Threshold measurement method for carbon sink of sandy land based on monitoring of functional traits of grass communities

Abstract

This invention relates to the field of carbon sink threshold measurement, specifically to a method for measuring carbon sink thresholds in sandy areas based on monitoring the functional traits of shrub and grass communities. The method includes: selecting gradient sequences of three degrees of fixation—mobile sandy areas, semi-fixed sandy areas, and fixed sandy areas—and setting up several fixed monitoring plots according to different vegetation types. By establishing these plots and collecting data, the complex environmental factors of sandy areas can be comprehensively considered. Through data standardization, the use of a random forest regression model, the constructed "functional trait-carbon sink" coupling model, and piecewise regression analysis applied to the response curves of plant functional traits and carbon sink indicators to the degree of sand fixation, the threshold of carbon sink capacity can be dynamically measured. This accurately reflects the impact of plant growth on carbon sinks, thereby effectively improving the accuracy of carbon sink threshold measurement.

Description

Technical Field

[0001] This invention belongs to the field of carbon sink threshold measurement, specifically relating to a method for measuring carbon sink thresholds in sandy areas based on monitoring the functional traits of shrub and grass communities. Background Technology

[0003] In terms of measuring carbon sequestration thresholds in sandy areas, many traditional methods rely heavily on empirical thresholds for estimation. These empirical thresholds are often derived from limited historical data or observations under specific conditions, and are often inaccurate for different environments. Therefore, overcoming these technical problems and shortcomings has become a key issue that urgently needs to be addressed in this field. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of many traditional methods in the background art that rely on empirical thresholds for estimation. Empirical thresholds are often derived from limited historical data or observations under specific conditions, and the thresholds are often inaccurate for different environments. This invention aims to realize a method for measuring the carbon sequestration threshold in sandy areas based on the monitoring of the functional traits of shrub and grass communities.

[0005] To achieve the above-mentioned objectives, the technical solution of this invention is: a method for measuring carbon sequestration thresholds in sandy areas based on monitoring the functional traits of shrub and grass communities, comprising: Select gradient sequences of three degrees of fixation: mobile sand dunes, semi-fixed sand dunes, and fixed sand dunes, and set up several fixed monitoring plots according to different vegetation types; Plant functional traits, environmental factors, and carbon sink indicators were collected in each plot, and plant functional traits were measured. The collected plant functional trait variables and environmental factor variables were standardized. The standardized variables were then screened. Using the screened plant functional trait variables and environmental factor variables as features, and carbon sink indicators as response variables, a random forest regression model was adopted. The contribution of each variable was calculated using permutation importance, and the calculated contributions were normalized. Then, an initial path model was constructed using plant functional trait variables as exogenous manifest variables, carbon sink indicators as endogenous variables, and environmental factor variables as covariates. The model was then corrected through a goodness-of-fit test to obtain the optimal model. Plotting the sand fixation degree on the x-axis and the carbon sink index and plant functional traits in the model on the y-axis, piecewise regression analysis was used to find the threshold of the relationship curve.

[0006] In the above-mentioned method for measuring the carbon sink threshold in sandy areas based on the monitoring of functional traits of shrub and grass communities, the vegetation type is shrub layer.

[0007] In the above-mentioned method for measuring the carbon sink threshold of sandy land based on the monitoring of the functional traits of shrub and grass communities, three plot groups were set up in sequence in the selected fixed monitoring plots. Each plot group contained three replicate 20m×20m quadrats, for a total of 18 fixed monitoring plots.

[0008] In the above-mentioned method for measuring the carbon sink threshold in sandy areas based on the monitoring of functional traits of shrub and grass communities, the collection of plant functional trait variables, environmental factor variables, and carbon sink indicators includes measuring the diameter at ground level, plant height, and crown width of the dominant shrub layer in each plot; collecting mixed samples of healthy mature leaves and fine roots <2mm in the four directions of east, west, south, and north; setting up three 1m×1m quadrats along the diagonal of each plot; recording the name, average height, number of plants, and coverage of all herbaceous species; harvesting all aboveground parts at ground level; digging root clods to a depth of 30cm; washing to obtain the root system; and using a soil auger to perform stratified soil sampling. The litter layer and dead wood layer are collected and biomass carbon is calculated. The carbon sink indicator is calculated using the unit area.

[0009] In the above-mentioned method for measuring the carbon sink threshold in sandy areas based on the monitoring of functional traits of shrub and grass communities, the measurement of collected plant functional trait variables includes the measurement of the stoichiometric characteristics of carbon, nitrogen, and phosphorus in leaves and roots, as well as their morphological characteristics, to ensure data standardization and comparability.

[0010] In the above-mentioned method for measuring the carbon sink threshold in sandy areas based on the monitoring of shrub and grass community functional traits, the standardization of the collected plant functional trait variables and environmental factor variables includes: establishing an n×m table; Where n: number of plots; m: number of variables, including plant functional traits, environmental factors, and carbon sink response variables; Outliers in the established tables were checked and processed. For a small number of missing data, multiple imputation or imputation based on plot / species averages was used to fill in the missing data. For variables that deviated significantly from the normal distribution, logarithmic, square root, or Box-Cox transformations were performed to meet the data distribution requirements of the subsequent linear model. To eliminate the influence of different variable dimensions, all independent variables were standardized using Z-score. Spearman correlation analysis was used to initially select plant functional trait variables and environmental factor variables that were significantly related to carbon sink indicators. LASSO regression was used to further compress variables and screen out the plant functional trait variables and environmental factor variables that were most important and made high contributions to carbon sink prediction. The screening of standardized variables includes: calculating the Spearman rank correlation coefficient (ρ) and its significance p-value between each plant functional trait variable and environmental factor variable and each carbon sink index; retaining variables that satisfy |ρ|>0.3 and p<0.05 with any carbon sink index to form an initial set of variables; performing LASSO regression with the initial set of variables as features X and the carbon sink index as the response Y; LASSO selects variables by adding an L1 regularization term λ∑|βᵢ| to the loss function, compressing the regression coefficient β of unimportant variables to exactly 0 during the fitting process; determining the optimal regularization parameter λ through 10-fold cross-validation; and selecting variables with non-zero regression coefficients under the optimal λ, outputting a list containing k variables for these variables.

[0011] In the aforementioned method for measuring the carbon sink threshold in sandy areas based on the monitoring of functional traits of shrub and grass communities, the calculation of permutation importance includes: First, calculating a baseline prediction accuracy MSE_0 on the out-of-bag data of the model; randomly shuffling the value of a certain feature; then recalculating the prediction accuracy MSE_i of the model using the shuffled data; the importance Imp_i of this feature represents the degree of decrease in prediction accuracy: Imp_i = MSE_i - MSE_0; then repeating this process for all core variables; summing the permutation importance values ​​of each feature to obtain a total of Total_Imp; then calculating the relative contribution of each feature: Contribution = (Imp_i / Total_Imp); creating a table listing each core variable and its corresponding percentage contribution, sorted from highest to lowest contribution.

[0012] In the above-mentioned method for measuring the carbon sink threshold in sandy land based on the monitoring of functional traits of shrub and grass communities, the initial path model includes: firstly, assuming the path as annual average precipitation (MAP) → herbaceous SLA → community photosynthetic capacity → aboveground carbon storage (AGC), and then converting the assumed path into a series of linear equations in the Lavaan in R software. Community photosynthetic capacity = λ1 * herbaceous SLA + λ2 * herbaceous Nleaf + error; Aboveground carbon storage (AGC) = β1 * community photosynthetic capacity + β2 * available soil nitrogen + error; Estimate path coefficients, variance, and covariance using the maximum likelihood method; calculate a series of variables to determine the model's fit with the data; If the fit is poor, based on the correction index and theoretical rationality, paths are gradually added or removed to form a competing model. The AIC / BIC information criterion is used to compare the merits of different correction models. The model with the smallest AIC / BIC value is the optimal one. All significant paths and their standardized path coefficients β values ​​are displayed. β values ​​can directly compare the effect size. The direct, indirect, and total effects of each variable on the carbon sink index are calculated. This is a key result for revealing the regulatory path and mechanism of action. By using the partial dependency graph output by the random forest model, we can visualize the relationship between plant functional traits and carbon sink indicators, as well as their interaction with environmental factor variables.

[0013] Compared with the prior art, the method for measuring carbon sink thresholds in sandy areas based on the monitoring of functional traits of shrub and grass communities in this invention has at least the following beneficial effects: 0. The present invention provides a method for measuring carbon sink thresholds in sandy areas based on monitoring the functional traits of shrub and grass communities. Through the establishment of sample plots and data collection, it can comprehensively consider the complex environmental factors of sandy areas. By standardizing data, using a random forest regression model, constructing a "functional trait-carbon sink" coupling model, and applying piecewise regression analysis to the response curves of plant functional traits and carbon sink indicators to the degree of sandy area fixation, it can dynamically measure the threshold of carbon sink capacity. This can accurately reflect the impact of plant growth on carbon sinks, thereby effectively improving the accuracy of carbon sink threshold measurement. It can also be used for current status assessment, predicting future carbon sink dynamics under different climate change scenarios or vegetation management schemes, and providing decision support for ecological restoration planning.

[0014] 0. The method for measuring the carbon sequestration threshold in sandy land based on the monitoring of functional traits of shrub and grass communities in this invention, by introducing multi-dimensional plant functional trait indicators, can not only more accurately assess the carbon sequestration capacity of sandy land, but also deeply reveal the underlying plant physiological and ecological mechanisms, effectively answering the question of the causes of differences in carbon sequestration capacity in different regions or communities. Attached Figure Description

[0015] Figure 1 This is a conceptual model diagram of the functional trait-carbon sink coupled SEM of the sandy land carbon sink threshold measurement method based on the monitoring of functional traits of shrub and grass community in this invention. Figure 2 This is a technical roadmap for the sandy land carbon sink threshold measurement method based on the monitoring of shrub and grass community functional traits, as proposed in this invention. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings and specific embodiments, provides a more detailed account of the method for measuring carbon sink thresholds in sandy areas based on the monitoring of functional traits of shrub and grass communities.

[0017] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0018] This embodiment discloses a method for measuring carbon sink thresholds in sandy areas based on monitoring the functional traits of shrub and grass communities, which mainly includes: Plot layout: Select a gradient sequence of three degrees of fixation in the eastern section of the Kubuqi Desert: mobile sand dunes, semi-fixed sand dunes, and fixed sand dunes, and set up several fixed monitoring plots according to different vegetation types. Data collection and measurement: Plant functional traits, environmental factors, and carbon sink indicators were collected in each plot, and plant functional traits were measured. Through plot layout and data collection, the complex environmental factors of sandy land can be comprehensively considered.

[0019] Model Construction: The collected plant functional trait variables and environmental factor variables were standardized, and the standardized variables were screened. Using the screened variables as features and carbon sink indicators as response variables, a random forest regression model was adopted. The contribution of each variable was calculated using permutation importance, and the calculated contribution was normalized. Then, an initial path model was constructed with plant functional traits as exogenous manifest variables, carbon sink indicators as endogenous variables, and environmental factor variables as covariates. The model was then corrected through goodness-of-fit tests to obtain the optimal model. Threshold measurement: Plotting sand fixation degree as the x-axis and carbon sink index and plant functional trait variables in the model as the y-axis, piecewise regression analysis is used to find the threshold of the relationship curve; the differences in plant functional trait combination, community structure and carbon sink pathway before and after the threshold are analyzed to clarify the trait configuration and vegetation conditions required to achieve optimal carbon sink.

[0020] The vegetation type is a shrub layer; the shrub layer consists of Caragana korshinskii and Salix psammophila.

[0021] Three plot groups were set up sequentially in the selected fixed monitoring plots of Caragana korshinskii and Salix psammophila. Each plot group contained three replicate 20m×20m quadrats, for a total of 18 fixed monitoring plots.

[0022] The collection of plant functional trait variables, environmental factor variables, and carbon sink indicators included measuring the diameter at breast height, plant height, and crown width of dominant *Caragana korshinskii* and *Salix psammophila* species in each plot; collecting mixed samples of healthy mature leaves and fine roots <2mm in diameter in four directions (north, south, east, and west); and setting up three 1m × 1m quadrats along the diagonal of each plot. The names, average height, number of plants, and canopy coverage of all herbaceous species were recorded. All aboveground parts were harvested at ground level, and root clods were dug to a depth of 30cm. Roots were obtained by rinsing. Soil samples were taken using a soil auger in layers of 0-10cm, 10-20cm, 20-30cm, 30-60cm, 60-80cm, and 80-100cm to determine soil organic carbon, bulk density, and nutrients. Carbon sink data collection included collecting samples from litter and deadwood layers, calculating biomass carbon, and calculating carbon sink indicators per unit area.

[0023] The stoichiometric and morphological characteristics of carbon, nitrogen, and phosphorus in leaves and roots were strictly determined according to the methods shown in Table 1 to ensure data standardization and comparability.

[0024] Table 1. Methods for determining plant functional traits and their ecological significance. Standardization of the collected plant functional trait variables and environmental factor variables includes: establishing an n×m table; Where n: number of plots (e.g., 18 plots); m: number of variables, including plant functional trait variables, environmental factor variables (average annual temperature MAT, average annual precipitation MAP, available soil nitrogen Soil_AN, soil pH, soil bulk density BD, etc.) and carbon sink response variables (total carbon storage per unit area Total_C, aboveground carbon storage AGC, soil organic carbon storage SOC). Outliers in the established tables were checked and processed. For a small number of missing data, multiple imputation or imputation based on plot / species averages was used to fill in the missing data. For variables that seriously deviated from the normal distribution (judged by Shapiro-Wilk test or QQ plot), logarithmic (log(x+1)), square root, or Box-Cox transformations were performed to meet the data distribution requirements of the subsequent linear model. To eliminate the influence of different variable dimensions, all variables (plant functional traits and environmental factors) were standardized using Z-score. After processing, the mean of each variable was 0 and the standard deviation was 1, indicating that they were all on the same order of magnitude. Spearman correlation analysis was used to initially select plant functional trait variables and environmental factor variables that were significantly correlated with carbon sink indicators (such as total carbon storage per unit area, aboveground carbon storage, and soil carbon storage). LASSO regression was used to further compress variables and screen out the plant functional trait variables and environmental factor variables (such as available nitrogen in soil and annual precipitation) that were most important for carbon sink prediction and had a high contribution. The screening of standardized variables includes: calculating the Spearman rank correlation coefficient (ρ) and its significance p-value between each plant functional trait variable and environmental factor variable and each carbon sink index (Total_C); Spearman correlation is suitable for nonlinear monotonic relationships and is more robust than Pearson correlation; variables that satisfy |ρ|>0.3 and p<0.05 with any carbon sink index are retained. This step initially eliminates variables with weak or no statistical significance association with carbon sink indexes, forming a preliminary set of variables; using the preliminary set of variables as feature X and the carbon sink index (such as Total_C) as response Y, LASSO regression is performed; LASSO adds an L1 regularization term λ∑|βᵢ| to the loss function during the fitting process. The regression coefficients β of unimportant variables are compressed to exactly 0, thus achieving variable selection. The optimal regularization parameter λ is determined through 10-fold cross-validation, using minimizing the average prediction error as the criterion. Specifically, the data is randomly divided into 10 parts, with 9 parts used for training the model and 1 part for validation. The average prediction error of the model under different λ values ​​is calculated. Typically, the λ value that minimizes the cross-validation error (λ_min) or the λ value of the simplest model with an error within one standard deviation (λ_1se) is chosen. λ_1se produces a sparser model (fewer variables) and is more interpretable. Based on this, the LASSO model is refitted using all data at this optimal λ value, and variables with non-zero regression coefficients are selected as the final subset of key variables. A list containing k variables (k much smaller than m) is output; this subset will be used for subsequent contribution quantification and SEM construction.

[0025] Random forest regression models, by constructing a large number of decision trees and integrating their results, can effectively handle nonlinear relationships and interactions, and are not sensitive to overfitting. The number of trees (e.g., 500 trees), the maximum number of features considered when splitting each tree (usually √k), etc., are used to select a subset of variables. A random forest regression model is used, with plant functional traits as features and carbon sink indicators as response variables. Permutation importance calculation is employed. First, the baseline prediction accuracy MSE_0 is calculated on the out-of-bag data. The values ​​of a certain feature (e.g., shrub Nleaf) are randomly shuffled (disrupting the relationship between this feature and the carbon sink response variable). Then, the prediction accuracy MSE_i is recalculated using the shuffled data. The importance Imp_i of this feature represents the degree of decrease in prediction accuracy: Imp_i = MSE_i - MSE_0. The greater the decrease, the more important the feature. This process is repeated for all variables. The permutation importance values ​​of each feature are summed to obtain a total_Imp. Then, the relative contribution of each feature is calculated: Contribution = (Imp_i / ... Total_Imp); Create a table listing each variable and its corresponding percentage contribution, sorted from highest to lowest contribution.

[0026] The initial path model includes: constructing an initial path hypothesis map based on prior ecological knowledge (such as "leaf traits → photosynthetic capacity → aboveground biomass", "fine root traits → nutrient absorption → productivity", "litter traits → decomposition rate → soil carbon"); forming an initial path map based on contribution ranking and ecological theory, initially assuming the initial path as: average annual precipitation (MAP) → herbaceous soil saturation (SLA) → community photosynthetic capacity latent variable → aboveground carbon storage (AGC); and transforming the assumed initial path into a series of linear equations in lavaan in R software. Measurement model (for latent variables): Community photosynthetic capacity = λ1 * herbaceous SLA + λ2 * herbaceous Nleaf + error; Structural model (relationships between variables): Aboveground carbon storage AGC = β1 * community photosynthetic capacity + β2 * soil available nitrogen + error; The path coefficients (λ, β), variance, and covariance are estimated using the maximum likelihood method. A series of variables are calculated to judge the model's fit to the data, where χ² / df (the ratio of chi-square value (χ²) to degrees of freedom (df), used to test the absolute fit of the model to the data) <3 (the smaller the better), RMSEA (measures the model's approximation error in the population, considering model complexity) <0.08 (the smaller the better), CFI (compares the goodness of fit between the target model and the independent model (the model without any paths)) >0.90 (the larger the better), and SRMR (measures the average of the standardized residuals of the model's predicted covariance and the sample covariance) <0.08 (the smaller the better). If the fit is poor, based on the correction index and theoretical rationality, pathways are gradually added or removed (e.g., adding the pathway "soil available nitrogen → shrub Nleaf") to form a competing model. The AIC / BIC information criterion is used to compare the merits of different modified models, and the model with the lowest AIC / BIC value is the optimal one. All significant pathways and their standardized pathway coefficients β values ​​are displayed; β values ​​can directly compare effect sizes (e.g., a pathway with β=0.5 has a stronger impact than a pathway with β=0.2), including indicators such as χ², df, RMSEA, CFI, and SRMR, to prove that the model is supported by the data; the direct effect, indirect effect (through mediating variables), and total effect of each variable on the carbon sink index are calculated. This is a key result in revealing the regulatory pathways and mechanisms of action.

[0027] By using the standardized path coefficients output by SEM, the direct and indirect effects of each plant functional trait on carbon sink indicators are quantified. Using the partial dependency graph output by the random forest model, the relationship between plant functional trait variables (such as SLA) and carbon sink indicators (linear and nonlinear) is visualized, as well as their interaction with environmental factor variables (such as temperature).

[0028] Piecewise regression analysis showed that when the sandy soil fixation level (expressed as vegetation cover) reached 45%-55%, both shrub Nroot and community carbon sequestration efficiency showed a significant increase; this range is considered the "critical threshold." Below this threshold, carbon sequestration is severely limited by water, and trait regulation is weak; above this threshold, the system enters a rapid carbon fixation stage, and shrub root traits become the key regulatory factors.

[0029] Management recommendations: In areas where sand fixation is below 45%, sand barriers should be used to stabilize sand and protect pioneer herbaceous plants. When the fixation rate enters the threshold range of 45%-55%, priority should be given to introducing or promoting shrub species with high fine root nitrogen content (such as Caragana korshinskii) and combining them with herbaceous plants with high phosphorus utilization efficiency to maximize carbon sequestration potential.

[0030] It should be noted that the carbon sink index can be specified as the amount of carbon fixed by the ecosystem per unit time and per unit area, or more practically characterized by the storage of different carbon pool components (aboveground biomass carbon, belowground biomass carbon, and soil organic carbon). The model is constructed as a multivariate statistical method that integrates factor analysis and path analysis. It can handle multiple dependent variables simultaneously and allows measurement errors to be included in both independent and dependent variables, making it very suitable for testing complex path hypotheses based on ecological theory. LASSO regression is a compressed estimation method that performs variable selection and complexity adjustment while fitting a generalized linear model, and is suitable for high-dimensional data. Out-of-bag data refers to sample data that is not used for training the current tree when building an ensemble model (such as random forest, gradient boosting tree, etc.).

[0031] Permutation importance calculation is a method used to quantify the contribution of different environmental variables or ecological factors to carbon sink function in a community. Its core principle is to randomly permutate (shuffle) the value of a variable, observe the changes in model performance (such as the accuracy of carbon sink prediction), and then assess the relative importance of the variable to carbon sink function.

[0032] AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion) are two important information criteria used in statistics for model selection. They evaluate the quality of different models by balancing the goodness of fit and complexity. The model with the smallest AIC / BIC value is usually considered the optimal model.

[0033] This invention is not a simple correlation analysis; it is a complete technical system that moves from hypothesis-driven (ecological theory) analysis to data collection, and then through multi-level statistical modeling to verify hypotheses, quantify contributions, and identify thresholds. Its core lies in establishing a causal inference framework (through SEM) of "trait combination → ecological process → carbon sink function," rather than merely describing statistical relationships.

[0034] Unless otherwise defined, the technical or scientific terms used herein should be understood in their ordinary sense as would be understood by one of ordinary skill in the art to which this invention pertains. The use of terms such as "a" or "an" in this specification and claims does not necessarily indicate a limitation of quantity. Terms such as "comprising" or "including" mean that the element or component preceding the word encompasses the element or component listed following the word and its equivalents, without excluding other elements or components. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0035] The exemplary embodiments of the present invention have been described in detail above with reference to preferred embodiments. However, those skilled in the art will understand that various modifications and alterations can be made to the above specific embodiments without departing from the concept of the present invention, and various combinations can be made to the various technical features and structures proposed in the present invention without exceeding the protection scope of the present invention.

Claims

1. A method for measuring carbon sequestration thresholds in sandy areas based on monitoring the functional traits of shrub and grass communities, characterized by: include: Select gradient sequences of three degrees of fixation: mobile sand dunes, semi-fixed sand dunes, and fixed sand dunes, and set up several fixed monitoring plots according to different vegetation types; Plant functional traits, environmental factors, and carbon sink indicators were collected in each plot, and plant functional traits were measured. The collected plant functional trait variables and environmental factor variables were standardized. The standardized variables were then screened. Using the screened plant functional trait variables and environmental factor variables as features, and carbon sink indicators as response variables, a random forest regression model was adopted. The contribution of each variable was calculated using permutation importance, and the calculated contributions were normalized. Then, an initial path model was constructed using plant functional trait variables as exogenous manifest variables, carbon sink indicators as endogenous variables, and environmental factor variables as covariates. The model was then corrected through a goodness-of-fit test to obtain the optimal model. Plotting the sand fixation degree on the x-axis and the carbon sink index and plant functional traits in the model on the y-axis, piecewise regression analysis was used to find the threshold of the relationship curve.

2. The method for measuring the carbon sequestration threshold in sandy land based on the monitoring of functional traits of shrub and grass communities according to claim 1, characterized in that: The vegetation type is shrub layer.

3. The method for measuring the carbon sequestration threshold in sandy land based on the monitoring of functional traits of shrub and grass communities according to claim 2, characterized in that: Three plot groups were set up sequentially in the selected fixed monitoring plots. Each plot group contained three replicate 20m×20m quadrats, for a total of 18 fixed monitoring plots.

4. The method for measuring the carbon sequestration threshold in sandy land based on the monitoring of functional traits of shrub and grass communities according to claim 1, characterized in that: The collection of plant functional trait variables, environmental factor variables, and carbon sink indicators included measuring the diameter at ground level, plant height, and crown width of the dominant shrub layer in each plot; collecting mixed samples of healthy mature leaves and fine roots <2mm in diameter in the four directions of east, west, south, and north; setting up three 1m×1m quadrats along the diagonal of each plot; recording the name, average height, number of plants, and canopy of all herbaceous species; harvesting all aboveground parts at ground level; digging root clods to a depth of 30cm; washing to obtain the root system; and using a soil auger to perform stratified soil sampling. Biomass carbon was collected from the litter layer and deadwood layer, and the carbon sink indicator was calculated using the unit area.

5. The method for measuring the carbon sequestration threshold in sandy areas based on the monitoring of functional traits of shrub and grass communities according to claim 4, characterized in that: The determination of collected plant functional trait variables includes measuring the stoichiometric characteristics of carbon, nitrogen, and phosphorus in leaves and roots, as well as their morphological characteristics, to ensure data standardization and comparability.

6. The method for measuring the carbon sequestration threshold in sandy land based on the monitoring of functional traits of shrub and grass communities according to claim 1, characterized in that: The standardization of the collected plant functional trait variables and environmental factor variables includes: establishing an n×m table; where n: number of plots; m: number of variables, including plant functional trait variables, environmental factor variables, and carbon sink response variables; Outliers in the established tables were checked and processed. For a small number of missing data, multiple imputation or imputation based on plot / species averages was used to fill in the missing data. For variables that deviated significantly from the normal distribution, logarithmic, square root, or Box-Cox transformations were performed to meet the data distribution requirements of the subsequent linear model. To eliminate the influence of different variable units, all variables were standardized using Z-score. Spearman correlation analysis was used to initially select plant functional trait variables and environmental factor variables that were significantly related to carbon sink indicators. LASSO regression was used to further compress variables and screen out the plant functional trait variables and environmental factor variables that were most important and made high contributions to carbon sink prediction. The screening of standardized variables includes: calculating the Spearman rank correlation coefficient (ρ) and its significance p-value between each plant functional trait variable and environmental factor variable and each carbon sink index; retaining variables that satisfy |ρ| > 0.3 and p < 0.05 with any carbon sink index to form an initial set of variables; performing LASSO regression with the initial set of variables as features X and the carbon sink index as the response Y; LASSO selects variables by adding an L1 regularization term λ∑|βᵢ| to the loss function, compressing the regression coefficient β of unimportant variables to exactly 0 during the fitting process; determining the optimal regularization parameter λ through 10-fold cross-validation; and selecting variables with non-zero regression coefficients under the optimal λ, outputting a list of k variables for these variables.

7. The method for measuring the carbon sequestration threshold in sandy land based on the monitoring of functional traits of shrub and grass communities according to claim 6, characterized in that: The permutation importance calculation includes: First, calculating a baseline prediction accuracy MSE_0 on the out-of-bag data of the model; randomly shuffling the values ​​of a certain feature; then recalculating the prediction accuracy MSE_i of the model using the shuffled data; the importance Imp_i of this feature represents the degree of decrease in prediction accuracy: Imp_i = MSE_i - MSE_0; then repeating this process for all core variables; summing the permutation importance values ​​of each feature to obtain a total_Imp; then calculating the relative contribution of each feature: Contribution = (Imp_i / Total_Imp); creating a table listing each variable and its corresponding percentage contribution, sorted from highest to lowest contribution.

8. The method for measuring the carbon sequestration threshold in sandy land based on the monitoring of functional traits of shrub and grass communities according to claim 7, characterized in that: The initial path model includes: First, assuming the initial path is average annual precipitation (MAP) → herbaceous SLA → community photosynthetic capacity → aboveground carbon storage (AGC). In the Lavaan in R software, the assumed initial path is transformed into a series of linear equations. Community photosynthetic capacity = λ1 * herbaceous SLA + λ2 * herbaceous Nleaf + error; Aboveground carbon storage (AGC) = β1 * community photosynthetic capacity + β2 * available soil nitrogen + error; Estimate path coefficients, variance, and covariance using the maximum likelihood method; calculate a series of variables to determine the model's fit with the data; If the fit is poor, based on the correction index and theoretical rationality, paths are gradually added or removed to form a competing model. The AIC / BIC information criterion is used to compare the merits of different correction models. The model with the smallest AIC / BIC value is the optimal one. All significant paths and their standardized path coefficients β values ​​are displayed. β values ​​can directly compare the effect size. The direct, indirect, and total effects of each variable on the carbon sink index are calculated. This is a key result for revealing the regulatory path and mechanism of action. By using the partial dependency graph output by the random forest model, we can visualize the relationship between plant functional traits and carbon sink indicators, as well as their interaction with environmental factor variables.

Images