Nonlinear evolution path and characteristic evaluation method for organic carbon in regional soil and application of nonlinear evolution path and characteristic evaluation method

Abstract

The invention belongs to the technical field of soil science, and provides a regional soil organic carbon nonlinear evolution path and characteristic evaluation method and application, and the method comprises the steps: firstly, calculating environment variable data and soil organic carbon data of each pixel in a region based on multi-source remote sensing image data; classifying the states of the organic carbon in the soil by using a self-organizing mapping network; identifying a nonlinear evolution path of the soil organic carbon state type in the constructed state space; and finally, based on the nonlinear evolution path of the soil organic carbon, determining the hysteresis, toughness, mutation threshold and other nonlinear characteristics. According to the method, the nonlinear evolution path is reconstructed by using the soil organic carbon data in different spaces at the same time, so that the problem that the historical data of the soil organic carbon is relatively lacking is solved, and the hysteresis quality, toughness, mutation threshold and other nonlinear characteristics of the soil organic carbon evolution path are quickly and accurately quantified; trend prediction and situation early warning of the soil organic carbon are realized, so that management of the soil organic carbon is served.

Description

Technical Field

[0001] The present invention belongs to the technical field of soil science, and particularly relates to a method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon. Background Art

[0002] At present, in view of the important role of soil organic carbon in mitigating global climate change and ensuring food security, a number of initiatives and goals for enhancing soil organic carbon have been successively proposed globally. As the Earth enters the "Anthropocene", under human land use and management, the unpredictable non-linear mutation of soil organic carbon makes it difficult to conduct accurate trend warning for soil organic carbon management. Therefore, how to identify the complex evolution path of soil organic carbon and quantitatively characterize its non-linear evolution characteristics is an urgent problem to be solved in current soil organic carbon management.

[0003] Ecological regime shift refers to a large-scale, continuous and sudden change in the structure and function of an ecosystem, where the ecosystem rapidly reorganizes from a relatively stable state to another stable state. For the phenomenon of regime shift, two types of models, the S-shaped conceptual model and the ball-cup conceptual model, have been developed. Among them, the S-shaped model describes the bistability of the ecosystem and the phenomenon of regime shift under the change of external environmental conditions in a two-dimensional plane, while the ball-cup model demonstrates the process of regime shift under external condition disturbances. Therefore, through the conceptual model, the multi-stable range of the ecosystem and the critical points where regime shift occurs can be determined to achieve the analysis and prediction of the non-linear evolution characteristics of the ecosystem. These conceptual models provide an important basis for the evaluation of the non-linear evolution path and characteristics of soil organic carbon.

[0004] Although the S-shaped and ball-cup conceptual models can effectively describe the non-linear evolution characteristics of the ecosystem under the change of the external environment, these characteristics cannot be quantitatively expressed in the conceptual model, especially it is difficult to quantitatively express the non-linear evolution characteristics of soil organic carbon under human land use and management through data.

[0005] In addition, since the data of soil organic carbon often depend on soil profiles, it is difficult to obtain long-term continuous measured data and difficult to describe the evolution path of soil organic carbon. Therefore, the present invention provides a method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon and its application. Summary of the Invention

[0006] The purpose of the present invention is to overcome the existing defects, and provide a method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon and its application, which uses the soil organic carbon data at different spaces at the same time to reconstruct its non-linear evolution path, and quickly and accurately quantifies the non-linear characteristics such as the hysteresis, resilience and mutation threshold of the soil organic carbon evolution path, so as to achieve the trend prediction and situation warning of soil organic carbon, and thus serve the improvement management of soil organic carbon.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: The first object of the present invention is a method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon, including: Obtain multi-source remote sensing image data of the target area, and each pixel in the multi-source remote sensing image data represents surface reflectance data, surface temperature data, and elevation data at the corresponding position; Using a machine learning model, based on the surface reflectance data, surface temperature data, and elevation data of each pixel in the multi-source remote sensing image data, calculate the soil organic carbon content and its environmental variable data of each pixel; Using a self-organizing mapping network model, taking the soil organic carbon content and environmental variable data of each pixel as input vectors, divide all pixels in the region into several different soil organic carbon state types; Construct a two-dimensional state space according to the average cover function index and average soil organic carbon content of each soil organic carbon state type, and use quadratic functions to fit the degradation path curve and the recovery path curve respectively in the constructed state space to identify the non-linear evolution path of soil organic carbon; Based on the non-linear evolution path of soil organic carbon, determine the mutation threshold, hysteresis, and resilience of the soil organic carbon evolution path.

[0008] Further, the calculation steps of the environmental variable data include: (1) Based on the surface reflectance data of each pixel in the remote sensing image data, use spectral mixture decomposition to obtain the endmember abundance values of all endmembers of each pixel in the region, and the categories of the endmembers include vegetation, sand, salt, and dark matter; Input the endmember abundance values of each pixel into a pre-trained cover type discrimination model to obtain the cover type of the corresponding pixel; (2) Calculate the cover function index according to the following formula: ; where, , and are the endmember abundance values of sand, salt, and dark matter in spring respectively; is the endmember abundance value of vegetation in summer; LCI is the cover function index, and its value range is from 0 to 1; (3) Calculate the vegetation production rate index of the corresponding pixel according to the surface temperature data of each pixel; (4) Calculate the topographic wetness index of the corresponding pixel according to the elevation data of each pixel.

[0009] Further, the calculation steps of the soil organic carbon content include: Select a certain number of pixels in the remote sensing image data as samples, and measure the soil organic carbon content of each pixel in the samples using laboratory analysis methods; Randomly divide all the pixels in the samples into a training set and a validation set according to a proportion. Input the measured values of the soil organic carbon content and the environmental variable data of each pixel in the training set into the random forest model. Using the grid search method, determine the optimal parameters of the number of trees, maximum depth, and maximum number of features of the random forest model according to the results of 5-fold cross-validation; Take the determined optimal parameters as the set parameters of the random forest model to obtain the optimal random forest prediction model for the soil organic carbon content in the region; Input the environmental variable data corresponding to each pixel into the optimal random forest prediction model and run it several times. Determine the mean value of the running results of several times as the soil organic carbon content of the corresponding pixel, output the mean value map of the predicted soil organic carbon content, and determine the standard deviation of the running results of multiple times as the uncertainty value of the soil organic carbon content; Input the measured values of the soil organic carbon content and the environmental variable data of each pixel in the validation set into the optimal random forest prediction model to obtain the predicted values of the soil organic carbon content of each pixel in the validation set. Based on the measured values and predicted values of each pixel in the validation set, use two indicators, the coefficient of determination and the root mean square error, to evaluate the accuracy of the mapping result of the mean value map.

[0010] Furthermore, using the self-organizing mapping network model, taking the soil organic carbon content and environmental variable data of each pixel as input vectors, divide all the pixels in the region into several different soil organic carbon state types, specifically including: Set the scale of the self-organizing mapping network. Input the cover type, cover function index, vegetation production rate, terrain humidity index, and soil organic carbon content of all the pixels in the region into the self-organizing mapping network model to obtain several different soil organic carbon state types; In the setting of the scale of the self-organizing mapping network, first set the initial scale value of the self-organizing mapping network according to the empirical formula. The initial scale value is five times the square root of the total number of regional pixels; then select several values close to the initial scale value to obtain a set of values to be determined; calculate the quantization error and the topographic error for any value in the set of values to be determined, and determine the scale value of the self-organizing mapping model according to the calculation results.

[0011] Furthermore, in the constructed state space, use quadratic functions to fit the degradation path curve and the restoration path curve respectively to identify the non-linear evolution path of soil organic carbon, specifically including: Taking the average cover function index and the average soil organic carbon content of all pixels in each soil organic carbon state type as fitting sample points, the degradation path curve and the recovery path curve are respectively fitted using a quadratic function; According to the S-shaped conceptual model, the degradation path curve and the recovery path curve are connected by a dotted line to form a complete S-shaped evolution path.

[0012] Furthermore, for the non-linear evolution path based on soil organic carbon, determining the mutation threshold, hysteresis, and resilience of the soil organic carbon evolution path specifically includes: In the non-linear evolution path of soil organic carbon, the part with the same cover function index in the degradation path curve and the recovery path curve is defined as the unstable state, the part of the degradation path curve with a cover function index less than the unstable state is defined as the good stable state, and the part of the recovery path curve with a cover function index greater than the unstable state is defined as the poor stable state. The good mutation threshold and the poor mutation threshold are respectively determined according to the cover function index division values between the two stable states and the unstable state; The hysteresis of the soil organic carbon evolution path is determined according to the range between the two mutation thresholds; The resilience of the soil organic carbon evolution path is evaluated according to the degree of change in the cover function index required for any soil organic carbon state type in the unstable state to be converted into another achievable soil organic carbon state type in the stable state.

[0013] Furthermore, the calculation method of the good mutation threshold includes: ; where S1 is the set of pixels corresponding to the soil organic carbon state type with the largest average cover function index in the good stable state, LCI(x) is the cover function index corresponding to pixel x, and LCI1 is the pixel with the largest cover function index in the good stable state, that is, the largest LCI(x); ; where S2 is the set of pixels corresponding to the soil organic carbon state type with the smallest average cover function index in the unstable state on the recovery path curve, LCI(x) is the cover function index corresponding to pixel x, and LCI2 is the pixel with the smallest cover function index in the unstable state, that is, the smallest LCI(x); ; where LCIa is the good mutation threshold, μ1 and μ2 are respectively the average values of the cover function indices of the pixels corresponding to S1 and S2, σ1 and σ2 are respectively the standard deviations of the cover function indices of the pixels corresponding to S1 and S2, n ∈ (0.1, 0.2, 0.3,..., 2), and the value of n should be determined to satisfy the following conditions: 。

[0014] Furthermore, the calculation method of the mutation threshold of the difference includes: ; where S3 is the set of pixels corresponding to the type of soil organic carbon state with the largest average cover function index among the unstable states on the degradation path, LCI(x) is the cover function index corresponding to pixel x, and LCI3 is the pixel with the largest cover function index among the unstable states, that is, the largest LCI(x); ; where S4 is the set of pixels corresponding to the type of soil organic carbon state with the smallest average cover function index among the stable states of the difference, LCI(x) is the cover function index corresponding to pixel x, and LCI4 is the pixel with the smallest cover function index among the stable states of the difference, that is, the smallest LCI(x).

[0015] ; where LCIb is the mutation threshold of the difference, μ3 and μ4 are the average values of the cover function indices of the pixels corresponding to the state types S3 and S4 respectively, σ3 and σ4 are the standard deviations of the cover function indices of the pixels corresponding to the state types S3 and S4 respectively, n ∈ (0.1, 0.2, 0.3,..., 2), and the value of n should be determined to satisfy the following conditions: 。

[0016] Furthermore, the resilience of the soil organic carbon evolution path is evaluated according to the degree of change in the cover function index required for any soil organic carbon state type in the unstable state to be converted into another achievable stable state type of soil organic carbon state type, specifically including: Determine all soil organic carbon state types in the good or poor stable state that any soil organic carbon state type in the unstable state can be restored or degraded to as candidate states; In the candidate states, screen according to the principle of consistent cover type, and screen out the set of soil organic carbon state types with the same cover type as the current soil organic carbon state type; According to the principle of the closest topographic wetness index, determine the soil organic carbon state type in the achievable stable state corresponding to the current soil organic carbon state type in the set of soil organic carbon state types; Calculate the absolute value of the difference between the cover function indices of the current soil organic carbon state type and the achievable soil organic carbon state type, which is the resilience of the current soil organic carbon state type to be restored or degraded to a good or poor stable state.

[0017] Another object of the present invention is to provide an electronic device, comprising a processor and a memory storing a computer program, wherein when the processor executes the computer program, the method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon provided by the first object of the present invention is realized.

[0018] Another object of the present invention is to provide a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon provided by the first object of the present invention is realized.

[0019] Combined with the above technical solutions, the beneficial effects of the present invention compared with the prior art are as follows: Based on the principle of trading space for time, in the construction of the state space, the present invention uses the soil organic carbon data of different spaces at a certain time to reconstruct the state evolution path of soil organic carbon in the same space, effectively solving the problem that the evolution path cannot be constructed due to the lack of long-term measured data of current soil organic carbon (especially historical data).

[0020] The present invention can realize the rapid and high-precision mapping of soil organic carbon and its environmental variables based on limited measured sample point data and multi-source remote sensing data, accurately evaluate soil organic carbon and its dynamic changes, and provide basic data support for carbon neutrality. On this basis, the present invention can accurately identify the non-linear evolution path of soil organic carbon under land use and management, and quantify its non-linear characteristics such as mutation threshold, hysteresis and resilience, effectively solving the problem that the current soil organic carbon management lacks trend prediction and early warning methods. The trend prediction of soil organic carbon can be realized according to the evolution path of soil organic carbon, while the mutation threshold can further provide early warning information of soil organic carbon, identify the early warning area to avoid the huge restoration cost caused by its mutation, and the evaluation result of resilience determines the priority for soil organic carbon management, providing decision-making support for the management of soil organic carbon in China; in addition, the data sources used in the present invention are mainly remote sensing data. On the one hand, the spatial and temporal resolutions of remote sensing data are relatively high, which can greatly improve the accuracy and update frequency of monitoring results. At the same time, the remote sensing data used in the present invention are all publicly available free data products, which can greatly reduce the human and material costs consumed by field sampling; on the other hand, remote sensing data can cover areas that are difficult to reach by traditional field surveys, realizing full coverage of the region.

[0021] The present invention can accurately identify the non-linear evolution path of soil organic carbon under land use and management based on limited measured sample point data and multi-source remote sensing data, and quantify its non-linear characteristics such as mutation threshold, hysteresis and resilience, so as to realize the trend prediction and situation warning of soil organic carbon and serve the management of soil organic carbon. In addition, the model constructed by the present invention has good stability. After evaluating the evolution path and its characteristics, it can evaluate the state of soil organic carbon only relying on multi-source remote sensing data, greatly reducing the cost consumed by on-site sampling of soil organic carbon. Description of the Drawings

[0022] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. In the drawings: Figure 1 is a flowchart of the method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon provided by the embodiment of the present invention; Figure 2 is a schematic diagram of endmember extraction based on geometric vertices provided by the embodiment of the present invention Figure 3 is a schematic diagram of the self-organizing mapping network model provided by the embodiment of the present invention; Figure 4 is a schematic diagram of the principle of the method for reconstructing the evolution path of soil organic carbon provided by the embodiment of the present invention; Figure 5 is a schematic diagram of the state space provided by the embodiment of the present invention; Figure 6 is a schematic diagram of the S conceptual model provided by the embodiment of the present invention; Figure 7 is a schematic diagram of the state evolution path provided by the embodiment of the present invention; Figure 8 is a schematic diagram of state type identification provided by the embodiment of the present invention; Figure 9 is a schematic diagram of the evaluation of state non-linear characteristics provided by the embodiment of the present invention. Detailed Embodiments

[0023] The following describes the preferred embodiments of the present invention with reference to the drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0024] Embodiment 1: As Figure 1 shown, it is an embodiment of the method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon provided by the present invention, which specifically includes the following steps: S1: Obtain multi-source remote sensing image data of the target area. Each pixel in the multi-source remote sensing image data represents surface reflectance data, surface temperature data, and elevation data at the corresponding position. S2: Using a machine learning model, based on the surface reflectance data, surface temperature data, and elevation data of each pixel in the multi-source remote sensing image data, calculate the soil organic carbon content and its environmental variable data of each pixel. The environmental variable data includes cover type, cover function index, vegetation production rate, and terrain humidity index. S3: Using a self-organizing mapping network model, with the soil organic carbon content and environmental variable data of each pixel as input vectors, divide all pixels in the region into several different soil organic carbon state types. S4: Construct a two-dimensional state space based on the average cover function index and average soil organic carbon content of each soil organic carbon state type. In the constructed state space, use quadratic functions to fit the degradation path curve and recovery path curve respectively, and identify the non-linear evolution path of soil organic carbon. S5: Based on the non-linear evolution path of soil organic carbon, determine the mutation threshold, hysteresis, and resilience of the soil organic carbon evolution path.

[0025] Preferably, the calculation steps of the environmental variable data in the embodiments of the present invention include: (1) Based on the surface reflectance data of each pixel in the remote sensing image data, use spectral mixture decomposition to obtain the endmember abundance values of all endmembers of each pixel in the region. The categories of the endmembers include vegetation, sand, salt, and dark matter. Input the endmember abundance values of each pixel into a pre-trained cover type discrimination model to obtain the cover type of the corresponding pixel. Specifically, according to the land cover classification system (such as forest land, grassland, cultivated land, construction land, and water area, etc.), select about 1000 samples for each land cover category through high-spatial-resolution images of Google Earth, and divide them into training samples and validation samples at a ratio of 75% and 25%. Input the training samples, their corresponding endmember abundance values, and the entire endmember abundance value image obtained by spectral mixture decomposition into a random forest model for cover type classification. Perform accuracy evaluation according to the cover type of the validation samples and the corresponding classification results. The evaluation indicators include overall accuracy p e and Kappa coefficient k. The calculation formulas are as follows:

[0026] ; Wherein, It is the sum of the number of correctly classified samples of each land cover category divided by the total number of samples, that is, the overall classification accuracy. , … is the number of true samples of each land cover category, and , … is the number of samples of each land cover category in the prediction results, and the total number of samples is n.

[0027] It should be noted that spectral mixture decomposition mainly includes three steps: principal component analysis, endmember extraction based on geometric vertices, and calculation of the endmember abundance value of each pixel in the study area.

[0028] First, the months with the highest vegetation coverage (such as August) and the lowest surface coverage (such as May) were determined based on local knowledge in the study area. The Landsat surface reflectance data images of these months were subjected to principal component analysis using ENVI software to obtain the transformed principal component images. The types and numbers of end members were determined based on existing research in the region. In arid areas, monitoring is generally carried out on four types of end members: vegetation (GV), sand (SL), salt (SA), and dark matter (DA). Vegetation (GV) end members were extracted in August, and sand (SL), salt (SA), and dark matter (DA) end members were extracted in May.

[0029] Then, if Figure 2 As shown in the figure, the two-dimensional scatter plot extraction method is used to extract the endmember spectral curve. Based on the mathematical mechanism of the linear mixed model, the scatter plot of the principal component should be triangularly distributed. The three vertices of the triangle can be regarded as pure endmembers, and the interior of the triangle is a mixed pixel, which is a linear combination of pure endmembers. In the ENVI software, the principal component image of the extraction month corresponding to the endmember is used to construct a scatter plot. 200-400 pixels are selected as endmember pixels at the vertices of the principal component scatter plot, and the average reflectance of the original image at the location of the endmember pixel is used as the endmember spectrum.

[0030] Finally, after obtaining the endmember spectra in the region, the endmember spectra and regional reflectance images were used to perform linear spectral mixed decomposition in the ENVI software to obtain the endmember abundance values ​​of each endmember in the study area.

[0031] Linear spectral mixture decomposition is the main method for estimating abundance values, and its expression is: ; Among them, i is the remote sensing image band, is the reflectivity of band i, m is the number of end members, for each end member j (1≤j≤m), is the reflectivity of end member j in band i, is the abundance value of endmember j, that is, the percentage of the endmember in the pixel area. is the residual. In practice, in order to make the estimated abundance value have physical significance, generally the fully constrained least squares method can be used to obtain the abundance value of each endmember in each pixel, and the corresponding abundance value image is obtained.

[0032] 。

[0033] (2)Calculate the cover function index according to the following formula: ; Among them, 、 and are the endmember abundance values of sand, salt, and dark matter in spring, respectively; is the endmember abundance value of vegetation in summer; LCI is the cover function index, and its value range is from 0 to 1.

[0034] (3)Calculate the vegetation production rate index of the corresponding pixel according to the surface temperature data of each pixel. The calculation formula is as follows: ; Among them, VPI represents the vegetation production rate index, is a function of the reference surface temperature curve, is a function of the pixel surface temperature curve, and are the start and end times of the main growing season, respectively; (4)Calculate the topographic wetness index of the corresponding pixel according to the elevation data of each pixel. The calculation formula is as follows: ; Among them, TWI represents the topographic wetness index, is the upstream area of the unit contour line where surface water flows through, is the slope.

[0035] Preferably, the calculation steps of the soil organic carbon content in the embodiments of the present invention include: (1)Select a certain number of pixels in the remote sensing image data as samples, and measure the soil organic carbon content of each pixel in the samples by laboratory analysis method; (2)Randomly divide all the pixels in the samples into a training data set and a validation data set according to the ratio of 75% and 25%. In the training data set, adopt the grid search method, and determine the optimal parameter combination of the number of trees, maximum depth, and maximum number of features of the random forest model according to the results of 5-fold cross-validation, and use the determined optimal parameter combination as the setting parameters of the random forest model; The specific operation is: 1) Preset the number of trees in the random forest model according to experience to be from 100 to 1000, with an interval of 50; the maximum depth to be from 4 to 10, with an interval of 1; and the maximum number of features to be from 2 to 5, with an interval of 1. 2) Grid search method, that is, combine all possible values of the three parameters within the preset range. 3) Then perform 5-fold cross-validation for each combination of parameter values: divide the training data set into 5 samples, successively use each sample as the validation sample, and the remaining four as the training samples, and calculate the average determination coefficient R 2 and the root mean square error RMSE, and select the highest value to determine the optimal model parameters.

[0036]

[0037] ; where and are respectively the predicted value and the measured value of the soil organic carbon content of pixel i, is the average value of the measured values of the soil organic carbon content of all pixels, and n is the number of pixels.

[0038] Using the random forest model of the "rpart" package in R language, input the measured values of the soil organic carbon content of each pixel in the training set and the environmental variable data, and according to the selected optimal model parameters as the setting parameters of the random forest model, the optimal random forest prediction model of the soil organic carbon content of each pixel in the study area can be obtained.

[0039] (3) Next, input the environmental variable data set into the optimal random forest prediction model, take the mean value of the results of running the optimal random forest prediction model 100 times as the final prediction result, output the soil organic carbon mean value map, and use the determination coefficient R 2 and the root mean square error RMSE to evaluate the accuracy of the mapping result, and at the same time use the standard deviation of the 100 predicted values for the evaluation of model uncertainty.

[0040] Preferably, in the embodiment of the present invention, using the self-organizing mapping network model, with the soil organic carbon content and environmental variable data of each pixel as the input vector, all pixels in the region are divided into several different soil organic carbon state types, specifically including: Set the scale of the self-organizing mapping network, input the cover type, cover function index, vegetation production rate, terrain humidity index and soil organic carbon content of all pixels in the region into the self-organizing mapping network model, and obtain several different soil organic carbon state types.

[0041] In setting the scale of the self-organizing mapping network, first, according to the empirical formula, set the initial scale value of the self-organizing mapping network. The initial scale value is five times the square root of the total number of regional pixels. Then, select several values close to the initial scale value to obtain a set of values to be determined. For any value in the set of values to be determined, calculate the quantity error and the structure error respectively, and determine the scale value of the self-organizing mapping model according to the calculation results.

[0042] Specifically, as Figure 3 shown. In the self-organizing mapping network, first, according to the Vesanto empirical formula, use to estimate the initial value of the scale of the self-organizing mapping network: ; where M is the scale value of the self-organizing mapping network and S is the number of samples; Suppose the number of samples is 253, then the initial scale is 80. Then, select several values on the left and right of the initial scale value to obtain a set of values to be determined. Since the self-organizing mapping network is a two-dimensional network, according to experience, set 3 scale combinations on the left and right of the initial scale 80. Therefore, the scales are preset as 10×7, 9×8, 11×7, 10×8, 11×8, 10×9, 12×8. Finally, for any value in the set of values to be determined, calculate the quantity error QE and the structure error TE under each scale respectively, and determine the scale value of the self-organizing mapping model (i.e., the number of types of soil organic carbon status) according to the calculation results.

[0043] ; where Num is the total number of samples, is the i-th input sample, is the sample corresponding to the weight vector of the best matching neuron, is the Euclidean distance between the two; ; where Num is the total number of samples, and are the first neighboring neuron and the second neighboring neuron of the sample respectively. When and are not adjacent, return 1, otherwise return 0.

[0044] In the embodiment of the present invention, in the two-dimensional state space constructed based on the average coverage function index and the average soil organic carbon content of each soil organic carbon state type, the construction of the state space is mainly based on the method of trading space for time. It is assumed that there are enough soil organic carbon state types in the space and the evolution time of the soil organic carbon state types is long enough. The set of soil organic carbon state types at different spatial positions at the same time is equivalent to the set of soil organic carbon state types at the same spatial position at different times. As Figure 4 shown, this principle is like reconstructing the set of plants at different growth stages within the same time into the set of different growth stages of the same plant. Therefore, the state evolution path of soil organic carbon at different times on the same space can be reconstructed using the soil organic carbon state types in different spaces at the same time.

[0045] According to the determined soil organic carbon state types, a two-dimensional state space of soil organic carbon is constructed with the average coverage function index of all pixels in each soil organic carbon state type as the X-axis and the average soil organic carbon content of all pixels in each soil organic carbon state type as the Y-axis. As Figure 5 shown, the colored points represent the corresponding positions of each soil organic carbon state type in the state space.

[0046] Preferably, in the constructed state space of the embodiment of the present invention, quadratic functions are used to fit the degradation path curve and the recovery path curve respectively. The specific steps for identifying the non-linear evolution path of soil organic carbon include: According to the S-shaped conceptual model, quadratic functions are used to fit the curves of the degradation path and the recovery path in the state space. The quadratic function fitting uses the polyfit function in the numpy library of the python language. The average coverage function index and the average soil organic carbon content of all pixels in each soil organic carbon state type obtained from the above self-organizing mapping network are used as the fitting sample points, and the function form is set as follows: ; where y is the soil organic carbon content, x is the coverage function index, and a, b, and c are coefficients; the polyfit function can solve for each coefficient using the least squares method.

[0047] It should be noted that when the external environmental conditions change, the system state of the land will respond accordingly and present bistability. Figure 6The solid lines in it represent the steady states. The green and red respectively represent the best and worst steady states under the corresponding environmental conditions. The dashed lines represent unstable non-equilibrium states, and the orange represents multiple unstable states under the same environmental conditions. The upper half of the S-shaped curve represents the degradation path. As the external environmental conditions gradually degrade to near the critical level, the system will undergo a sudden change from one steady state to another through a threshold, that is, a steady-state transition. The points where the mutation occurs are called critical points (P1 and P2). However, this forward steady-state transition path is not completely reversible, and there is an obvious hysteresis in the land system, that is, even if the external environmental conditions are restored to the level before the mutation, the land system cannot return to its previous state.

[0048] As Figure 7 shown, the quadratic function curve fitted to the green points in the upper half of the state space is defined as the degradation path. Along with the increase in the cover function index (the worse the natural environment and human management), the state of soil organic carbon degrades along the solid green line to the right (that is, the content of soil organic carbon gradually decreases), and as the cover function index becomes larger and larger (the worse the natural environment and human management), the degradation of the organic carbon state (soil organic carbon content) caused by the increase in the unit cover function index gradually increases. Until a certain threshold is reached and a mutation occurs, that is, Figure 7 the path of the green dashed line in

[0049] Use the same method to identify the restoration path of soil organic carbon. As Figure 7 shown, the quadratic function curve fitted to the orange points in the lower half of the state space is defined as the restoration path. Along with the decrease in the cover function index (the better the natural environment and human management), the state of soil organic carbon recovers along the solid orange line to the left (that is, the content of soil organic carbon gradually increases), and as the cover function index becomes smaller and smaller (the better the natural environment and human management), the recovery of the organic carbon state (soil organic carbon content) caused by the decrease in the unit cover function index gradually increases. Until a certain threshold is reached and a mutation occurs, that is, Figure 7 the path of the orange dashed line in

[0050] As Figure 8 shown, according to the S-shaped conceptual model, the degradation path curve and the restoration path curve are connected with dashed lines to form a complete S-shaped evolution path.

[0051] Preferably, in the embodiments of the present invention, based on the non-linear evolution path of soil organic carbon, the specific steps for determining the mutation threshold, hysteresis and resilience of the soil organic carbon evolution path include: In the non-linear evolution path of soil organic carbon, the part with the same cover function index in the degradation path curve and the restoration path curve is defined as the unstable state. The part of the degradation path curve where the cover function index is less than the unstable state is defined as the good stable state, and the part of the restoration path curve where the cover function index is greater than the unstable state is defined as the poor stable state. The good mutation threshold and the poor mutation threshold are respectively determined according to the cover function index division values between the two stable states and the unstable state; The hysteresis of the soil organic carbon evolution path is determined according to the range between the two mutation thresholds; The resilience of the soil organic carbon evolution path is evaluated according to the degree of change in the cover function index required for any soil organic carbon state type in the unstable state to be converted into the soil organic carbon state type in another achievable stable state.

[0052] Specifically, as Figure 8 shown, in the S-shaped path, the soil organic carbon state types located on the orange path in the folding interval are defined as the unstable state. In this interval, the same cover function index can correspond to multiple soil organic carbon state types with different soil organic carbon contents. When the natural environment or human management changes slightly, the soil organic carbon state types will transform along the path. Among them, the states on the dotted path are non-equilibrium states, and these states are only instantaneous states existing at the observation moment and will eventually evolve to the equilibrium states on the solid path. Therefore, these non-equilibrium states are not included in the evolution path. On the left side of the folding interval of the S-shaped path, the soil organic carbon state types on the green path represent the good stable state, while on the right side of the folding interval, the soil organic carbon state types on the red path represent the poor stable state. Compared with the unstable state, when the natural environment or human management changes slightly, the soil organic carbon state types in these stable states can rely on their own resilience to resist these interferences and maintain the stability of the state.

[0053] The evaluation of the non-linear characteristics includes the quantitative characterization of mutation thresholds, hysteresis, and resilience. As Figure 9 shown, the mutation threshold refers to the point at which, as the cover function index (natural environment or human management) changes, the state of soil organic carbon will gradually degrade or recover to the critical level and mutate into another stable state, and the cover function index value corresponding to this mutation point is called the mutation threshold. The mutation threshold includes the good mutation threshold ( Figure 9 LCIa in Figure 9 ) and the poor mutation threshold ( LCIb in

[0054] ). These thresholds are determined by the cover function index division values between the two stable states and the unstable state. Among them, S1 is the set of pixels corresponding to the soil organic carbon state type with the largest average cover function index in the good stable state, LCI(x) is the cover function index corresponding to pixel x, and LCI1 is the largest LCI(x); ; Among them, S2 is the set of pixels corresponding to the soil organic carbon state type with the smallest average cover function index in the unstable state on the recovery path, LCI(x) is the cover function index corresponding to pixel x, and LCI2 is the smallest LCI(x); ; Among them, LCIa is the good mutation threshold, μ1 and μ2 are the average values of the cover function indices of the pixels corresponding to S1 and S2 respectively, σ1 and σ2 are the standard deviations of the cover function indices of the pixels corresponding to S1 and S2 respectively, n ∈ (0.1, 0.2, 0.3,..., 2), and the value of n should be determined to satisfy the following conditions: .

[0055] The calculation method of the poor mutation threshold includes: ; Among them, S3 is the set of pixels corresponding to the soil organic carbon state type with the largest average cover function index in the unstable state on the degradation path, LCI(x) is the cover function index corresponding to pixel x, and LCI3 is the largest LCI(x); ; Among them, S4 is the set of pixels corresponding to the soil organic carbon state type with the smallest average cover function index in the poor stable state, LCI(x) is the cover function index corresponding to pixel x, and LCI4 is the smallest LCI(x).

[0056] ; Among them, LCIb is the poor mutation threshold, μ3 and μ4 are the average values of the cover function indices of the pixels corresponding to state types S3 and S4 respectively, σ3 and σ4 are the standard deviations of the cover function indices of the pixels corresponding to state types S3 and S4 respectively, n ∈ (0.1, 0.2, 0.3,..., 2), and the value of n should be determined to satisfy the following conditions: .

[0057] Hysteresis means that the degradation path of the mutation of the soil organic carbon state is irreversible. When the cover function index (natural environment or human management) returns to the level before the mutation, the state of the soil organic carbon cannot return to the original state. The level of hysteresis depends on the range between the two mutation thresholds, that is Figure 9The absolute value of the difference between LCIa and LCIb in it. The larger this range is, the stronger the hysteresis is, and the more difficult it is for soil organic carbon to return to its previous state after a mutation occurs.

[0058] Resilience refers to the degree of change in the cover function index required for the current state of soil organic carbon to transition to another achievable stable state. For the achievable states corresponding to a certain state of soil organic carbon, in addition to being affected by the evolution path, the influence of its cover type and topographic wetness index should also be considered. As Figure 9 shown, taking the restoration of soil organic carbon state type B to soil organic carbon state type A1 in a good stable state as an example: (1) First, determine all the good stable states to which soil organic carbon state type B may be restored as candidate states; (2) Among the candidate states, screen out the set of soil organic carbon state types with the same cover type as soil organic carbon state type B according to the principle of consistent cover type; (3) In the screened set of soil organic carbon state types, determine soil organic carbon state type A1 in the achievable stable state corresponding to soil organic carbon state type B according to the principle of the closest topographic wetness index; (4) Calculate the absolute value of the difference between the cover function indices of soil organic carbon state type B and soil organic carbon state type A1, which is the resilience of soil organic carbon state type B to restore to a good stable state. The calculation method for the resilience of the state degenerating into a poor stable state C1 is the same as the above method.

[0059] Example 2: An embodiment of the present invention provides an electronic device, including a processor and a memory storing a computer program. When the processor executes the computer program, it implements the method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon provided in Embodiment 1 of the present invention.

[0060] Example 3: Another object of the present invention is to provide a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon provided in Embodiment 1 of the present invention.

[0061] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown in sequence according to the indications of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in the present invention, the execution of these steps has no strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same moment, but can be executed at different moments. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0062] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided by the present invention can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0063] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon, characterized in that The method includes: Obtaining multi-source remote sensing image data of the target area, where each pixel in the multi-source remote sensing image data represents surface reflectance data, surface temperature data, and elevation data at the corresponding position; Using a machine learning model, based on the surface reflectance data, surface temperature data, and elevation data of each pixel in the multi-source remote sensing image data, calculating the soil organic carbon content and its environmental variable data of each pixel; Using a self-organizing mapping network model, with the soil organic carbon content and environmental variable data of each pixel as input vectors, dividing all pixels in the area into several different soil organic carbon state types; Constructing a two-dimensional state space based on the average cover function index and average soil organic carbon content of each soil organic carbon state type, and using quadratic functions to fit the degradation path curve and recovery path curve respectively in the constructed state space to identify the non-linear evolution path of soil organic carbon; Based on the non-linear evolution path of soil organic carbon, determining the mutation threshold, hysteresis, and resilience of the soil organic carbon evolution path.

2. The method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon according to claim 1, wherein The calculation steps of the environmental variable data include: (1) Based on the surface reflectance data of each pixel in the remote sensing image data, using spectral mixture decomposition to obtain the endmember abundance values of all endmembers of each pixel in the area, and the categories of the endmembers include vegetation, sand, salt, and dark matter; Inputting the endmember abundance values of each pixel into a pre-trained cover type discrimination model to obtain the cover type of the corresponding pixel; (2) Calculating the cover function index according to the following formula: ； Among them, , and are the end-member abundance values of sand, salt, and dark matter in spring, respectively; is the end-member abundance value of vegetation in summer; LCI is the cover function index, and its value range is from 0 to 1. (3) Calculating the vegetation production rate index of the corresponding pixel according to the surface temperature data of each pixel; (4) Calculating the topographic wetness index of the corresponding pixel according to the elevation data of each pixel.

3. The method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon according to claim 1, characterized in that The calculation steps of the soil organic carbon content include: Selecting a certain number of pixels in the remote sensing image data as samples, and measuring the soil organic carbon content of each pixel in the samples using laboratory analysis methods; Randomly dividing all pixels in the samples into a training set and a validation set according to a ratio, inputting the measured values of the soil organic carbon content and environmental variable data of each pixel in the training set into a random forest model, and using the grid search method to determine the optimal parameters of the number of trees, maximum depth, and maximum number of features of the random forest model according to the results of 5-fold cross-validation; Taking the determined optimal parameters as the set parameters of the random forest model to obtain the optimal random forest prediction model of the soil organic carbon content in the area; Inputting the environmental variable data corresponding to each pixel into the optimal random forest prediction model and running it several times, determining the mean value of the running results of several times as the soil organic carbon content of the corresponding pixel, outputting the mean value map of the predicted soil organic carbon content, and determining the standard deviation of the running results of multiple times as the uncertainty value of the soil organic carbon content; Inputting the measured values of the soil organic carbon content and environmental variable data of each pixel in the validation set into the optimal random forest prediction model to obtain the predicted values of the soil organic carbon content of each pixel in the validation set, and based on the measured values and predicted values of each pixel in the validation set, using the coefficient of determination and root mean square error as two indicators to evaluate the accuracy of the mapping result of the mean value map.

4. The method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon according to claim 1, wherein Using the self-organizing mapping network model, with the soil organic carbon content and environmental variable data of each pixel as the input vector, all pixels in the region are divided into several different soil organic carbon state types, specifically including: Set the scale of the self-organizing mapping network, input the cover type, cover function index, vegetation production rate, terrain humidity index, and soil organic carbon content of all pixels in the region into the self-organizing mapping network model to obtain several different soil organic carbon state types; In setting the scale of the self-organizing mapping network, first set the initial scale value of the self-organizing mapping network according to the empirical formula, and the initial scale value is five times the square root of the total number of regional pixels; then select several values close to the initial scale value to obtain a set of values to be determined; calculate the quantity error and structural error respectively for any value in the set of values to be determined, and determine the scale quantity of the self-organizing mapping model according to the calculation results.

5. The method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon according to claim 1, characterized in that, In the constructed state space, use quadratic functions to fit the degradation path curve and the recovery path curve respectively to identify the non-linear evolution path of soil organic carbon, specifically including: Take the average cover function index and average soil organic carbon content of all pixels in each soil organic carbon state type as the fitting sample points, and use quadratic functions to fit the degradation path curve and the recovery path curve respectively; According to the S-shaped concept model, connect the degradation path curve and the recovery path curve with a dotted line to form a complete S-shaped evolution path.

6. The method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon according to claim 1, wherein Based on the non-linear evolution path of soil organic carbon, determine the mutation threshold, hysteresis, and resilience of the soil organic carbon evolution path, specifically including: In the non-linear evolution path of soil organic carbon, define the part with the same cover function index in the degradation path curve and the recovery path curve as the unstable state, the part with the cover function index less than the unstable state in the degradation path curve as the good stable state, and the part with the cover function index greater than the unstable state in the recovery path curve as the poor stable state, and determine the good mutation threshold and the poor mutation threshold respectively according to the cover function index division values between the two stable states and the unstable state; Determine the hysteresis of the soil organic carbon evolution path according to the range between the two mutation thresholds; Evaluate the resilience of the soil organic carbon evolution path according to the degree of change in the cover function index required for any soil organic carbon state type in the unstable state to be converted into the soil organic carbon state type in another achievable stable state.

7. The method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon according to claim 6, wherein The calculation method of the good mutation threshold includes: ； where S1 is the set of pixels corresponding to the soil organic carbon state type with the largest average cover function index in the good stable state, LCI(x) is the cover function index corresponding to pixel x, and LCI1 is the largest LCI(x); ； where S2 is the set of pixels corresponding to the soil organic carbon state type with the smallest average cover function index in the unstable state on the recovery path curve, LCI(x) is the cover function index corresponding to pixel x, and LCI2 is the smallest LCI(x); ； Among them, LCIa is a good mutation threshold, μ1 and μ2 are respectively the average values of the coverage function indices of the pixels corresponding to S1 and S2, σ1 and σ2 are respectively the standard deviations of the coverage function indices of the pixels corresponding to S1 and S2, n ∈ (0.1, 0.2, 0.3, …, 2), and the value of n should be determined to meet the following conditions: 。 8. The method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon according to claim 6, wherein The calculation method of the poor mutation threshold includes: ； Among them, S3 is the set of pixels corresponding to the soil organic carbon state type with the largest average coverage function index in the unstable state on the degradation path, LCI(x) is the coverage function index corresponding to pixel x, and LCI3 is the pixel with the largest coverage function index in the unstable state, that is, the largest LCI(x); ； Among them, S4 is the set of pixels corresponding to the soil organic carbon state type with the smallest average coverage function index in the poor stable state, LCI(x) is the coverage function index corresponding to pixel x, and LCI4 is the pixel with the smallest coverage function index in the poor stable state, that is, the smallest LCI(x); ； Among them, LCIb is the poor mutation threshold, μ3 and μ4 are respectively the average values of the coverage function indices of the pixels corresponding to state types S3 and S4, σ3 and σ4 are respectively the standard deviations of the coverage function indices of the pixels corresponding to state types S3 and S4, n ∈ (0.1, 0.2, 0.3, …, 2), and the value of n should be determined to meet the following conditions: 。 9. The method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon according to claim 6, wherein Evaluating the resilience of the soil organic carbon evolution path according to the degree of change in the coverage function index required for any soil organic carbon state type in the unstable state to be converted into another achievable stable state type of the soil organic carbon state type specifically includes: Determining all soil organic carbon state types in the good or poor stable states that any soil organic carbon state type in the unstable state can be restored or degraded to as candidate states; Among the candidate states, screening according to the principle of consistent coverage type, and screening out the set of soil organic carbon state types with the same coverage type as the current soil organic carbon state type; Determining the soil organic carbon state type in the achievable stable state corresponding to the current soil organic carbon state type in the set of soil organic carbon state types; Calculating the absolute value of the difference between the coverage function indices of the current soil organic carbon state type and the achievable soil organic carbon state type, which is the resilience of the current soil organic carbon state type to be restored or degraded to a good or poor stable state.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method for evaluating the non-linear evolution path and characteristics of regional soil organic carbon according to any one of claims 1 to 9.

Images