Ecological system service function evaluation method based on multi-source geographic spatio-temporal data

Abstract

The invention discloses an ecological system service function evaluation method based on multi-source geographic spatio-temporal data, which comprises the following steps: S1, collecting the multi-source geographic spatio-temporal data to form a geographic spatio-temporal data set; s2, encoding spatial position information and time sequence information in the geographic spatio-temporal data set, and forming a spatio-temporal association feature set through an association relationship of data in a spatio-temporal dimension; s3, inputting the time-space correlation feature set into the constructed ecosystem service function evaluation model, obtaining a nonlinear mapping relationship between different features and ecosystem service functions, and establishing quantitative correlation between the features and the ecosystem service functions; s4, optimizing the model by adopting a training strategy combining reinforcement learning and supervised learning; and S5, inputting the space-time correlation characteristics of the to-be-evaluated region into the trained ecological system service function evaluation model, and obtaining a quantitative evaluation result. According to the invention, comprehensive, accurate and dynamic evaluation of ecological system service functions is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of ecosystem service assessment, and in particular to an ecosystem service function assessment method based on multi-source geographic spatiotemporal data. Background Art

[0002] As global ecological and environmental challenges become increasingly severe, accurately assessing ecosystem service functions is crucial for ecological protection, resource management, and sustainable development planning. Good water conservation ensures a stable supply of water resources, soil conservation reduces soil erosion and maintains soil fertility, windbreaks and sand fixation are crucial for ecological security in arid and semi-arid regions, and biodiversity conservation helps maintain ecological balance and ecosystem stability. However, ecosystems are complex and dynamic systems, and their service functions are influenced by a variety of geographic and temporal factors, including land use type, ecosystem type, vegetation cover, topography, and meteorological and hydrological conditions. To accurately assess ecosystem service functions, a method for evaluating ecosystem service functions is needed.

[0003] Traditional methods for assessing ecosystem service functions rely primarily on data from ground-based monitoring stations, obtaining ecosystem-related indicators through field measurements and surveys. For example, when assessing soil conservation, soil erosion is measured at specific plots to infer soil conservation capacity for the entire region. While this approach can obtain relatively accurate data in local areas, it has significant limitations. Firstly, ground-based monitoring stations are sparsely distributed, making it difficult to cover large areas. This results in insufficient spatial representation of the data and an inability to reflect the spatial heterogeneity of ecosystem service functions within a region. Secondly, field measurements require significant human, material, and time resources, resulting in low efficiency and insufficient long-term dynamic monitoring and assessment of ecosystem service functions.

[0004] Existing technologies are beginning to utilize satellite remote sensing data for assessments, but these often simply rely on a single band or a combination of bands from remote sensing images to estimate indicators related to ecosystem services. For example, vegetation coverage is assessed solely based on vegetation indices, which infer biodiversity maintenance. While this approach improves data acquisition efficiency and spatial coverage, it neglects the complexity of ecosystems and the interplay of multiple factors, resulting in poor accuracy and reliability. For example, vegetation indices cannot accurately reflect the impact of topography on ecosystem services. In mountainous areas, topographical fluctuations affect conditions such as light and water, which in turn affect vegetation growth and the performance of ecosystem services, failing to effectively account for these factors.

[0005] At the same time, existing ecosystem service function assessment technologies also have shortcomings in data processing, model construction and computing efficiency, and cannot meet the requirements for comprehensive, accurate and dynamic assessment of ecosystem service functions. Summary of the Invention

[0006] Based on the above, this application discloses an ecosystem service function assessment method based on multi-source geographic spatiotemporal data, including:

[0007] S1. Collect multi-source geographic spatiotemporal data. The land use type, ecosystem type, and vegetation coverage obtained by classification of satellite remote sensing imagery data are combined with elevation, slope, and aspect data extracted from topographic data to form a geographic spatiotemporal dataset through data fusion technology.

[0008] S2. Encode the spatial location information and time series information in the geographic spatiotemporal dataset, obtain the spatiotemporal location encoding vector through the spatiotemporal association analysis model, and form a spatiotemporal association feature set through the association relationship of the data in the spatiotemporal dimension;

[0009] S3. Take the spatiotemporal correlation feature set as input, combine it with the type of ecosystem service function, and input it into the constructed ecosystem service function assessment model to obtain the nonlinear mapping relationship between different features and ecosystem service functions, and preliminarily establish the quantitative association between features and ecosystem service functions;

[0010] S4. Use a training strategy that combines reinforcement learning and supervised learning to conduct iterative training, so that the ecosystem service function assessment model can accurately predict ecosystem service functions based on spatiotemporal correlation characteristics;

[0011] S5. Input the spatiotemporal correlation characteristics of the area to be evaluated into the trained ecosystem service function evaluation model to obtain quantitative evaluation results of water conservation, soil conservation, windbreak and sand fixation, and biodiversity maintenance in each area.

[0012] Preferably, the multi-source geographic spatiotemporal data in S1 include high-resolution satellite remote sensing image data with spatial geographic information at different times, high-precision topographic data, rainfall, evapotranspiration, wind speed, runoff, soil type, soil moisture, soil physical and chemical properties, species list and species distribution data.

[0013] Preferably, the step of classifying the land use type and ecosystem type according to the satellite remote sensing image data in S1 includes: constructing a spatiotemporal fusion feature classification model to classify the satellite remote sensing image data through a fusion architecture of a multi-scale convolutional neural network and a spatiotemporal attention mechanism, wherein the formula is: in, For location The classification results at For location Multispectral image features at For location The spatiotemporal context features at is the time scale number, is the number of spatial attention heads, and The attention weights of time and space dimensions respectively, For feature fusion operations, weights are adaptively allocated through the spatiotemporal attention mechanism to obtain the temporal change characteristics and spatial correlation characteristics in image data at different times, and output classification results including land use types such as cultivated land, woodland, grassland, and water, as well as ecosystem types such as forest ecosystems, grassland ecosystems, and wetland ecosystems.

[0014] Preferably, the vegetation coverage is obtained by classifying high-resolution satellite remote sensing image data in S1, specifically:

[0015] The vegetation coverage is calculated by constructing a spatiotemporal spectral fusion vegetation index model. The formula is: ,in, For location The vegetation coverage, For location The converted soil adjusted vegetation index at is: , is the reflectivity in the near-infrared band, is the reflectivity of red light band, is the soil adjustment coefficient, is the noise correction coefficient, and are the TSI values ​​of pure soil and pure vegetation areas, is the temporal attention weight, It is a spatiotemporal attention mechanism used to obtain the temporal variation characteristics of vegetation growth at different times; the spatiotemporal spectral fusion vegetation index model adaptively enhances the weight of vegetation spectral features through the spatiotemporal attention mechanism to classify vegetation coverage under complex terrain.

[0016] Preferably, the S1 uses data fusion technology to form a geographic spatiotemporal dataset, and by constructing a multi-source data fusion framework of a spatiotemporal feature pyramid network, the land use type, ecosystem type and vegetation coverage data obtained by classification of satellite remote sensing images are structured with the elevation, slope and aspect data extracted from high-precision terrain and landform data to generate feature vectors under a unified spatiotemporal benchmark, and the attention mechanism is used to dynamically allocate weights of different data types. Through a hierarchical feature extraction and fusion strategy, the spatial topological features of the terrain data and the land feature attribute features of the remote sensing classification data are cross-coupled to form a geographic spatiotemporal dataset containing spatiotemporal attributes, land feature categories, and terrain features.

[0017] Preferably, the construction of the spatiotemporal correlation analysis model in S2 to obtain the spatiotemporal position coding vector is specifically as follows:

[0018] Construct a spatiotemporal dual attention encoding network to encode the spatial location information and time series information in the geographic spatiotemporal dataset. The formula is: ,in, For location At the moment The spatiotemporal position encoding vector of is the spatial position encoding function, which generates spatial position embedding through two-dimensional sine-cosine function. is the time position encoding function, which generates time position embedding through a periodic function; Indicates location At the moment The multi-source feature vectors of It is a spatial convolution operation used to extract local spatial features. Long short-term memory network to capture temporal dependencies; and They are spatial and temporal attention mechanisms, generating weight matrices through self-attention calculations. represents element-wise multiplication, It uses a multi-layer perceptron to perform feature fusion and nonlinear transformation; it encodes the complex spatiotemporal correlation in geographic spatiotemporal data through a spatiotemporal dual attention encoding network to obtain the spatiotemporal position encoding vector.

[0019] Preferably, the ecosystem service function evaluation model is constructed in S3, specifically:

[0020] A hierarchical heterogeneous attention fusion network is constructed based on the acquired spatiotemporal correlation feature set. The multi-scale spatiotemporal feature decomposition module is used to hierarchically decompose the spatiotemporal correlation features according to spatial scale and time granularity. Functional feature enhancement subnetworks are constructed for different ecosystem service functions such as water conservation, soil conservation, windbreak and sand fixation, and biodiversity maintenance. Each subnetwork adopts a dynamic graph convolution structure to adaptively adjust the association weights between nodes according to functional characteristics. The functional fusion features are calculated through the cross-functional attention fusion mechanism. The formula is: ,in For the The characteristics of ecosystem services, For the graph convolutional layers, is the spatiotemporal feature after splitting, The function-specific attention weights are used to form an ecosystem service function evaluation model based on a hierarchical heterogeneous network model that comprehensively evaluates multiple ecosystem service functions.

[0021] Preferably, in S4, the ecosystem service function evaluation model adopts a training strategy combining reinforcement learning and supervised learning to optimize the model structure. Through the dual-space parameter optimization framework, the actual observation data of the known ecosystem service function is used as the supervised learning label, and the mean square error between the model prediction result and the actual observation data is calculated as the supervised learning loss function. The reinforcement learning mechanism is used, the environmental state is used as the deviation between the model prediction result and the actual situation, and the action space is used as the model parameter adjustment strategy. The accuracy is evaluated by the reward function, so that the model continuously adjusts the parameters in a dynamic environment to optimize the model structure.

[0022] Preferably, the dual-space parameter optimization framework is used to optimize the model structure, and the formula is: ,in, are model parameters, Optimize model parameters, is the mean square error loss function for supervised learning, and the formula is: , is the actual observation data, Predict the results for the model; in the reinforcement learning part, the state is the mean absolute error of the prediction deviation, action is the parameter adjustment amount, is the reward function, is the balance coefficient, is the discount factor, is the total number of time steps, is the total number of samples.

[0023] Preferably, the quantitative assessment results of water conservation, soil conservation, windbreak and sand fixation, and biodiversity maintenance in each region are obtained in S5, and the assessment results are integrated with the geographic map in the form of a three-dimensional layer through the geographic information system and using three-dimensional visualization technology, and the size and distribution of the functional values ​​are distinguished by different colors, heights and textures.

[0024] Compared with the prior art, the technical solution of this application has the following technical effects:

[0025] The present invention collects multi-source geographic spatiotemporal data, covering various data types such as high-resolution satellite remote sensing images and high-precision topography, and uses the spatiotemporal fusion feature classification model to obtain accurate land use types and ecosystem types. It calculates vegetation coverage with the help of the spatiotemporal spectral fusion vegetation index model, and fuses multi-source data through the spatiotemporal feature pyramid network, making data processing comprehensive and accurate, and able to deeply mine the information behind the data and fully present the characteristics of the ecosystem in the spatiotemporal dimension.

[0026] The present invention constructs a hierarchical heterogeneous attention fusion network in the ecosystem service function evaluation model, splits spatiotemporal correlation features according to spatial scale and temporal granularity, constructs functional feature enhancement sub-networks for different ecosystem service functions, uses a dynamic graph convolution structure to adaptively adjust weights, and comprehensively evaluates multiple functions through a cross-functional attention fusion mechanism. It can effectively capture the nonlinear mapping relationship between different features and ecosystem service functions. Compared with traditional models, it is more efficient in processing complex ecosystem data, significantly improves the accuracy and adaptability of the evaluation model, and accurately quantifies the relationship between features and ecosystem service functions.

[0027] The present invention adopts a training strategy that combines reinforcement learning and supervised learning, using actual observation data as supervised learning labels and calculating the loss function through mean square error. In reinforcement learning, the prediction deviation is used as the environmental state, parameter adjustment is used as the action space, and the improvement of evaluation accuracy is used as the reward function. Under the dual-space parameter optimization framework, the model continuously adjusts parameters in a dynamic environment to achieve adaptive fusion of supervision signals and environmental feedback. This training strategy optimizes the model structure, enables the model to better fit actual data, enhances the model's generalization ability, and ensures that ecosystem service functions can be accurately predicted in different scenarios.

[0028] With the help of geographic information systems and three-dimensional visualization technology, this invention integrates the quantitative assessment results of water conservation, soil conservation and other functions in each region into three-dimensional layers and geographic maps, and uses different colors, heights and textures to distinguish the size and distribution of functional values, so as to facilitate decision makers to intuitively understand the spatial distribution characteristics of ecosystem service functions. At the same time, it can analyze the changing trends and evolution laws based on the assessment results at different times, providing intuitive and scientific decision-making basis for ecological protection, resource management and regional planning, and helping to achieve sustainable development of the ecosystem.

[0029] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application so that it can be implemented in accordance with the contents of the specification, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following is a detailed description of the preferred embodiment of the present application in conjunction with the accompanying drawings.

[0030] Based on the detailed description of the specific embodiments of the present application in conjunction with the accompanying drawings below, those skilled in the art will become more aware of the above and other objects, advantages and features of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without inventive work. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0032] Figure 1 This is a flow chart of an ecosystem service function assessment method based on multi-source geographic spatiotemporal data according to the present invention;

[0033] Figure 2 This is the multi-source data fusion framework structure diagram of the spatiotemporal feature pyramid network;

[0034] Figure 3 Schematic diagram of sample site selection for ecosystem service function assessment method;

[0035] Figure 4 Comparison chart of land use classification accuracy for ecosystem service function assessment methods;

[0036] Figure 5 A comparison curve of vegetation coverage between the present application and the prior art at sample point 1;

[0037] Figure 6 This is a comparison chart of the absolute errors of the present application and the prior art at sample point 1;

[0038] Figure 7 This is a comparison chart of the error indicators of the present application and the prior art at sample point 1. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. In the following description, specific details such as specific configurations and components are provided only to help fully understand the embodiments of the present application. Therefore, it should be clear to those skilled in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. In addition, for clarity and brevity, the description of known functions and structures has been omitted in the embodiments.

[0040] It should be understood that references throughout this specification to "one embodiment" or "this embodiment" mean that a particular feature, structure, or characteristic associated with the embodiment is included in at least one embodiment of the present application. Therefore, the appearance of "one embodiment" or "this embodiment" throughout this specification does not necessarily refer to the same embodiment. Furthermore, these particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0041] In addition, the present application may repeat reference numerals and / or letters in different examples. This repetition is for the purpose of simplicity and clarity and does not in itself indicate the relationship between the various embodiments and / or settings discussed.

[0042] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" in this article describes another type of association object relationship, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the previous and subsequent associated objects are in an "or" relationship.

[0043] The term "at least one" in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, at least one of A and B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0044] It should also be noted that, in this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include," "comprises," or any other variations thereof are intended to cover non-exclusive inclusion.

[0045] Example 1

[0046] This example mainly describes an ecosystem service function evaluation method based on multi-source geographic spatiotemporal data. Figure 1 Shown, including:

[0047] S1. Collect multi-source geographic spatiotemporal data. The land use type, ecosystem type, and vegetation coverage obtained by classification of satellite remote sensing imagery data are combined with elevation, slope, and aspect data extracted from topographic data to form a geographic spatiotemporal dataset through data fusion technology.

[0048] S2. Encode the spatial location information and time series information in the geographic spatiotemporal dataset, obtain the spatiotemporal location encoding vector through the spatiotemporal association analysis model, and form a spatiotemporal association feature set through the association relationship of the data in the spatiotemporal dimension;

[0049] S3. Take the spatiotemporal correlation feature set as input, combine it with the type of ecosystem service function, and input it into the constructed ecosystem service function assessment model to obtain the nonlinear mapping relationship between different features and ecosystem service functions, and preliminarily establish the quantitative association between features and ecosystem service functions;

[0050] S4. Use a training strategy that combines reinforcement learning and supervised learning to conduct iterative training, so that the ecosystem service function assessment model can accurately predict ecosystem service functions based on spatiotemporal correlation characteristics;

[0051] S5. Input the spatiotemporal correlation characteristics of the area to be evaluated into the trained ecosystem service function evaluation model to obtain quantitative evaluation results of water conservation, soil conservation, windbreak and sand fixation, and biodiversity maintenance in each area.

[0052] Furthermore, the multi-source geographic spatiotemporal data in S1 include high-resolution satellite remote sensing image data with spatial geographic information at different times, high-precision topographic and geomorphological data, rainfall, evapotranspiration, wind speed, runoff, soil type, soil moisture, soil physical and chemical properties, species list and species distribution data.

[0053] Furthermore, the land use types and ecosystem types obtained by classification of satellite remote sensing image data in S1 include: through the fusion architecture of multi-scale convolutional neural network and spatiotemporal attention mechanism, a spatiotemporal fusion feature classification model is constructed to classify satellite remote sensing image data. The formula is: in, For location The classification results at For location Multispectral image features at For location The spatiotemporal context features at is the time scale number, is the number of spatial attention heads, and The attention weights of time and space dimensions respectively, For feature fusion operations, weights are adaptively allocated through the spatiotemporal attention mechanism to obtain the temporal change characteristics and spatial correlation characteristics in image data at different times, and output classification results including land use types such as cultivated land, woodland, grassland, and water, as well as ecosystem types such as forest ecosystems, grassland ecosystems, and wetland ecosystems.

[0054] Furthermore, in S1, vegetation coverage is obtained by classifying high-resolution satellite remote sensing image data, specifically:

[0055] The vegetation coverage is calculated by constructing a spatiotemporal spectral fusion vegetation index model. The formula is: ,in, For location The vegetation coverage, For location The converted soil adjusted vegetation index at is: , is the reflectivity in the near-infrared band, is the reflectivity of red light band, is the soil adjustment coefficient, is the noise correction coefficient, and are the TSI values ​​of pure soil and pure vegetation areas, is the temporal attention weight, It is a spatiotemporal attention mechanism used to obtain the temporal variation characteristics of vegetation growth at different times; the spatiotemporal spectral fusion vegetation index model adaptively enhances the weight of vegetation spectral features through the spatiotemporal attention mechanism to classify vegetation coverage under complex terrain.

[0056] Furthermore, data fusion technology is used in S1 to form a geographic spatiotemporal dataset. By constructing a multi-source data fusion framework of a spatiotemporal feature pyramid network, the land use type, ecosystem type and vegetation coverage data obtained by satellite remote sensing image classification are structured with the elevation, slope and aspect data extracted from high-precision terrain and landform data to generate feature vectors under a unified spatiotemporal benchmark. The attention mechanism is used to dynamically allocate weights of different data types. Through a hierarchical feature extraction and fusion strategy, the spatial topological features of the terrain data and the land feature attribute features of the remote sensing classification data are cross-coupled to form a geographic spatiotemporal dataset containing spatiotemporal attributes, land feature categories and terrain features.

[0057] Furthermore, a spatiotemporal correlation analysis model is constructed in S2 to obtain the spatiotemporal position encoding vector, specifically:

[0058] Construct a spatiotemporal dual attention encoding network to encode the spatial location information and time series information in the geographic spatiotemporal dataset. The formula is: ,in, For location At the moment The spatiotemporal position encoding vector of is the spatial position encoding function, which generates spatial position embedding through two-dimensional sine-cosine function. is the time position encoding function, which generates time position embedding through a periodic function; Indicates location At the moment The multi-source feature vectors of It is a spatial convolution operation used to extract local spatial features. Long short-term memory network to capture temporal dependencies; and They are spatial and temporal attention mechanisms, generating weight matrices through self-attention calculations. represents element-wise multiplication, It uses a multi-layer perceptron to perform feature fusion and nonlinear transformation; it encodes the complex spatiotemporal correlation in geographic spatiotemporal data through a spatiotemporal dual attention encoding network to obtain the spatiotemporal position encoding vector.

[0059] Furthermore, an ecosystem service function evaluation model is constructed in S3, specifically:

[0060] A hierarchical heterogeneous attention fusion network is constructed based on the acquired spatiotemporal correlation feature set. The multi-scale spatiotemporal feature decomposition module is used to hierarchically decompose the spatiotemporal correlation features according to spatial scale and time granularity. Functional feature enhancement subnetworks are constructed for different ecosystem service functions such as water conservation, soil conservation, windbreak and sand fixation, and biodiversity maintenance. Each subnetwork adopts a dynamic graph convolution structure to adaptively adjust the association weights between nodes according to functional characteristics. The functional fusion features are calculated through the cross-functional attention fusion mechanism. The formula is: ,in For the The characteristics of ecosystem services, For the graph convolutional layers, is the spatiotemporal feature after splitting, The function-specific attention weights are used to form an ecosystem service function evaluation model based on a hierarchical heterogeneous network model that comprehensively evaluates multiple ecosystem service functions.

[0061] Furthermore, S4 uses a training strategy that combines reinforcement learning and supervised learning to optimize the model structure of the ecosystem service function evaluation model. Through the dual-space parameter optimization framework, the actual observation data of known ecosystem service functions are used as supervised learning labels, and the mean square error between the model prediction results and the actual observation data is calculated as the supervised learning loss function. The reinforcement learning mechanism is used, the environmental state is used as the deviation between the model prediction results and the actual situation, and the action space is used as the model parameter adjustment strategy. The accuracy is evaluated through the reward function, so that the model can continuously adjust the parameters in a dynamic environment to optimize the model structure.

[0062] Furthermore, the dual-space parameter optimization framework is used to optimize the model structure, and the formula is: ,in, are model parameters, Optimize model parameters, is the mean square error loss function for supervised learning, and the formula is: , is the actual observation data, Predict the results for the model; in the reinforcement learning part, the state is the mean absolute error of the prediction deviation, action is the parameter adjustment amount, is the reward function, is the balance coefficient, is the discount factor, is the total number of time steps, is the total number of samples.

[0063] Furthermore, S5 obtains quantitative assessment results of water conservation, soil conservation, windbreak and sand fixation, and biodiversity maintenance in each region. Through the geographic information system and using three-dimensional visualization technology, the assessment results are integrated with the geographic map in the form of a three-dimensional layer, and the size and distribution of functional values ​​are distinguished by different colors, heights, and textures.

[0064] This embodiment describes in detail how to collect multi-source geographic spatiotemporal data, classify, fuse and encode them, construct an evaluation model, and adopt a strategy training optimization that combines reinforcement learning with supervised learning to achieve accurate evaluation. It can comprehensively and accurately quantify ecosystem service functions and clearly present the status of functions such as water conservation and soil conservation in different regions. Through visual display, the evaluation results are intuitive and easy to understand, making it easier for decision makers to obtain information and providing a scientific basis for ecological protection and resource management.

[0065] Based on Example 1, this implementation describes in detail that this application constructs a multi-source data fusion framework of a spatiotemporal feature pyramid network to fuse and form a geographic spatiotemporal dataset containing spatiotemporal attributes, land feature categories, and terrain features, such as Figure 2 As shown, specifically:

[0066] A multi-source data fusion framework for a spatiotemporal feature pyramid network is constructed. Land use type, ecosystem type, and vegetation coverage data obtained through classification of satellite remote sensing images, as well as elevation, slope, and aspect data extracted from high-precision topographic and geomorphological data, are pre-processed. These data are then divided according to their spatiotemporal characteristics. A hierarchical architecture is used to design a spatiotemporal feature pyramid network. From the bottom layer to the top layer, the spatial resolution of the data gradually decreases, but the level of feature abstraction gradually increases. At the bottom layer, the original high-resolution data is directly accessed, and features are extracted using a convolutional neural network (CNN). For land use type data, CNN can capture the boundaries and texture features of different landforms. For topographic and geomorphological data, CNN can extract the undulating characteristics of the terrain. As the network layer increases, the data resolution is reduced through pooling operations, and skip connections are used to fuse the detailed features of the bottom layer with the abstract features of the high layer to form a multi-scale feature representation.

[0067] Achieve effective fusion of different types of data, unify all data to the same spatiotemporal reference, generate a unified feature vector, and make remote sensing classification data such as land use type, ecosystem type and vegetation cover spatially consistent with topographic data through coordinate conversion and resampling. Convert the coordinate system of remote sensing images to the same coordinate system as the topographic data, and resample the remote sensing data according to the resolution of the topographic data.

[0068] In the temporal dimension, for multi-source data with time series, interpolation or aggregation operations are performed at a uniform time interval. If the temporal resolution of some data is high and that of other data is low, the high-resolution data can be aggregated using mean, maximum, or minimum statistical methods to align them with the low-resolution data in time. The processed data are then combined into feature vectors according to certain rules. For example, the independent codes of land use types, the values ​​of vegetation coverage, elevation values, slope values, and aspect values ​​can be arranged in sequence to form a feature vector containing multiple types of information. In this way, the feature vector of each location can fully reflect the comprehensive information of that location in the temporal and spatial dimensions.

[0069] After data preprocessing and feature vector generation, the attention mechanism dynamically assigns weights to different data types. A hierarchical feature extraction and fusion strategy is then used to cross-couple the spatial topological features of terrain data with the feature attributes of remote sensing classification data. The core idea of ​​the attention mechanism is to allow the model to automatically learn the importance of different data features, thereby assigning higher weights to more important features. For example, in mountainous areas, where topography has a greater impact on ecosystems, the attention mechanism will assign higher weights to terrain data such as elevation, slope, and aspect. In plains, land use type and vegetation cover have a more significant impact on ecosystem services, so the weights of these data features will be increased accordingly.

[0070] The hierarchical feature extraction and fusion strategy gradually mines and fuses features in the data through multiple levels of processing. At the lower levels, it mainly extracts and fuses local features, such as performing convolution operations on feature vectors of adjacent positions to extract local spatial relationship features. As the level increases, the scope of feature fusion gradually expands, and features from different regions are comprehensively analyzed to mine more macroscopic spatiotemporal correlation features. In this way, the inherent connection between terrain data and remote sensing classification data can be fully explored, achieving deep fusion of features and forming a geographic spatiotemporal dataset that includes spatiotemporal attributes, land feature categories, and terrain characteristics.

[0071] This embodiment describes in detail how to fuse multi-source data by constructing a spatiotemporal feature pyramid network to generate a unified spatiotemporal benchmark feature vector. It also uses an attention mechanism and a hierarchical strategy to deeply fuse terrain and remote sensing data features. This can accurately integrate multi-source geographic spatiotemporal data, mine complex spatiotemporal associations, and form a comprehensive dataset. This provides an accurate and reliable data foundation for ecosystem service function assessment, thereby improving the accuracy and scientific nature of the assessment.

[0072] Based on Example 1, this implementation describes in detail the land use type and ecosystem type classification accuracy data and vegetation coverage calculation accuracy data of this application, specifically:

[0073] This application uses the fusion architecture of multi-scale convolutional neural network and spatiotemporal attention mechanism to build a spatiotemporal fusion feature classification model, and selects a large number of samples in the experimental area, such as Figure 3 As shown in the figure, classification work is carried out on satellite remote sensing image data, and sample 1 is selected, as shown in the figure. Figure 4 As shown, it can be seen that the classification accuracy of land use types is as high as 93.56%. Among them, there are 80 cultivated land samples, 74 of which are correctly classified; 65 forest land samples, 60 of which are correctly classified; in terms of ecosystem types, there are 50 forest ecosystem samples, 46 of which are correctly classified, with a classification accuracy of 92%; there are 45 grassland ecosystem samples, 42 of which are correctly classified, with a classification accuracy of 93.33%.

[0074] Using single satellite remote sensing image data and maximum likelihood classification for classification, in the same experimental area, sample 1 was selected, and the land use classification accuracy was only 70.32%. Woodland and grassland, due to their similar spectral characteristics, are easily misclassified using existing technologies, resulting in large deviations in classification results. The figure clearly shows that the curve of the proposed technology consistently outperforms existing technologies in terms of classification accuracy for different land use types. In terms of cultivated land classification accuracy, the proposed technology outperforms existing technologies by approximately 23.24%. The gap is even more significant in woodland classification.

[0075] For vegetation cover, Figure 5-7 As shown, this application constructs a spatiotemporal spectral fusion vegetation index model to calculate vegetation coverage, selects sample 1, and compares the field measurement results of 20 sample points with the model calculation results. The average absolute error is 0.038, the root mean square error is 0.053, and the correlation coefficient is as high as 0.925; for example, the field measurement vegetation coverage of sample point 1 is 0.583, the model calculation value is 0.615, and the absolute error is 0.032; the field measurement value is 0.502, the calculation value is 0.534, and the absolute error is 0.032. The overall error is controlled in an extremely small range, and the correlation coefficient curve of this application is highly consistent with the fitting curve.

[0076] Existing technologies are mostly based on simple vegetation index algorithms and rely only on a few bands of satellite remote sensing images to calculate vegetation coverage. In comparisons of the same sample points, the average absolute error is 0.156, the root mean square error is 0.201, and the correlation coefficient is only 0.701; taking sample point 1 as an example, the deviation between the vegetation coverage calculated by the existing technology and the actual value is significantly greater than that of the technology of this application. In terms of the root mean square error, the technology of this application is only about a quarter of the existing technology. This fully proves that the technology of this application is more accurate in calculating vegetation coverage, can more accurately reflect the actual vegetation coverage, and provide more accurate data support for the evaluation of ecosystem service functions such as biodiversity maintenance and soil conservation.

[0077] This embodiment describes in detail how the present application constructs a classification and vegetation coverage calculation model by integrating a multi-scale convolutional neural network with a spatiotemporal attention mechanism. The classification accuracy of land use and ecosystem types is significantly ahead of the existing technology. In the calculation of vegetation coverage, the mean absolute error and root mean square error are smaller, and the correlation coefficient is higher. It can provide accurate data and strongly support the evaluation of ecosystem service functions.

[0078] The above are only preferred embodiments of the present invention, which do not limit the scope of protection of the present invention. For those skilled in the art, the present invention can be modified and varied in various ways. Any changes, modifications, replacements, integrations and parameter changes to these embodiments through conventional substitutions or that can achieve the same functions without departing from the principles and spirit of the present invention fall within the scope of protection of the present invention.

Claims

1. A method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data, characterized in that: The following steps are involved: S1. Collect multi-source geographic spatiotemporal data. The land use type, ecosystem type, and vegetation coverage obtained by classification of satellite remote sensing imagery data are combined with elevation, slope, and aspect data extracted from topographic data to form a geographic spatiotemporal dataset through data fusion technology. S2. Encode the spatial location information and time series information in the geographic spatiotemporal dataset, obtain the spatiotemporal location encoding vector through the spatiotemporal association analysis model, and form a spatiotemporal association feature set through the association relationship of the data in the spatiotemporal dimension; S3. Take the spatiotemporal correlation feature set as input, combine it with the type of ecosystem service function, and input it into the constructed ecosystem service function assessment model to obtain the nonlinear mapping relationship between different features and ecosystem service functions, and preliminarily establish the quantitative association between features and ecosystem service functions; S4. Use a training strategy that combines reinforcement learning and supervised learning to perform iterative training, so that the ecosystem service function assessment model can accurately assess ecosystem service functions based on spatiotemporal correlation characteristics; S5. Input the spatiotemporal correlation characteristics of the area to be evaluated into the trained ecosystem service function evaluation model to obtain quantitative evaluation results of water conservation, soil conservation, windbreak and sand fixation, and biodiversity maintenance in each area.

2. The method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data according to claim 1, characterized in that: The multi-source geographic spatiotemporal data in S1 include high-resolution satellite remote sensing image data with spatial geographic information at different times, high-precision topographic data, rainfall, evapotranspiration, wind speed, runoff, soil type, soil moisture, soil physical and chemical properties, species list and species distribution data.

3. The method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data according to claim 1, characterized in that: The land use type and ecosystem type obtained by classifying satellite remote sensing image data in S1 include: constructing a spatiotemporal fusion feature classification model to classify satellite remote sensing image data through a fusion architecture of a multi-scale convolutional neural network and a spatiotemporal attention mechanism, and the formula is: in, For location The classification results at For location Multispectral image features at For location The spatiotemporal context features at is the time scale number, is the number of spatial attention heads, and The attention weights of time and space dimensions respectively, For feature fusion operations, weights are adaptively allocated through the spatiotemporal attention mechanism to obtain the temporal change characteristics and spatial correlation characteristics in image data at different times, and output classification results including land use types such as cultivated land, woodland, grassland, and water, as well as ecosystem types such as forest ecosystems, grassland ecosystems, and wetland ecosystems.

4. The method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data according to claim 1 or 3, characterized in that: In S1, the vegetation coverage is obtained by classifying high-resolution satellite remote sensing image data, specifically: The vegetation coverage is calculated by constructing a spatiotemporal spectral fusion vegetation index model. The formula is: ,in, For location The vegetation coverage, For location The converted soil adjusted vegetation index at is: , is the reflectivity in the near-infrared band, is the reflectivity of red light band, is the soil adjustment coefficient, is the noise correction coefficient, and are the TSI values ​​of pure soil and pure vegetation areas, is the temporal attention weight, It is a spatiotemporal attention mechanism used to obtain the temporal variation characteristics of vegetation growth at different times; the spatiotemporal spectral fusion vegetation index model adaptively enhances the weight of vegetation spectral features through the spatiotemporal attention mechanism to classify vegetation coverage under complex terrain.

5. The method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data according to claim 1, characterized in that: In the S1, data fusion technology is used to form a geographic spatiotemporal dataset. By constructing a multi-source data fusion framework of a spatiotemporal feature pyramid network, the land use type, ecosystem type and vegetation coverage data obtained by satellite remote sensing image classification are structured with the elevation, slope and aspect data extracted from high-precision terrain and landform data to generate feature vectors under a unified spatiotemporal benchmark. The attention mechanism is used to dynamically allocate weights of different data types. Through a hierarchical feature extraction and fusion strategy, the spatial topological features of the terrain data and the land feature attribute features of the remote sensing classification data are cross-coupled to form a geographic spatiotemporal dataset containing spatiotemporal attributes, land feature categories and terrain features.

6. The method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data according to claim 1, characterized in that: In S2, a spatiotemporal correlation analysis model is constructed to obtain a spatiotemporal position coding vector, specifically: Construct a spatiotemporal dual attention encoding network to encode the spatial location information and time series information in the geographic spatiotemporal dataset. The formula is: ,in, For location At the moment The spatiotemporal position encoding vector of is the spatial position encoding function, which generates spatial position embedding through two-dimensional sine-cosine function. is the time position encoding function, which generates time position embedding through a periodic function; Indicates location At the moment The multi-source feature vectors of It is a spatial convolution operation used to extract local spatial features. Long short-term memory network to capture temporal dependencies; and They are spatial and temporal attention mechanisms, generating weight matrices through self-attention calculations. represents element-wise multiplication, It uses a multi-layer perceptron to perform feature fusion and nonlinear transformation; it encodes the complex spatiotemporal correlation in geographic spatiotemporal data through a spatiotemporal dual attention encoding network to obtain the spatiotemporal position encoding vector.

7. The method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data according to claim 1, characterized in that: The ecosystem service function evaluation model is constructed in S3, specifically: A hierarchical heterogeneous attention fusion network is constructed based on the acquired spatiotemporal correlation feature set. The multi-scale spatiotemporal feature decomposition module is used to hierarchically decompose the spatiotemporal correlation features according to spatial scale and time granularity. Functional feature enhancement subnetworks are constructed for different ecosystem service functions such as water conservation, soil conservation, windbreak and sand fixation, and biodiversity maintenance. Each subnetwork adopts a dynamic graph convolution structure to adaptively adjust the association weights between nodes according to functional characteristics. The functional fusion features are calculated through the cross-functional attention fusion mechanism. The formula is: ,in For the The characteristics of ecosystem services, For the graph convolutional layers, is the spatiotemporal feature after splitting, The function-specific attention weights are used to form an ecosystem service function evaluation model based on a hierarchical heterogeneous network model that comprehensively evaluates multiple ecosystem service functions.

8. The method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data according to claim 1, characterized in that: In the S4, a training strategy combining reinforcement learning and supervised learning is used to optimize the model structure of the ecosystem service function evaluation model. Through a dual-space parameter optimization framework, actual observation data of known ecosystem service functions are used as supervised learning labels. The mean square error between the model prediction results and the actual observation data is calculated as the supervised learning loss function. The reinforcement learning mechanism is utilized, the environmental state is used as the deviation between the model prediction results and the actual situation, and the action space is used as the model parameter adjustment strategy. The accuracy is evaluated through the reward function, so that the model can continuously adjust the parameters in a dynamic environment to optimize the ecosystem service function evaluation model structure.

9. The method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data according to claim 8, characterized in that: The dual-space parameter optimization framework is used to optimize the ecosystem service function evaluation model structure, and the formula is: ,in, are model parameters, To optimize the model parameters, is the mean square error loss function for supervised learning, and the formula is: , is the actual observation data, Predict the results for the model; in the reinforcement learning part, the state is the mean absolute error of the prediction deviation, action is the parameter adjustment amount, is the reward function, is the balance coefficient, is the discount factor, is the total number of time steps, is the total number of samples.

10. The method for evaluating ecosystem service functions based on multi-source geographic spatiotemporal data according to claim 1, characterized in that: The S5 obtains the quantitative assessment results of water conservation, soil conservation, windbreak and sand fixation, and biodiversity maintenance in each region. Through the geographic information system and using three-dimensional visualization technology, the assessment results are integrated with the geographic map in the form of a three-dimensional layer and displayed, and the size and distribution of the functional values ​​are distinguished by different colors, heights, and textures.

Images