Evaluation method for evolution characteristics and evolution driving factors of space-time pattern of park green land

Abstract

The invention discloses a method for evaluating evolution characteristics and evolution driving factors of a space-time pattern of a park green land, and particularly relates to the technical field of crossing of smart cities and landscape ecology, and the method comprises the steps: obtaining multi-period remote sensing image data and planning maps, and constructing a multi-source database; interpreting the image based on a random forest supervised classification algorithm to generate a park green space distribution map, and optimizing the precision; adopting a landscape ecology theory to select a landscape pattern index to quantify space-time evolution characteristics; constructing a driving evaluation index system, and analyzing a driving mechanism by means of a multiple regression model and a space measurement model; and generating a three-level spatial pattern optimization strategy based on the result and outputting a visual decision map. According to the method, the spatial-temporal pattern evolution characteristics and driving factors of the urban park green land can be comprehensively analyzed, and a scientific basis and decision support are provided for planning, management and protection of the urban park green land.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart city and landscape ecology, more particularly, the present application relates to a method for evaluating the spatiotemporal pattern evolution characteristics and evolution driving factors of park green space. BACKGROUND

[0002] With the acceleration of urbanization, the city scale is expanding, and the spatiotemporal pattern of urban park green space, as an important part of urban ecological system, has changed significantly. Urban park green space not only can improve the urban ecological environment, alleviate the heat island effect, provide leisure and entertainment space, but also has important significance for biodiversity protection. However, in the current urban development process, park green space is facing problems such as occupation and intensification of fragmentation, which seriously affects the play of its ecological service function.

[0003] In the existing research and practice, there are certain limitations in the analysis method of spatiotemporal pattern evolution of urban park green space. Some researches only rely on a single data source, and the completeness and accuracy of the data are insufficient, which is difficult to fully reflect the true situation of park green space. In image interpretation, the traditional classification method has low precision and is easily disturbed by noise and complex ground objects, resulting in deviation in the obtained park green space information. When studying the evolution characteristics of park green space, there is a lack of systematic quantitative indicators and methods, which makes it difficult to accurately grasp the change rule of its landscape structure and composition.

[0004] For the driving factor analysis of spatiotemporal pattern evolution of park green space, previous studies often only consider a single type of factor, such as only focusing on economic or environmental factors, ignoring the comprehensive action of natural factors and human factors, and failing to fully consider the interaction between factors and the influence difference in different spatial positions. This makes the understanding of driving mechanism not deep enough, and it is difficult to provide comprehensive and accurate basis for scientific planning and effective management of urban park green space.

[0005] In the practice of urban park green space planning and management, due to the lack of in-depth analysis of spatiotemporal pattern evolution characteristics and driving mechanism, the planning scheme often lacks pertinence and scientificity, and it is difficult to effectively deal with the problems of park green space, resulting in that the effect of improving the quality of urban ecological environment is not obvious, and it is difficult to meet the demand of residents for high-quality ecological space.

[0006] Therefore, the present application provides a method for evaluating the spatiotemporal pattern evolution characteristics and evolution driving factors of park green space. SUMMARY

[0007] In order to overcome the above-mentioned defects of the prior art, the present application provides a method for evaluating the spatiotemporal pattern evolution characteristics and evolution driving factors of park green space, to solve the problems raised in the background art.

[0008] In order to achieve the above object, the present application provides the following technical scheme: a park green space spatiotemporal pattern evolution characteristic and evolution driving factor evaluation method, comprising the following steps:

[0009] S1. Obtain multi-period remote sensing image data and planning map of the target city, and construct a multi-source database containing social, economic, environmental and cultural factors;

[0010] S2. Based on the random forest supervised classification algorithm, the remote sensing image is interpreted, the spatial distribution map of the park green space in different periods is generated, and the accuracy is optimized through post-classification processing;

[0011] S3. Adopting the theory of landscape ecology, the landscape pattern index is selected at the landscape level and the type level, and the spatiotemporal evolution characteristics are quantified by using Fragstats software;

[0012] S4. Constructing a driving evaluation index system containing natural factors and human factors, analyzing the driving mechanism by using multiple regression model and spatial econometric model;

[0013] S5. Based on the evolution characteristics and driving mechanism, the "macro-micro-micro" three-level spatial pattern optimization strategy is generated and the visual decision graph is output.

[0014] Preferably, in the step S2, the random forest supervised classification is realized by using ENVI5.6 platform, and the post-classification processing includes:

[0015] The accuracy is verified by calculating the Kappa coefficient through the confusion matrix, and the threshold is greater than or equal to 0.85;

[0016] The morphological filtering is adopted to eliminate the salt and pepper noise, and the park green space evolution patch is extracted by using GIS spatial overlay analysis.

[0017] Preferably, in the step S3, the landscape pattern index includes:

[0018] At the landscape level: aggregation index (AI), Shannon diversity index (SHDI), contagion index (CONTAG);

[0019] At the type level: patch density (PD), maximum patch index (LPI), shape index (LSI);

[0020] The standard deviation ellipse analysis and the center of gravity migration model are used to represent the spatial evolution directionality characteristics.

[0021] Preferably, in the step S4, the driving evaluation index system construction includes:

[0022] Social factors: population density, urbanization rate, public service facility density;

[0023] Economic factors: GDP growth rate, land transfer price, tertiary industry proportion;

[0024] Environmental factors: NDVI vegetation index, heat island intensity, water body buffer distance;

[0025] Human factors: Cultural heritage distribution, public green land satisfaction survey data;

[0026] The geographic detector model is used to quantify the interaction strength of each factor.

[0027] Preferably, the driving mechanism analysis further comprises:

[0028] The spatial autocorrelation analysis is used to identify the aggregation and heterogeneity of the park green land evolution;

[0029] The spatial non-stationary influence of the driving factors is revealed based on the MGWR (multi-scale geographic weighted regression) model.

[0030] Preferably, the three-level optimization strategy in the step S5 comprises:

[0031] Macro level: based on the ecological security pattern, the park green land growth boundary is determined;

[0032] Medium level: the minimum cumulative resistance model is used to construct the green land corridor network;

[0033] Micro level: the community-level pocket park layout is optimized through the accessibility analysis.

[0034] Preferably, it further comprises:

[0035] A dynamic monitoring and early warning module is constructed, and an alarm is triggered when the annual change rate of the park green land fragmentation index is greater than 5%;

[0036] The ARCGISEngine is integrated to develop an interactive decision support system, and multi-scenario simulation deduction is realized.

[0037] Preferably, the method uses the time series InSAR technology to assist in monitoring the ground deformation data of the park green land, and is used for evaluating the influence of underground development activities on the stability of the green land.

[0038] Preferably, the driving factor analysis result and the optimization strategy are associated and stored through the knowledge graph technology, and the intelligent planning scheme generation based on semantic retrieval is supported.

[0039] The technical effects and advantages of the present application are as follows:

[0040] 1. By acquiring multi-period remote sensing image data, planning map data, and multi-source data such as social, economic, environmental, and cultural data, a database is constructed, and Python scripts are used for spatio-temporal alignment and indexing to ensure the spatio-temporal consistency and query efficiency of the data. In addition, InSAR technology is used to monitor the impact of underground activities on the stability of green spaces, supplementing the shortcomings of traditional remote sensing and providing comprehensive and accurate data support for subsequent analysis.

[0041] 2. The random forest supervised classification algorithm is used to interpret the remote sensing image on the ENVI5.6 platform. By setting reasonable sample size, feature band, and model parameters, and strictly following the post-classification processing procedures such as accuracy verification, morphological filtering, and evolution patch extraction, the accuracy of the generated park green space spatial distribution map is high and conforms to the actual planning unit scale.

[0042] 3. The landscape ecology theory is used to select landscape pattern indices at the landscape level and type level, and the standard deviation ellipse analysis and barycenter migration model are used to quantify the spatio-temporal evolution characteristics of park green spaces from the aspects of structure and composition, directly reflecting the impact of urbanization on green spaces, such as green space fragmentation, expansion direction, and barycenter migration.

[0043] 4. By constructing a driving evaluation index system containing natural factors and human factors, the geographic detector model is used to quantify the explanatory power and interaction strength of each factor, avoiding the interference of multiple collinearity; the MGWR model is used to reveal the spatial non-stationary influence of driving factors, which can clearly identify the dominant driving factors and influence differences in different regions, providing scientific basis for regional policy making.

[0044] 5. Based on the evolution characteristics and driving mechanism, a "macro-meso-micro" three-level spatial pattern optimization strategy is generated. Macroscopically, the MCR model is used to delineate the growth boundary of park green space to ensure ecological safety; mesoscopically, the minimum cumulative resistance model is used to construct a green space corridor network to improve landscape connectivity; microscopically, the accessibility analysis is used to optimize the layout of community-level pocket parks to directly respond to people's livelihood needs. At the same time, the driving factors, evolution characteristics, and optimization strategies are related through the knowledge graph, realizing multi-dimensional data intelligent correlation, improving the pertinence and scientificity of the planning scheme.

[0045] 6. By constructing a dynamic monitoring and early warning module, the annual change rate of landscape indices is calculated in real time for dynamic monitoring to avoid lagging governance. An interactive decision support system is developed based on ARCGISEngine to support planners to interactively adjust parameters and conduct multi-scenario simulation and deduction to assist scientific decision-making and reduce the threshold for using the model. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0047] All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative labor fall within the scope of protection of the present application.

[0048] The evaluation method of the spatiotemporal pattern evolution characteristics and evolution driving factors of the park green space is shown in the accompanying drawings, and includes the following steps: Figure 1

[0049] S1. Obtain multi-period remote sensing image data and planning drawings of the target city, and construct a multi-source database containing social, economic, environmental and cultural factors;

[0050] In specific implementation, the data sources are as follows:

[0051] Remote sensing data: Obtain Landsat and Sentinel series multispectral remote sensing images (time span ≥ 20 years), spatial resolution ≤ 30 meters, and cloud coverage rate < 10%;

[0052] Planning drawings: Integrate the land space planning, green space system special planning and historical land use change map of a city;

[0053] Social and economic data: Statistical yearbook, population census data, land transaction records and public survey questionnaire (covering ≥ 80% of the streets);

[0054] Environmental data: Temperature / rainfall data of weather stations, MODIS NDVI products and urban heat island intensity inversion results.

[0055] Multi-source database construction:

[0056] Use Python script to perform spatiotemporal alignment (WGS84 coordinate system, time granularity unified to year), store through PostgreSQL database, and establish spatial-attribute association index.

[0057] InSAR technology assisted monitoring:

[0058] Obtain Sentinel-1 SAR image, generate surface deformation rate map using SBAS-InSAR technology, spatial resolution 20 meters, accuracy ± 3mm / year, superimpose deformation data and park green space layer, identify areas with settlement > 10mm caused by underground engineering (such as subway), and mark as "high interference area".

[0059] Specifically, the spatiotemporal consistency is ensured through multi-source data fusion, the underground activities that cannot be captured by traditional remote sensing are supplemented by InSAR technology, the query efficiency is improved through database indexing, the subsequent model is efficiently called, the surface deformation data can quantify the hidden damage of human activities to the stability of green space, and engineering constraint conditions are provided for optimization strategy.

[0060] ​S2. Interpret remote sensing images based on random forest supervised classification algorithm, generate spatial distribution maps of parks in different periods, and optimize accuracy through post-classification processing;

[0061] In specific implementation, random forest classification:

[0062] In ENVI5.6, select training samples (parks, water, buildings, bare land, etc. 6 categories), sample number ≥ 500 / class, feature bands include RGB, NIR, NDVI, texture features (GLCM);

[0063] Set the number of decision trees = 200, maximum depth = 15, OOB error < 5%.

[0064] Post-classification processing:

[0065] Accuracy verification: generate confusion matrix, calculate overall accuracy (OA ≥ 90%) and Kappa coefficient (≥ 0.85), and manually correct areas that do not meet the standard;

[0066] Morphological filtering: use 3x3 kernel "open operation" to eliminate salt and pepper noise and retain green patches with an area greater than 0.5 hectares;

[0067] Evolution patch extraction: overlay multiple classification results in ArcGIS, identify new / lost patches using "raster calculator", and generate spatiotemporal evolution trajectory map.

[0068] Specifically, random forest is used for noise reduction, suitable for medium and low resolution image classification; morphological filtering avoids small area misclassification patch interference, ensures that the result meets the actual planning unit scale, and Kappa coefficient threshold setting is based on industry standard (Cohen's Kappa > 0.8 for high consistency), ensuring the credibility of the classification results.

[0069] S3. Use landscape ecology theory to select landscape pattern indices at landscape and type levels, and use Fragstats software to quantify spatiotemporal evolution characteristics;

[0070] In specific implementation, index selection and calculation:

[0071] Landscape level:

[0072] Aggregation index (AI): quantifies green space connectivity, higher values indicate stronger aggregation;

[0073] Shannon diversity index (SHDI): reflects green space type richness, higher values indicate increased landscape heterogeneity;

[0074] CONTAG: assesses the spread of dominant green space types, lower values indicate increased fragmentation;

[0075] Type level:

[0076] Patch density (PD): patch number per unit area, value increase indicates fragmentation;

[0077] Largest patch index (LPI): dominant patch area proportion, value decrease indicates core green space loss.

[0078] Spatial orientation analysis:

[0079] Standard deviation ellipse: calculate the main axis direction (azimuth of long axis) and dispersion degree (ratio of short axis to long axis) of green space distribution.

[0080] Barycenter migration model: track the displacement distance (Euclidean distance) and direction of green space barycenter coordinates in each period.

[0081] Specifically, the evolution characteristics are quantified from the aspects of structure and composition through landscape indices, for example, the decrease of AI and CONTAG indicates that urbanization leads to green space fragmentation; the standard deviation ellipse reveals whether the expansion direction is consistent with the planning direction. The barycenter migration can intuitively reflect the squeezing effect of urban development axis on green space.

[0082] S4. Construct a driving evaluation index system containing natural and human factors, and analyze the driving mechanism through multiple regression model and spatial econometric model;

[0083] In specific implementation, the index system construction:

[0084] Social factors: population density (person / km²), urbanization rate (%), number of parks per 10,000 people;

[0085] Economic factors: land GDP (ten thousand yuan / km²), land premium rate (%), third industry investment;

[0086] Environmental factors: annual average NDVI, heat island intensity (℃), distance to water system (m);

[0087] Human factors: cultural heritage point density, public satisfaction (Likert 5-level scale);

[0088] Geographical detector model:

[0089] Calculate the explanatory power of each factor (R²), formula:

[0090]

[0091] Wherein, is the number of sub-area samples, is the variance of sub-area, is the number of sub-areas;

[0092] Interaction detection: judge whether the factors are nonlinear enhancement (​ )) or independent action.

[0093] MGWR model:

[0094] Model formula:

[0095]

[0096] where, is the spatial coordinate, is the spatially varying coefficient;

[0097] Bandwidth optimization: Optimal bandwidth is selected using AICc criterion, revealing the spatial heterogeneity of driving factors.

[0098] Specifically, the geographic detector quantifies the influence of single factors, avoiding multiple collinearity interference; the MGWR model captures the spatial non-stationarity of factor effects (e.g., central city is more strongly driven by economy, while suburban area is more significantly constrained by environment), providing basis for zoning policy.

[0099] S5. Based on the evolution characteristics and driving mechanism, generate "macro-meso-micro" three-level spatial pattern optimization strategy and output visual decision map.

[0100] In specific implementation, the macro-strategy (ecological safety pattern):

[0101] Based on MCR model: Identify ecological sources (areas with current green land LPI > 30%), calculate terrain and land use resistance surface, generate minimum cost path, and delineate green land growth boundary.

[0102] Meso-strategy (green corridor):

[0103] Use Graphab software to build corridor network, set corridor width = 50 meters, connection degree index (γ) > 0.6, and ensure biological migration accessibility.

[0104] Micro-strategy (community park):

[0105] Network analysis (ArcGIS Network Analyst): Cover 95% of residential areas with a service radius of 500 meters, add pocket parks in gap areas, and optimize site selection priority (population density > 10,000 people / km² first).

[0106] Visual decision map:

[0107] Based on knowledge graph (Neo4j graph database), associate "driving factor-evolution characteristics-optimization strategy", support semantic query (e.g., "areas with high heat island effect should increase what kind of green land").

[0108] Specifically, the MCR model is used to constrain the disorderly expansion from the perspective of ecological safety; the corridor network improves landscape connectivity; the community-level optimization directly responds to people's livelihood needs; the knowledge graph realizes intelligent association of multi-dimensional data, improving the pertinence and scientificity of the planning scheme.

[0109] Also includes:

[0110] A dynamic monitoring and early warning module is constructed, and an alarm is triggered when the park green space fragmentation index annual change rate is greater than 5%;

[0111] An interactive decision support system is developed by integrating ARCGIS Engine to realize multi-scenario simulation and deduction.

[0112] Specifically, the early warning module is used to calculate the annual change rate of landscape index in real time, and when the fragmentation (PD) change rate is greater than 5%, an early warning is triggered and the mutation area (such as the surrounding of newly developed land) is located. A C / S architecture system is developed, integrating the "data management-model calculation-scenario simulation" module, supporting planners to interactively adjust parameters (such as green space rate threshold), and real-time previewing the optimization scheme effect, so as to dynamically monitor and avoid lagging governance. The ARCGIS system reduces the threshold for using models, and multi-scenario simulation (such as "high-density development" vs. "ecological priority") helps scientific decision-making. Specific embodiments

[0113] Taking the analysis of Lu'an City from 2010 to 2020 as an example:

[0114] Temporal and spatial evolution:

[0115] The green space area decreased by 12%, and the center of gravity shifted 2.3 km to the northeast (consistent with the development direction of the economic development zone).

[0116] The AI index decreased from 78.5 to 65.2, and the fragmentation was significant.

[0117] Driving mechanism:

[0118] The dominant factors are land premium rate ( ) and population density ( ), and the interaction is synergistic ( ).

[0119] Calculate the explanatory power of each factor ( value), formula:

[0120]

[0121] Among them, is the number of samples in the first sub-area after stratification (divided into 3 layers according to the land premium rate), the variance of the change in park green space area in the first sub-area, For the total number of samples (the total number of administrative units in Liu'an City, assuming 100), For the overall variance of the change in the area of park green space in the city;

[0122] Take the land premium rate as an example:

[0123] Stratification: divide the land premium rate into high, medium, and low three layers (high, medium, and low) , and assume that the sample size of each layer is 30, 40, and 30, respectively.

[0124] Calculate the sub-area variance:

[0125] High premium area: variance of decrease in park green space area ;

[0126] Medium premium area: variance ;

[0127] Low premium area: variance ;

[0128] Numerator calculation:

[0129]

[0130] Denominator calculation:

[0131] Overall variance , then:

[0132]

[0133] Value calculation:

[0134]

[0135] Interaction verification:

[0136] Compare the interaction value with the maximum value of the single factor:

[0137]

[0138] Therefore, the interaction between the land premium rate and population density presents a nonlinear increase, indicating that the combined effect of the two factors on green space evolution is much greater than that of a single factor.

[0139] MGWR shows that the economic factor has a greater influence on the city center ( ) than the suburbs ( ).

[0140] According to the MGWR model formula:

[0141]

[0142] where, is the park green area change rate of the th spatial unit, is the land premium rate of the th unit, is the regression coefficient varying with spatial location.

[0143] Calculation steps:

[0144] Bandwidth optimization:

[0145] The optimal bandwidth (search range 1~50km) is selected by AICc criterion, the minimum AICc corresponds to bandwidth=15km.

[0146] Local weighted regression:

[0147] For the central unit (coordinates , ), a Gaussian kernel weight matrix is constructed with 15km as the radius, the weight of adjacent units decays with distance.

[0148] Solve by weighted least squares:

[0149]

[0150] Input data example:

[0151] The land premium rate of the central unit , the average premium rate of the adjacent 10 units ;

[0152] The park green area change rate , by substituting the calculation , it shows that the green area decreases for every increase in land premium.

[0153] Suburban coefficient calculation:

[0154] The land premium rate of the suburban unit , the average premium rate of the adjacent units ;

[0155] The park green area change rate , by substituting the calculation , it shows that the economic factor has a weaker impact in the suburbs.

[0156] The land development intensity is high in the city center, and the marginal effect of premium rate on green invasion is significant ( );

[0157] The suburban area is mainly agricultural / ecological land, with weak economic driving force , which is consistent with the actual urbanization gradient.

[0158] The significance of combining geographic detector with MGWR:

[0159] The geographic detector identifies that the land premium rate is the key factor ( =0.58), and MGWR further reveals the differentiation of its influence in space (city center > suburb ), which shows that the land development premium in the city center needs to be strictly controlled, and the suburb focuses on ecological compensation.

[0160] Mathematical verification:

[0161] If the land premium rate of the city center unit increases from 8% to 10%, the predicted green land reduction is:

[0162]

[0163] The actual observation of the city center green land reduction from 2010 to 2020 is 15%, and the model predicts a cumulative reduction of 14% (93% agreement).

[0164] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. The evaluation method of the spatiotemporal pattern evolution characteristics and evolution driving factors of park green spaces is characterized by: The following steps are involved: S1. Obtain multi-period remote sensing imagery and planning maps of the target city and construct a multi-source database encompassing social, economic, environmental, and humanistic factors; S2. Interpret remote sensing images using a random forest supervised classification algorithm to generate spatial distribution maps of park green spaces at different time periods, and optimize accuracy through post-classification processing. S3. Using landscape ecology theory, we selected landscape pattern indices at the landscape and type levels and used Fragstats software to quantify spatiotemporal evolution characteristics. S4. Construct a driving evaluation index system that includes both natural and human factors, and analyze the driving mechanism through multiple regression models and spatial econometric models; S5. Based on the evolutionary characteristics and driving mechanisms, a three-level spatial pattern optimization strategy (macro-meso-micro) is generated and a visual decision map is output.

2. The method for evaluating the spatiotemporal evolution characteristics and driving factors of park green space according to claim 1 is characterized in that: In step S2, random forest supervised classification is implemented using the ENVI5.6 platform, and post-classification processing includes: The Kappa coefficient was calculated by confusion matrix for accuracy verification, with a threshold of ≥0.85; Morphological filtering was used to eliminate salt and pepper noise, and GIS spatial overlay analysis was used to extract park green space evolution patches.

3. The method for evaluating the spatiotemporal evolution characteristics and driving factors of park green space according to claim 1 is characterized in that: In step S3, the landscape pattern index includes: Landscape level: aggregation index, Shannon diversity index, and spread index; Type level: plaque density, maximum plaque index, shape index; The directional characteristics of spatial evolution are characterized by standard deviation ellipse analysis and center of gravity migration model.

4. The method for evaluating the spatiotemporal evolution characteristics and driving factors of park green space according to claim 1 is characterized in that: The construction of the driving evaluation index system in step S4 includes: Social factors: population density, urbanization rate, and density of public service facilities; Economic factors: GDP growth rate, land transfer price, and proportion of the tertiary industry; Environmental factors: NDVI vegetation index, heat island intensity, and distance to water buffer zones; Human factors: cultural heritage distribution and public green space satisfaction survey data; The geographic detector model was used to quantify the interaction strength of each factor.

5. The method for evaluating the spatiotemporal evolution characteristics and driving factors of park green space according to claim 4 is characterized in that: The driving mechanism analysis further includes: Identify the clustering and heterogeneity of park green space evolution through spatial autocorrelation analysis; The spatial non-stationary impact of driving factors is revealed based on the MGWR model.

6. The method for evaluating the spatiotemporal evolution characteristics and driving factors of park green space according to claim 1 is characterized in that: The three-level optimization strategy in step S5 includes: Macro level: Delineate the growth boundaries of park green spaces based on the ecological security pattern; Meso-level: Use the minimum cumulative resistance model to construct a green corridor network; Micro level: Optimize the layout of community-level pocket parks through accessibility analysis.

7. The method for evaluating the spatiotemporal evolution characteristics and driving factors of park green space according to claim 1 is characterized in that: Also includes: Build a dynamic monitoring and early warning module to trigger an alarm when the annual change rate of the park green space fragmentation index is greater than 5%; Integrate ARCGIS Engine to develop an interactive decision support system and realize multi-scenario simulation and deduction.

8. The method for evaluating the spatiotemporal evolution characteristics and driving factors of park green space according to claim 1 is characterized in that: The method uses time-series InSAR technology to assist in monitoring the surface deformation data of park green spaces and to evaluate the impact of underground development activities on the stability of green spaces.

9. The method for evaluating the spatiotemporal evolution characteristics and driving factors of park green space according to claim 1 is characterized in that: The driving factor analysis results and optimization strategies are associated and stored through knowledge graph technology, supporting the generation of intelligent planning solutions based on semantic retrieval.

Images