the present invention are covered within the protection scope of the present invention patent. No reference numerals in the claims should be construed as limiting the scope of the claims. Any simple variations or equivalent substitutions of technical solutions that can be readily obtained by those skilled in the art within the scope of the technology disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A method for diagnosing blue space evolution and degradation, characterized in that, include: Acquire remote sensing images, climate data, and human activity data of the blue space to be diagnosed, and extract water body images at multiple time points from the time series of the remote sensing images; The water body image is divided into multiple grids, and multiple landscape morphology indices are calculated for each grid in the water body image. The water body image is clustered based on the multiple landscape morphology indices of each grid, and the clustering category of each grid is used as the blue space combination pattern corresponding to that grid. The climate data, human activity data, landscape morphology index, and area time series data corresponding to the blue space combination pattern are input into a pre-trained random forest model to obtain the contribution of the climate data, human activity data, and landscape morphology index to the area time series data. The area time series data is determined based on the area of ​​the blue space combination pattern in the water body image at different time points. Based on the contribution, water body change driving factors are screened from the climate data, human activity data, and landscape morphology index; the degradation diagnosis result of the blue space to be diagnosed is determined by the change trend of the area time series data and the relationship between the area time series data and the critical threshold, wherein the critical threshold is determined based on the relationship between the water body change driving factors and the area time series data.

2. The blue space evolution / degradation diagnostic method according to claim 1, characterized in that, Extracting water body images at multiple time points from the time series of the remote sensing images includes: Water bodies are identified pixel by pixel from the time series of the remote sensing images based on the MNDWI index, resulting in water body images at different time points.

3. The blue space evolution / degradation diagnostic method according to claim 1, characterized in that, The landscape morphology indices include patch density, Shannon diversity index, fractal dimension index, maximum patch index, landscape shape index, patch cohesion index, spread index, and edge density.

4. The blue space evolution / degradation diagnostic method according to claim 1, characterized in that, Based on the multi-class landscape morphology index clustering of water body images for each grid, the clustering category of each grid is used as the blue space combination pattern corresponding to that grid, including: K-means clustering is performed on each grid in the water image based on the Euclidean distance of the multiple landscape morphology indices of each grid to obtain the blue space combination pattern corresponding to each grid. The grid scale of the K-means clustering is determined by the relative scale variation of the statistical characteristics of the landscape morphology index, and the number of K-means clusters is determined by the differences between grids using the elbow method.

5. The blue space evolution / degradation diagnostic method according to claim 1, characterized in that, The grid includes a biased grid, where the difference between the water body type probability and the blue space combination pattern of the biased grid is higher than a preset value. After taking the clustering category of each grid as the blue space combination pattern corresponding to that grid, the following is also included: The probability of water body type in each grid is determined by the multi-class landscape morphology index of each grid through a pre-trained XGBoost model. The contribution of various landscape morphology indices to the probability of water body type was analyzed using SHAP. The biased grid is re-divided based on the contribution of the various landscape morphology indices to the probability of the water body type, resulting in a corrected blue space combination pattern for each grid.

6. The blue space evolution / degradation diagnostic method according to claim 1, characterized in that, After identifying the driving factors of water body changes through screening, the process also includes: A partial dependency graph is constructed based on the relationship between water body change drivers and the area time series data. Inflection points are identified from the partial dependency graph using a second-order difference method based on curvature variation, and critical thresholds are determined from these inflection points.

7. The blue space evolution / degradation diagnostic method according to claim 1, characterized in that, The blue space combination patterns include low fragmentation-high connectivity patterns and high fragmentation-low connectivity patterns, and the area time series data change trends include the area change trends of low fragmentation-high connectivity patterns and high fragmentation-low connectivity patterns.

8. A blue space evolution / degradation diagnostic system, characterized in that, include: The image acquisition module is used to acquire remote sensing images, climate data, and human activity data of the blue space to be diagnosed, and to extract water body images at multiple time points from the time series of the remote sensing images. The water body clustering module is used to divide the water body image into multiple grids, calculate the multi-class landscape morphology index of each grid in the water body image, and cluster the water body image based on the multi-class landscape morphology index of each grid, and use the clustering category of each grid as the blue space combination pattern corresponding to that grid. The contribution calculation module is used to input the climate data, human activity data, landscape morphology index, and area time series data corresponding to the blue space combination pattern into a pre-trained random forest model to obtain the contribution of the climate data, human activity data, and landscape morphology index to the area time series data. The area time series data is determined based on the area of ​​the blue space combination pattern in the water body image at different time points. The degradation diagnosis module is used to screen water body change driving factors from the climate data, human activity data, and landscape morphology index based on the contribution degree; and to determine the degradation diagnosis result of the blue space to be diagnosed based on the changing trend of the area time series data and the relationship between the area time series data and the critical threshold, wherein the critical threshold is determined based on the relationship between the water body change driving factors and the area time series data.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the blue space evolution / degeneration diagnostic method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the blue space evolution / degeneration diagnostic method according to any one of claims 1 to 7.

Images