A method and system for evaluating ecological restoration effects based on satellite remote sensing

Abstract

The present invention discloses a method and system for evaluating ecological restoration effects based on satellite remote sensing, comprising collecting satellite remote sensing data and field soil data of a preset area, using the remote sensing data of the preset area within a specified time as data to be analyzed; calculating time series deviation and contrast based on the satellite remote sensing data to obtain graded areas; performing deviation correction on the satellite remote sensing data according to the field soil data within the graded areas, and inputting the corrected satellite remote sensing data into a spatiotemporal ecological development prediction model to obtain ecological development prediction data; performing restoration uncertainty analysis based on the ecological development prediction data and the field soil data to obtain ecological restoration effect confidence, performing elasticity analysis based on the ecological restoration effect confidence and the satellite remote sensing data to obtain an ecological resilience index; constructing a time-varying ecological restoration effect evaluation model based on the ecological resilience index, and outputting an evaluation result based on the data to be analyzed.

Description

Technical Field

[0001] The present invention relates to the field of ecological assessment, and in particular to an ecological restoration effect assessment method and system based on satellite remote sensing. Background Art

[0002] With the combined impacts of global climate change and human activities, ecosystem degradation is becoming increasingly prominent, and ecological restoration has become a key driver of regional sustainable development. Traditional ecological restoration effectiveness assessments often rely on small-scale field surveys, which are subject to bottlenecks such as long data collection cycles, limited spatial coverage, and insufficient dynamic monitoring capabilities. These assessments are unable to meet the needs of large-scale, multi-factor, and time-series assessments. Although satellite remote sensing technology can provide continuous observations of wide-area surface information, its application in ecological restoration scenarios still faces three challenges: First, the integration mechanism of multi-source remote sensing data with ground-based measured data is not yet perfected, resulting in the accuracy of ecological parameter inversion being limited by the inherent bias of a single data source; second, ecosystem evolution has significant spatiotemporal heterogeneity, making it difficult for conventional static assessment models to capture the nonlinear responses and elastic recovery thresholds during the restoration process; third, the ecological effects of restoration projects are often affected by uncertainties such as climate fluctuations and human interference. Existing methods lack theoretical knowledge to quantify the confidence level of restoration effects and the system's ability to resist interference.

[0003] In recent years, dynamic monitoring technology based on time-series remote sensing data has provided new ideas for the above-mentioned problems. Studies have shown that by integrating multi-dimensional ecological parameters such as vegetation cover, surface temperature, and biodiversity remote sensing index, a comprehensive representation system for ecosystem health can be constructed. However, existing studies mostly focus on the status assessment of a single time node, lacking a comparative analysis of the temporal evolution of restoration areas and non-restoration areas, resulting in insufficient accuracy in identifying key restoration driving factors. In addition, as the material carrier of ecological restoration, the spatial coupling relationship between the physical and chemical properties of soil and remote sensing inversion parameters has not been fully explored, which limits the construction of a "sky-ground" collaborative assessment model. How to lock in ecologically sensitive areas through time-series deviation analysis, correct remote sensing inversion errors in combination with ground soil data, and then establish a dynamic assessment model that integrates elastic resilience has become a key technical breakthrough direction for improving the quantitative accuracy of ecological restoration effects. Summary of the Invention

[0004] The purpose of the present invention is to provide an ecological restoration effect evaluation method based on satellite remote sensing.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0006] The present invention comprises the following steps:

[0007] Collect satellite remote sensing data and field soil data of a preset area, and use the remote sensing data of the preset area within a specified time as data to be analyzed; the data to be analyzed includes vegetation information, land use and land cover data, topography data, hydrological data, thermal environment data, and biodiversity data;

[0008] Calculating the time series deviation based on the satellite remote sensing data, taking the preset area where the time series deviation is greater than 0 as the study area, and otherwise as the non-study area, and classifying the changes in the study area according to the contrast to obtain a graded area; comprising:

[0009] Calculate timing deviation:

[0010]

[0011] The timing deviation is θ, and the upper limit of the monitoring time is T s , the ath remote sensing data at time s2 is h a (s2), the ath remote sensing data at time s1 is h a (s1), the monitoring start time is s o , the altitude of the remote sensing satellite is D, the radius of the earth is R, and the viewing angle is

[0012] performing deviation correction on the satellite remote sensing data according to the field soil data in the grading area, and inputting the corrected satellite remote sensing data into a spatiotemporal ecological development prediction model to obtain ecological development prediction data;

[0013] Performing restoration uncertainty analysis based on the ecological development forecast data and the field soil data to obtain an ecological restoration effect confidence level, and performing elasticity analysis based on the ecological restoration effect confidence level and the satellite remote sensing data to obtain an ecological resilience index;

[0014] A time-varying ecological restoration effect evaluation model is constructed based on the ecological resilience index, the data to be analyzed is input into the time-varying ecological restoration effect evaluation model, and an evaluation result is output.

[0015] Furthermore, the method for grading the changes in the study area according to the contrast to obtain graded areas includes:

[0016] The subgrids were obtained by dividing the grid according to the size and shape of the study area. The change features within the subgrids were extracted according to the change ratio of the remote sensing satellite data. The change features with a change ratio greater than 0.013 were taken as key features.

[0017] Taking the subgrid as a unit, a subgrid-based clustering feature tree is constructed according to the contrast and key features. The key features are used as clustering feature nodes, and the abnormality degree index of the clustering feature nodes is calculated.

[0018] The abnormality index is input into the student psychology optimization algorithm, abnormal clustering feature nodes are eliminated according to the optimization characteristics, the optimized key feature set is obtained, the clustering feature nodes in the optimized key feature set are integrated and adjusted, and the clustering feature tree is updated;

[0019] Calculate the contrast of the subgrid based on the temporal and spatial changes according to the updated cluster feature tree:

[0020]

[0021] The contrast of the w-th subgrid is l w , the number of key features of the optimized key feature set in the w-th sub-grid is N w , the vth key feature at time s2 is H v (s2), the vth distinguishing feature at the s1th moment is H v (s1), the weight coefficient of the vth key feature is ζ v , the sign function is sgn(·);

[0022] When the contrast is less than 0, the subgrid is a poor-quality restoration area; when the contrast is greater than 0 and less than 0.0051, the subgrid is a poor-quality restoration area; when the contrast is greater than 0.0051 and less than 0.015, the subgrid is a medium-quality restoration area; when the contrast is greater than 0.015 and less than 0.0297, the subgrid is a good-quality restoration area; when the contrast is greater than 0.0297, the subgrid is an excellent-quality restoration area.

[0023] Adjacent subgrids of the same level are connected according to the contrast, and the connected results are output as the graded area.

[0024] Furthermore, the method of the spatiotemporal ecological development prediction model includes:

[0025] Collect climate and environmental data within the preset area, and perform time series impact analysis on the climate and environmental data to obtain environmental factors;

[0026] Spatiotemporal ecological development prediction models include graph networks, time-guided embeddings, and long short-term memory neural networks;

[0027] Graph networks represent pixels, regions, or objects in remote sensing data as graph nodes, and the relationships between graph nodes as edges. Graph neural network architecture is used to capture the complex relationships between graph nodes and extract the complex spatiotemporal features of remote sensing data based on these complex relationships.

[0028] Time-guided embedding incorporates temporal information into the embedding vector by learning spatiotemporal ecological development, and combines complex spatiotemporal features to capture the temporal cyclical pattern of ecological development;

[0029] Long short-term memory neural networks capture long-term dependencies in time series through gating mechanisms and cell states, and make ecological development predictions based on temporal cycle patterns and long-term dependencies.

[0030] Furthermore, a method for performing restoration uncertainty analysis based on the ecological development prediction data and the field soil data to obtain the confidence level of the ecological restoration effect includes:

[0031] The ecological development prediction data and field soil data are converted into multidimensional vectors, and the ecological characteristic matrix is ​​constructed based on the multidimensional vectors;

[0032] Obtain ecological restoration effect indicators and ecological characteristics of the classified areas, and construct the least absolute shrinkage and selection operator Lasso regression model based on the ecological restoration effect indicators and ecological characteristics;

[0033] In the Lasso regression model, the square sum of the standardized coefficients of the ecological characteristics is used as the sensitivity coefficient, the non-zero frequency of the weight of the ecological characteristics in the Lasso regression model is obtained, and the importance of the ecological characteristics is calculated:

[0034]

[0035] The sensitivity coefficient of the wth ecological characteristic is χ w , the non-zero frequency of the wth ecological characteristic is F w , the importance of the wth ecological feature is φ w ;

[0036] The Pearson correlation coefficient was used to calculate the correlation between ecological characteristics, and ecological characteristics with correlation greater than 0.673 were regarded as key characteristics. The elbow rule was used to determine the optimal number of clusters for key characteristics, and the hierarchical regions were divided into multiple ecologically homogeneous groups according to the relationship characteristics;

[0037] Random sampling was performed on the ecologically homogeneous groups according to the proportion, and traditional ecological indicators were used for random sampling. A control group of the same area was set up to obtain the ecological environment quality score, soil quality index, and vegetation coverage of the ecologically homogeneous group and the control group. The ecological restoration effects of the ecologically homogeneous group and the control group were calculated based on the following:

[0038]

[0039] The ecological restoration effect of group c is The experimental area of ​​group c is S c , the vegetation coverage at time s2 is The vegetation coverage at time s1 is The ecological environment quality score at time s2 is The ecological environment quality score at time s1 is The soil quality index at time s2 is The soil quality index at time s1 is The weight coefficients are β1, β2, and β3;

[0040] Draw the temporal ecological restoration effect reference curve based on ecological characteristics and the temporal ecological restoration effect curve based on key characteristics, and calculate the difference between the temporal ecological restoration effect reference curve and the temporal ecological restoration effect curve:

[0041]

[0042] The number of features is N c , the i-th feature is b i , feature b at time s2 i The ecological restoration effect is Feature b at time s1 i The ecological restoration effect is referenced as The difference between the cth ecological homogeneous group is K c ;

[0043] The ratio of the difference of the sampling results to the mean of the difference is taken as the coefficient of variation. When the coefficient of variation is greater than 0.279, the difference between 1 and the coefficient of variation is output as the confidence level of the ecological restoration effect.

[0044] Furthermore, the method for obtaining an ecological resilience index by performing resilience analysis based on the ecological restoration effect confidence and the satellite remote sensing data includes:

[0045] A multidimensional ecological indicator system was constructed based on satellite remote sensing data. Resilience assessment indicators were screened based on the 4R theory, refining the indicators to identify key ecological functions according to the type of damage risk, eliminating redundant indicators based on the substitutability of species within the ecosystem, and selecting intelligent indicators based on the ability to self-regulate and mobilize recovery when faced with damage risks.

[0046] The ecological resilience change process is divided into the preparation stage, the resistance and absorption stage, the adaptive adjustment stage and the recovery stage according to time. Based on the performance indicators of each stage of the ecosystem and the detailed indicators of the 4R theory, a comprehensive ecological resilience measurement model is constructed.

[0047] Calculate the Ecological Resilience Index:

[0048]

[0049] The ecological resilience index of the kth graded area is The number of ecological resilience assessment indicators is M r , the weight of the rth indicator in the tth stage is ρ r(t), the elasticity evaluation index of the t stage is f r (t), the preparation stage is t1, the recovery stage is t4, and the confidence level of the ecological restoration effect of the kth graded area in the tth stage is The indicator weight of the preparation stage t1 is ρ r (t1), the elasticity evaluation index of the preparation stage t1 is f r (t1), the confidence level of ecological restoration effect of the kth graded area in the preparation stage t1 is

[0050] Furthermore, a method for constructing a time-varying ecological restoration effect evaluation model based on the ecological resilience index includes:

[0051] Obtain the ecological performance evaluation indicators and corresponding weights of the study area, and construct the objective function based on the ecological resilience index and loss function. The expression is:

[0052]

[0053]

[0054] The loss function is The initial ecosystem performance of the kth classification area is HO k , the number of graded areas in the study area is M k , the weight of the rth indicator at the sth moment is η r (s), the ecological performance evaluation index at the sth moment is g r (s), the ecosystem performance of the k-th classification area at the s-th moment is H k (s), the ecological resilience index of the kth graded area at the sth moment is The objective function at time s is

[0055] Time-varying ecological restoration effect evaluation models include random forest algorithms, time series analysis algorithms, and spatiotemporal graph neural networks;

[0056] The random forest algorithm uses an integrated learning algorithm to build multiple decision trees based on satellite remote sensing data, using randomly selected features and samples to construct multiple decision trees, and then determines the classification results through a voting mechanism.

[0057] The time series analysis algorithm is based on the time-space spectrum feature. It extracts the time, space and spectral features of the pixel points and combines them with the support vector machine classifier for change detection. The time series change pattern of the surface features is extracted based on the change detection.

[0058] The spatiotemporal graph neural network is constructed as a spatiotemporal graph based on the law of temporal changes. The spatiotemporal module of the graph neural network is used to model spatial and temporal dependencies respectively. The ecological restoration effect is predicted based on the dependency relationship, and the predicted ecological restoration effect is output.

[0059] The second aspect is an ecological restoration effect evaluation system based on satellite remote sensing, including:

[0060] Data acquisition module: used to collect satellite remote sensing data and field soil data of a preset area, and use the remote sensing data of the preset area within a specified time as the data to be analyzed; the data to be analyzed includes vegetation information, land use and land cover data, topographic data, hydrological data, thermal environment data and biodiversity data;

[0061] A partitioning and grading module is configured to calculate the time series deviation based on the satellite remote sensing data, and to define the preset area where the time series deviation is greater than 0 as the study area, and vice versa as the non-study area, and to classify the changes within the study area according to the contrast to obtain the graded area;

[0062] A correction prediction module is configured to correct the deviation of the satellite remote sensing data according to the field soil data in the classification area, and input the corrected satellite remote sensing data into the spatiotemporal ecological development prediction model to obtain ecological development prediction data;

[0063] Confidence elasticity module: used to perform restoration uncertainty analysis based on the ecological development prediction data and the field soil data to obtain the ecological restoration effect confidence, and to perform elasticity analysis based on the ecological restoration effect confidence and the satellite remote sensing data to obtain the ecological resilience index;

[0064] Modeling output module: used to construct a time-varying ecological restoration effect evaluation model based on the ecological resilience index, input the data to be analyzed into the time-varying ecological restoration effect evaluation model, and output the evaluation results.

[0065] The present invention is a method and system for evaluating ecological restoration effects based on satellite remote sensing. Compared with the existing technology, the present invention has the following technical effects:

[0066] The present invention can improve the accuracy of ecological restoration effect evaluation through deviation correction, acquisition of ecological development prediction data, restoration uncertainty analysis, elasticity analysis and model construction steps, thereby improving the precision of ecological restoration effect evaluation and optimizing ecological restoration effect evaluation, which can greatly save resources and improve work efficiency. It can realize intelligent evaluation of ecological restoration effect and perform restoration uncertainty analysis and elasticity analysis on ecological restoration effect evaluation in real time, which is of great significance to ecological restoration effect evaluation, can adapt to ecological restoration effect evaluation of different standards and different ecological restoration effect evaluation needs, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a flowchart of the steps of an ecological restoration effect evaluation method based on satellite remote sensing of the present invention. DETAILED DESCRIPTION

[0068] The present invention will be further described below through specific examples. The illustrative examples and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0069] The present invention provides an ecological restoration effect evaluation method and system based on satellite remote sensing, comprising the following steps:

[0070] like Figure 1 As shown, in this embodiment, the following steps are included:

[0071] Collect satellite remote sensing data and field soil data of a preset area, and use the remote sensing data of the preset area within a specified time as data to be analyzed; the data to be analyzed includes vegetation information, land use and land cover data, topography data, hydrological data, thermal environment data, and biodiversity data;

[0072] In the actual evaluation, a certain ecological restoration area was selected as the research object, and satellite remote sensing data and field soil data from 2019 to 2023 were used as historical data to train the time-varying ecological restoration effect evaluation model. The satellite remote sensing data of Region A and Region B in the spring of 2024 were used as the data to be analyzed.

[0073] Region A: Vegetation coverage is 85%, dominated by broad-leaved and coniferous forests, mixed with a small amount of shrubs and herbaceous plants, and the vegetation health index is 0.8. Land use and land cover data are forest coverage of 70%, grassland of 15%, water area of ​​10%, and construction land of 5%. Topography data are an average altitude of 800 meters, an average slope of 15°, with slopes ranging from 10° to 20° in most areas, and a predominance of mountains and hills, with mountains accounting for 60% and hills accounting for 40%. Hydrological data are a river density of 0.3 km / km2, an average annual precipitation of 1,200 mm, and an average groundwater depth of 5 meters. Thermal environment data are an average summer surface temperature of 25°C, a weak heat island effect, and an average temperature difference of less than 2°C. Biodiversity data are a species richness index of 12 and the number of rare species is 10.

[0074] Region B: Vegetation information includes a vegetation coverage rate of 30%, dominated by sparse herbaceous plants and a small number of drought-resistant shrubs, with herbaceous plants accounting for 71% and shrubs accounting for 29%, and a vegetation health index of 0.3. Land use and land cover data include forest coverage of 10%, grassland of 40%, water area of ​​5%, construction land area of ​​25%, and wasteland area of ​​20%. Topography and landform data include an average altitude of 500 meters, an average slope of 5°, with slopes ranging from 0° to 10° in most areas, and mainly plains with a small number of low hills, accounting for 80% and low hills accounting for 20%. Hydrological data include a river density of 0.1 km / km2, an average annual precipitation of 600 mm, and an average groundwater depth of 15 meters. Thermal environment data include an average summer surface temperature of 32°C, a strong heat island effect, and an average temperature difference of more than 5°C. Biodiversity data include a species richness index of 4 and the number of rare species of 2.

[0075] Field soil data for Area A in March were as follows: sand content 30%, silt content 40%, clay content 30%, organic matter content 3.5%, total nitrogen content 0.2%, available phosphorus content 25 mg / kg, available potassium content 180 mg / kg, pH = 6.8, average moisture content 25%, microbial biomass carbon 1500 mg / kg, and soil bulk density 1.2 g / cm3;

[0076] Field soil data for Area B in March were as follows: sand content 60%, silt content 20%, clay content 20%, organic matter content 1.0%, total nitrogen content 0.08%, available phosphorus content 10 mg / kg, available potassium content 80 mg / kg, pH 8.2, average moisture content 10%, microbial biomass carbon 500 mg / kg, and soil bulk density 1.5 g / cm3;

[0077] Calculating the time series deviation based on the satellite remote sensing data, taking the preset area where the time series deviation is greater than 0 as the study area, and otherwise as the non-study area, and classifying the changes in the study area according to the contrast to obtain a graded area; comprising:

[0078] Calculate timing deviation:

[0079]

[0080] The timing deviation is θ, and the upper limit of the monitoring time is T s , the ath remote sensing data at time s2 is h a (s2), the ath remote sensing data at time s1 is h a (s1), the monitoring start time is s o , the altitude of the remote sensing satellite is D, the radius of the earth is R, and the viewing angle is

[0081] In the actual assessment, the study areas are Area A and Area B; the sub-areas of Area A are A1, A2, A3, and A4; the sub-areas of Area B are B1, B2, and B3; the graded areas of Area A are A1, A2, A3, and A4; the graded areas of Area B are B1 and B3;

[0082] performing deviation correction on the satellite remote sensing data according to the field soil data in the grading area, and inputting the corrected satellite remote sensing data into a spatiotemporal ecological development prediction model to obtain ecological development prediction data;

[0083] Performing restoration uncertainty analysis based on the ecological development forecast data and the field soil data to obtain an ecological restoration effect confidence level, and performing elasticity analysis based on the ecological restoration effect confidence level and the satellite remote sensing data to obtain an ecological resilience index;

[0084] In the actual assessment, the confidence levels of ecological restoration effects in Region A and Region B were 0.877 and 0.861, respectively; the ecological resilience indexes of Region A and Region B were 0.429 and 0.218, respectively;

[0085] Constructing a time-varying ecological restoration effect evaluation model based on the ecological resilience index, inputting the data to be analyzed into the time-varying ecological restoration effect evaluation model, and outputting an evaluation result;

[0086] In the actual evaluation, the repair effects of region A and region B are 0.607 and 0.211 respectively.

[0087] In this embodiment, the method for grading changes in the study area according to contrast to obtain graded areas includes:

[0088] The subgrids were obtained by dividing the grid according to the size and shape of the study area. The change features within the subgrids were extracted according to the change ratio of the remote sensing satellite data. The change features with a change ratio greater than 0.013 were taken as key features.

[0089] Taking the subgrid as a unit, a subgrid-based clustering feature tree is constructed according to the contrast and key features. The key features are used as clustering feature nodes, and the abnormality degree index of the clustering feature nodes is calculated.

[0090] The abnormality index is input into the student psychology optimization algorithm, abnormal clustering feature nodes are eliminated according to the optimization characteristics, the optimized key feature set is obtained, the clustering feature nodes in the optimized key feature set are integrated and adjusted, and the clustering feature tree is updated;

[0091] Calculate the contrast of the subgrid based on the temporal and spatial changes according to the updated cluster feature tree:

[0092]

[0093] The contrast of the w-th subgrid is l w , the number of key features of the optimized key feature set in the w-th sub-grid is N w , the vth key feature at time s2 is H v (s2), the vth distinguishing feature at the s1th moment is H v (s1), the weight coefficient of the vth key feature is ζ v , the sign function is sgn(·);

[0094] When the contrast is less than 0, the subgrid is a poor-quality restoration area; when the contrast is greater than 0 and less than 0.0051, the subgrid is a poor-quality restoration area; when the contrast is greater than 0.0051 and less than 0.015, the subgrid is a medium-quality restoration area; when the contrast is greater than 0.015 and less than 0.0297, the subgrid is a good-quality restoration area; when the contrast is greater than 0.0297, the subgrid is an excellent-quality restoration area.

[0095] Adjacent subgrids of the same level are connected according to the contrast, and the connected results are output as the graded area.

[0096] In this embodiment, the method of the spatiotemporal ecological development prediction model includes:

[0097] Collect climate and environmental data within the preset area, and perform time series impact analysis on the climate and environmental data to obtain environmental factors;

[0098] Spatiotemporal ecological development prediction models include graph networks, time-guided embeddings, and long short-term memory neural networks;

[0099] Graph networks represent pixels, regions, or objects in remote sensing data as graph nodes, and the relationships between graph nodes as edges. Graph neural network architecture is used to capture the complex relationships between graph nodes and extract the complex spatiotemporal features of remote sensing data based on these complex relationships.

[0100] Time-guided embedding incorporates temporal information into the embedding vector by learning spatiotemporal ecological development, and combines complex spatiotemporal features to capture the temporal cyclical pattern of ecological development;

[0101] Long short-term memory neural networks capture long-term dependencies in time series through gating mechanisms and cell states, and make ecological development predictions based on temporal cycle patterns and long-term dependencies.

[0102] In this embodiment, the method for performing restoration uncertainty analysis based on the ecological development prediction data and the field soil data to obtain the confidence level of the ecological restoration effect includes:

[0103] The ecological development prediction data and field soil data are converted into multidimensional vectors, and the ecological characteristic matrix is ​​constructed based on the multidimensional vectors;

[0104] Obtain ecological restoration effect indicators and ecological characteristics of the classified areas, and construct the least absolute shrinkage and selection operator Lasso regression model based on the ecological restoration effect indicators and ecological characteristics;

[0105] In the Lasso regression model, the square sum of the standardized coefficients of the ecological characteristics is used as the sensitivity coefficient, the non-zero frequency of the weight of the ecological characteristics in the Lasso regression model is obtained, and the importance of the ecological characteristics is calculated:

[0106]

[0107] The sensitivity coefficient of the wth ecological characteristic is χ w , the non-zero frequency of the wth ecological characteristic is F w , the importance of the wth ecological feature is φ w ;

[0108] The Pearson correlation coefficient was used to calculate the correlation between ecological characteristics, and ecological characteristics with correlation greater than 0.673 were regarded as key characteristics. The elbow rule was used to determine the optimal number of clusters for key characteristics, and the hierarchical regions were divided into multiple ecologically homogeneous groups according to the relationship characteristics;

[0109] Random sampling was performed on the ecologically homogeneous groups according to the proportion, and traditional ecological indicators were used for random sampling. A control group of the same area was set up to obtain the ecological environment quality score, soil quality index, and vegetation coverage of the ecologically homogeneous group and the control group. The ecological restoration effects of the ecologically homogeneous group and the control group were calculated based on the following:

[0110]

[0111] The ecological restoration effect of group c is The experimental area of ​​group c is S c , the vegetation coverage at time s2 is The vegetation coverage at time s1 is The ecological environment quality score at time s2 is The ecological environment quality score at time s1 is The soil quality index at time s2 is The soil quality index at time s1 is The weight coefficients are β1, β2, and β3;

[0112] Draw the temporal ecological restoration effect reference curve based on ecological characteristics and the temporal ecological restoration effect curve based on key characteristics, and calculate the difference between the temporal ecological restoration effect reference curve and the temporal ecological restoration effect curve:

[0113]

[0114] The number of features is N c , the i-th feature is b i , feature b at time s2 i The ecological restoration effect is Feature b at time s1 i The ecological restoration effect is referenced as The difference between the cth ecological homogeneous group is K c ;

[0115] The ratio of the difference of the sampling results to the mean of the difference is taken as the coefficient of variation. When the coefficient of variation is greater than 0.279, the difference between 1 and the coefficient of variation is output as the confidence level of the ecological restoration effect.

[0116] In this embodiment, the method for obtaining an ecological resilience index by performing resilience analysis based on the ecological restoration effect confidence level and the satellite remote sensing data includes:

[0117] A multidimensional ecological indicator system was constructed based on satellite remote sensing data. Resilience assessment indicators were screened based on the 4R theory, refining the indicators to identify key ecological functions according to the type of damage risk, eliminating redundant indicators based on the substitutability of species within the ecosystem, and selecting intelligent indicators based on the ability to self-regulate and mobilize recovery when faced with damage risks.

[0118] The ecological resilience change process is divided into the preparation stage, the resistance and absorption stage, the adaptive adjustment stage and the recovery stage according to time. Based on the performance indicators of each stage of the ecosystem and the detailed indicators of the 4R theory, a comprehensive ecological resilience measurement model is constructed.

[0119] Calculate the Ecological Resilience Index:

[0120]

[0121] The ecological resilience index of the kth graded area is The number of ecological resilience assessment indicators is M r , the weight of the rth indicator in the tth stage is ρ r (t), the elasticity evaluation index of the t stage is f r (t), the preparation stage is t1, the recovery stage is t4, and the confidence level of the ecological restoration effect of the kth graded area in the tth stage is The indicator weight of the preparation stage t1 is ρ r (t1), the elasticity evaluation index of the preparation stage t1 is f r (t1), the confidence level of ecological restoration effect of the kth graded area in the preparation stage t1 is

[0122] In this embodiment, the method for constructing a time-varying ecological restoration effect evaluation model based on the ecological resilience index includes:

[0123] Obtain the ecological performance evaluation indicators and corresponding weights of the study area, and construct the objective function based on the ecological resilience index and loss function. The expression is:

[0124]

[0125] The loss function is The initial ecosystem performance of the kth classification area is HO k , the number of graded areas in the study area is M k , the weight of the rth indicator at the sth moment is η r (s), the ecological performance evaluation index at the sth moment is g r (s), the ecosystem performance of the k-th classification area at the s-th moment is H k (s), the ecological resilience index of the kth graded area at the sth moment is The objective function at time s is

[0126] Time-varying ecological restoration effect evaluation models include random forest algorithms, time series analysis algorithms, and spatiotemporal graph neural networks;

[0127] The random forest algorithm uses an integrated learning algorithm to build multiple decision trees based on satellite remote sensing data, using randomly selected features and samples to construct multiple decision trees, and then determines the classification results through a voting mechanism.

[0128] The time series analysis algorithm is based on the time-space spectrum feature. It extracts the time, space and spectral features of the pixel points and combines them with the support vector machine classifier for change detection. The time series change pattern of the surface features is extracted based on the change detection.

[0129] The spatiotemporal graph neural network is constructed as a spatiotemporal graph based on the law of temporal changes. The spatiotemporal module of the graph neural network is used to model spatial and temporal dependencies respectively. The ecological restoration effect is predicted based on the dependency relationship, and the predicted ecological restoration effect is output.

[0130] The second aspect is an ecological restoration effect evaluation system based on satellite remote sensing, including:

[0131] Data acquisition module: used to collect satellite remote sensing data and field soil data of a preset area, and use the remote sensing data of the preset area within a specified time as the data to be analyzed; the data to be analyzed includes vegetation information, land use and land cover data, topographic data, hydrological data, thermal environment data and biodiversity data;

[0132] A partitioning and grading module is configured to calculate the time series deviation based on the satellite remote sensing data, and to define the preset area where the time series deviation is greater than 0 as the study area, and vice versa as the non-study area, and to classify the changes within the study area according to the contrast to obtain the graded area;

[0133] A correction prediction module is configured to correct the deviation of the satellite remote sensing data according to the field soil data in the classification area, and input the corrected satellite remote sensing data into the spatiotemporal ecological development prediction model to obtain ecological development prediction data;

[0134] Confidence elasticity module: used to perform restoration uncertainty analysis based on the ecological development prediction data and the field soil data to obtain the ecological restoration effect confidence, and to perform elasticity analysis based on the ecological restoration effect confidence and the satellite remote sensing data to obtain the ecological resilience index;

[0135] Modeling output module: used to construct a time-varying ecological restoration effect evaluation model based on the ecological resilience index, input the data to be analyzed into the time-varying ecological restoration effect evaluation model, and output the evaluation results.

[0136] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating ecological restoration effects based on satellite remote sensing, characterized in that: The following steps are involved: Collect satellite remote sensing data and field soil data of a preset area, and use the remote sensing data of the preset area within a specified time as data to be analyzed; the data to be analyzed includes vegetation information, land use and land cover data, topography data, hydrological data, thermal environment data, and biodiversity data; Calculating the time series deviation based on the satellite remote sensing data, taking the preset area where the time series deviation is greater than 0 as the study area, and otherwise as the non-study area, and classifying the changes in the study area according to the contrast to obtain a graded area; comprising: Calculate timing deviation: The timing deviation is The upper limit of monitoring time is , No. The ath remote sensing data at time is , No. The ath remote sensing data at time is The monitoring start time is , the altitude of the remote sensing satellite is , the radius of the earth is R, and the viewing angle is ; performing deviation correction on the satellite remote sensing data according to the field soil data in the grading area, and inputting the corrected satellite remote sensing data into a spatiotemporal ecological development prediction model to obtain ecological development prediction data; Performing restoration uncertainty analysis based on the ecological development forecast data and the field soil data to obtain an ecological restoration effect confidence level, and performing elasticity analysis based on the ecological restoration effect confidence level and the satellite remote sensing data to obtain an ecological resilience index; Constructing a time-varying ecological restoration effect evaluation model based on the ecological resilience index, inputting the data to be analyzed into the time-varying ecological restoration effect evaluation model, and outputting an evaluation result; The method for obtaining an ecological resilience index by performing resilience analysis based on the ecological restoration effect confidence level and the satellite remote sensing data includes: A multidimensional ecological indicator system was constructed based on satellite remote sensing data. Resilience assessment indicators were screened based on the 4R theory, refining the indicators to identify key ecological functions according to the type of damage risk, eliminating redundant indicators based on the substitutability of species within the ecosystem, and selecting intelligent indicators based on the ability to self-regulate and mobilize recovery when faced with damage risks. The ecological resilience change process is divided into the preparation stage, the resistance and absorption stage, the adaptive adjustment stage and the recovery stage according to time. Based on the performance indicators of each stage of the ecosystem and the detailed indicators of the 4R theory, a comprehensive ecological resilience measurement model is constructed. Calculate the Ecological Resilience Index: The ecological resilience index of the kth graded area is , the number of ecological resilience assessment indicators is , the weight of the rth indicator in the tth stage is , the elasticity evaluation index of the tth stage is The preparation stage is The recovery phase is , the confidence level of ecological restoration effect of the kth graded area in the tth stage is , preparation stage The indicator weight is , preparation stage The elasticity evaluation index is , preparation stage The confidence level of ecological restoration effect of the kth graded area is .

2. The method for evaluating ecological restoration effects based on satellite remote sensing according to claim 1, characterized in that: The method for obtaining graded regions by grading changes in the study area according to contrast comprises: The subgrids were obtained by dividing the grid according to the size and shape of the study area. The change features within the subgrids were extracted according to the change ratio of the remote sensing satellite data. The change features with a change ratio greater than 0.013 were taken as key features. Taking the subgrid as a unit, a subgrid-based clustering feature tree is constructed according to the contrast and key features. The key features are used as clustering feature nodes, and the abnormality degree index of the clustering feature nodes is calculated. The abnormality index is input into the student psychology optimization algorithm, abnormal clustering feature nodes are eliminated according to the optimization characteristics, the optimized key feature set is obtained, the clustering feature nodes in the optimized key feature set are integrated and adjusted, and the clustering feature tree is updated; Calculate the contrast of the subgrid based on the temporal and spatial changes according to the updated cluster feature tree: The contrast of the w-th subgrid is , the number of key features of the optimized key feature set in the w-th sub-grid is , No. The vth key feature at time is , No. The vth distinguishing feature at time is , the weight coefficient of the vth key feature is , the symbolic function is ; When the contrast is less than 0, the subgrid is a poor-quality restoration area; when the contrast is greater than 0 and less than 0.0051, the subgrid is a poor-quality restoration area; when the contrast is greater than 0.0051 and less than 0.015, the subgrid is a medium-quality restoration area; when the contrast is greater than 0.015 and less than 0.0297, the subgrid is a good-quality restoration area; when the contrast is greater than 0.0297, the subgrid is an excellent-quality restoration area. Adjacent subgrids of the same level are connected according to the contrast, and the connected results are output as the graded area.

3. The ecological restoration effect evaluation method based on satellite remote sensing according to claim 1, characterized in that: The method of the spatiotemporal ecological development prediction model comprises: Collect climate and environmental data within the preset area, and perform time series impact analysis on the climate and environmental data to obtain environmental factors; Spatiotemporal ecological development prediction models include graph networks, time-guided embeddings, and long short-term memory neural networks; Graph networks represent pixels, regions, or objects in remote sensing data as graph nodes, and the relationships between graph nodes as edges. Graph neural network architecture is used to capture the complex relationships between graph nodes and extract the complex spatiotemporal features of remote sensing data based on these complex relationships. Time-guided embedding incorporates temporal information into the embedding vector by learning spatiotemporal ecological development, and combines complex spatiotemporal features to capture the temporal cyclical pattern of ecological development; Long short-term memory neural networks capture long-term dependencies in time series through gating mechanisms and cell states, and make ecological development predictions based on temporal cycle patterns and long-term dependencies.

4. The method for evaluating ecological restoration effects based on satellite remote sensing according to claim 1, wherein: The method for performing restoration uncertainty analysis based on the ecological development prediction data and the field soil data to obtain the confidence level of the ecological restoration effect includes: The ecological development prediction data and field soil data are converted into multidimensional vectors, and the ecological characteristic matrix is ​​constructed based on the multidimensional vectors; Obtain ecological restoration effect indicators and ecological characteristics of the classified areas, and construct the least absolute shrinkage and selection operator Lasso regression model based on the ecological restoration effect indicators and ecological characteristics; In the Lasso regression model, the square sum of the standardized coefficients of the ecological characteristics is used as the sensitivity coefficient, the non-zero frequency of the weight of the ecological characteristics in the Lasso regression model is obtained, and the importance of the ecological characteristics is calculated: The sensitivity coefficient of the wth ecological characteristic is , the non-zero frequency of the wth ecological characteristic is , the importance of the wth ecological feature is ; The Pearson correlation coefficient was used to calculate the correlation between ecological characteristics, and ecological characteristics with correlation greater than 0.673 were regarded as key characteristics. The elbow rule was used to determine the optimal number of clusters for key characteristics, and the hierarchical regions were divided into multiple ecologically homogeneous groups according to the relationship characteristics; Random sampling was performed on the ecologically homogeneous groups according to the proportion, and traditional ecological indicators were used for random sampling. A control group of the same area was set up to obtain the ecological environment quality score, soil quality index, and vegetation coverage of the ecologically homogeneous group and the control group. The ecological restoration effects of the ecologically homogeneous group and the control group were calculated based on the following: The ecological restoration effect of group c is , the experimental area of ​​group c is , No. The vegetation coverage at the time is , No. The vegetation coverage at the time is , No. The ecological environment quality score at the moment is , No. The ecological environment quality score at the moment is , No. The soil quality index at the time is , No. The soil quality index at the time is , the weight coefficients are 、 、 ; Draw the temporal ecological restoration effect reference curve based on ecological characteristics and the temporal ecological restoration effect curve based on key characteristics, and calculate the difference between the temporal ecological restoration effect reference curve and the temporal ecological restoration effect curve: The number of features is , the i-th feature is , No. Moment Features The ecological restoration effect is , No. Moment Features The ecological restoration effect is referenced as , the difference of the cth ecological homogeneous group is ; The ratio of the difference of the sampling results to the mean of the difference is taken as the coefficient of variation. When the coefficient of variation is greater than 0.279, the difference between 1 and the coefficient of variation is output as the confidence level of the ecological restoration effect.

5. The ecological restoration effect evaluation method based on satellite remote sensing according to claim 1, characterized in that: The method for constructing a time-varying ecological restoration effect evaluation model based on the ecological resilience index includes: Obtain the ecological performance evaluation indicators and corresponding weights of the study area, and construct the objective function based on the ecological resilience index and loss function. The expression is: The loss function is , the initial ecosystem performance of the kth classification area is , the number of graded areas in the study area is , the weight of the rth indicator at the sth moment is , the ecological performance evaluation index at the sth moment is , the ecosystem performance of the kth graded area at the sth moment is , the ecological resilience index of the kth graded area at the sth moment is , the objective function at the sth moment is ; Time-varying ecological restoration effect evaluation models include random forest algorithms, time series analysis algorithms, and spatiotemporal graph neural networks; The random forest algorithm uses an integrated learning algorithm to build multiple decision trees based on satellite remote sensing data, using randomly selected features and samples to construct multiple decision trees, and then determines the classification results through a voting mechanism. The time series analysis algorithm is based on the time-space spectrum feature. It extracts the time, space and spectral features of the pixel points and combines them with the support vector machine classifier for change detection. The time series change pattern of the surface features is extracted based on the change detection. The spatiotemporal graph neural network is constructed as a spatiotemporal graph based on the law of temporal changes. The spatiotemporal module of the graph neural network is used to model spatial and temporal dependencies respectively. The ecological restoration effect is predicted based on the dependency relationship, and the predicted ecological restoration effect is output.

6. An ecological restoration effect evaluation system based on satellite remote sensing, used to implement the method according to any one of claims 1 to 5, characterized in that: include: Data acquisition module: used to collect satellite remote sensing data and field soil data of a preset area, and use the remote sensing data of the preset area within a specified time as the data to be analyzed; The data to be analyzed include vegetation information, land use and land cover data, topography data, hydrological data, thermal environment data and biodiversity data; A partitioning and grading module is configured to calculate the time series deviation based on the satellite remote sensing data, and to define the preset area where the time series deviation is greater than 0 as the study area, and vice versa as the non-study area, and to classify the changes within the study area according to the contrast to obtain the graded area; A correction prediction module is configured to correct the deviation of the satellite remote sensing data according to the field soil data in the classification area, and input the corrected satellite remote sensing data into the spatiotemporal ecological development prediction model to obtain ecological development prediction data; Confidence elasticity module: used to perform restoration uncertainty analysis based on the ecological development prediction data and the field soil data to obtain the ecological restoration effect confidence, and to perform elasticity analysis based on the ecological restoration effect confidence and the satellite remote sensing data to obtain the ecological resilience index; Modeling output module: used to construct a time-varying ecological restoration effect evaluation model based on the ecological resilience index, input the data to be analyzed into the time-varying ecological restoration effect evaluation model, and output the evaluation results.

Images