A method for dynamically evaluating soil carbon sequestration rate of pinus massoniana plantation

Abstract

The application discloses a kind of dynamic evaluation method of soil carbon sequestration rate of Pinus massoniana artificial forest, and relates to carbon sequestration evaluation technical field, the method comprises the following steps: step S1, Pinus massoniana artificial forest development sequence sample plot is selected in target area, and soil sample of different depths is collected and pretreated;Step S2, the soil sample collected is grouped and the content of each grouping component and soil total organic carbon is determined;Step S3, the density of each organic carbon component in soil profile is calculated respectively;Step S4, the apparent carbon sequestration rate of each organic carbon component between adjacent development stages is calculated;Step S5, machine learning model is constructed to predict carbon sequestration rate, and model is trained and verified;Step S6, the trained model is used to predict future carbon sequestration rate, problem diagnosis and risk level assessment are carried out, and decision suggestion is generated based on multi-objective optimization problem.The present application realizes the whole process support from evaluation to prediction, then to decision for forestry carbon neutralization.

Description

Technical Field

[0001] This invention relates to the field of carbon sequestration assessment technology, specifically to a method for dynamic assessment of soil carbon sequestration rate in Masson pine plantations. Background Technology

[0002] Soil is the largest carbon sink on land. The carbon storage at a depth of 2 meters globally far exceeds the combined carbon storage of the atmosphere and vegetation. 80% of the total carbon involved in the Earth's terrestrial carbon cycle exists in the form of soil organic carbon. Even small changes in soil organic carbon can cause significant fluctuations in atmospheric carbon dioxide, impacting global climate change. Assessments indicate that, under appropriate ecosystem management measures, soil organic carbon contributes up to 25% to natural climate solutions. Under the dual carbon objectives, the ability of soil to fix and accumulate organic carbon has received considerable attention.

[0003] Soil organic carbon (OCC) is a series of carbon components that differ in composition, structure, form, degree of decomposition, and turnover time. Various organic carbon components with different properties can be separated according to different grouping methods. Based on density grouping, soil organic matter can be divided into light organic carbon and recombinant organic carbon. Light organic carbon is easily decomposed and quickly turnovers, making it more sensitive to changes in soil organic matter, but it cannot be truly fixed by the soil and is considered unstable organic carbon. Recombinant organic carbon, on the other hand, is protected by soil organic carbon stabilization mechanisms and does not fluctuate in the short term, reflecting the soil's long-term accumulation and carbon sequestration capacity; it is considered stable organic carbon. Exploring changes in organic carbon components based on density grouping is of great significance for assessing soil carbon pool stability and understanding changes in soil quality. Since soil is the largest carbon pool in terrestrial ecosystems, the accuracy of its carbon sequestration potential assessment is crucial.

[0004] For example, Chinese patent CN118937573A discloses a method for assessing the aboveground carbon pool of Masson pine forests. This method involves first establishing a fixed monitoring base, using drones to photograph the entire Masson pine forest to obtain high-resolution images, analyzing the average age of the forest to determine its age group, establishing growth curves based on age groups, assigning growth weights to each age group, and combining tree age with monitoring indicators to conduct real-time monitoring and assessment of carbon sequestration, thereby obtaining the carbon storage of the Masson pine forest. This invention modifies traditional methods for estimating the carbon storage of Masson pine forests by analyzing the average age of the forest using drone photography to determine its age group, and then quickly estimating the carbon storage of the forest by linking it to the tree age using an intelligent carbon sequestration monitoring system.

[0005] For example, Chinese patent CN106779160B discloses a method for regulating stand structure to improve the carbon sequestration capacity of Masson pine forests, including the following steps: (1) Selecting sample plots, determining the tree species names within the sample plots, measuring the diameter at breast height (DBH) and height of individual trees within the sample plots, and recording the location coordinates of individual trees using the adjacent grid method; (2) Calculating the stand structure indicators of the sample plots based on the tree species names, DBH, height, and location coordinates: crown overlap index, health index, mixing degree, size ratio, angular scale, and diameter order distribution q value; and calculating the harvesting determination index; (3) Determining the stand structure regulation method based on the stand structure indicators and sample plot survey factors to regulate the stand structure. This invention method can not only promote the construction of a highly efficient forest production system, but also promote the construction of a benign cycle ecosystem. The annual carbon sequestration per unit area of ​​the regulated stand is higher than that of the control stand, which has significant economic and ecological benefits and a very broad application prospect.

[0006] All of the above patents suffer from the problem described in the background section: existing technologies are mostly limited to measuring the static stockpile of total organic carbon (TOC) in soil, failing to reveal its inherent stability and dynamic change patterns. Different organic carbon components with varying stability (such as physically protected aggregated carbon, chemically stable mineral-bound carbon, and biochemically inert carbon) respond differently to forest management. Therefore, developing a method that can distinguish different carbon components, quantify their dynamic carbon sequestration rates, and predict future changes to guide management is of great significance for scientifically enhancing the carbon sequestration function of forests. Summary of the Invention

[0007] This invention aims to overcome the shortcomings of existing technologies by using time series studies, physical-chemical joint grouping techniques, dynamic rate calculation models, and machine learning algorithms to achieve full-process support from current status assessment to future prediction and optimization decision-making. It provides a systematic and predictive method for assessing and optimizing the soil carbon sequestration capacity of Masson pine plantations.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] A method for dynamically assessing the soil carbon sequestration rate of Masson pine plantations includes the following steps:

[0010] Step S1: Select a developmental sequence plot of Masson pine plantation in the target area, which includes five developmental stages: young forest, middle-aged forest, near-mature forest, mature forest and over-mature forest. Collect soil samples at different depths and pre-process them.

[0011] Step S2: Divide the collected soil samples into groups and determine the content of each group component and the total organic carbon content of the soil.

[0012] Step S3: Calculate the density of each organic carbon component in the soil profile by combining soil bulk density and soil layer thickness;

[0013] Step S4: Based on the organic carbon density data of sample plots at different developmental stages, calculate the apparent carbon fixation rate of each organic carbon component between adjacent developmental stages.

[0014] Step S5: Construct a machine learning model, taking forest stand characteristic factors, soil environmental factors and current organic carbon density as inputs, and the carbon sequestration rate of each organic carbon component in the future stage as output, and train and validate the model.

[0015] Step S6: Use the trained model to predict the future carbon sequestration rate, perform problem diagnosis and risk level assessment, and generate decision suggestions based on multi-objective optimization problems.

[0016] Further, step S2 specifically includes: physically grouping the soil sample to obtain large aggregates, micro aggregates and silty aggregates, grouping by density to obtain light and heavy organic carbon components, and chemically grouping to obtain inert organic carbon components.

[0017] Furthermore, in step S2, the physical grouping adopts the wet sieving method, the density grouping reference adopts the Fourier transform method, and the chemical grouping adopts the acid hydrolysis method.

[0018] Furthermore, in step S3, the specific formula for calculating the density of each organic carbon component in the soil profile is as follows:

[0019]

[0020] in, Indicates organic carbon components. This represents the total density of organic carbon component x in the 0-80cm soil layer. Indicates the number of soil layers. Represents the i-th soil layer. This represents the organic carbon content of organic carbon component x in the i-th soil layer. This represents the soil bulk density of the i-th soil layer. Indicates the thickness of the i-th soil layer, 10 -1 This indicates the unit conversion factor.

[0021] Further, in step S4, the apparent carbon fixation rate is calculated as follows: the difference between the total density of the same organic carbon component in the later developmental stage and the total density in the previous developmental stage is calculated, and then the difference in total density is divided by the difference between the average age of the stand in the later developmental stage and the average age of the stand in the previous developmental stage. The result is the average apparent carbon fixation rate of the organic carbon component from developmental stage m to stage n.

[0022] Further, in step S5, the stand characteristic factors include: stand age, stand density, average diameter at breast height (DBH), average tree height, and litter biomass; the soil environmental factors include: soil clay content and soil pH; and the current organic carbon density status includes: macroaggregate organic carbon density, microaggregate organic carbon density, silty aggregate organic carbon density, light aggregate organic carbon density, recombinant organic carbon density, inert organic carbon density, and total organic carbon density.

[0023] Furthermore, in step S5, the machine learning model is a multilayer perceptron neural network model, wherein the number of neurons in the input layer of the multilayer perceptron is 14, which is the same as the dimension of the input features; there are 2 hidden layers, with 128 neurons in the first hidden layer and 64 neurons in the second hidden layer; and the number of neurons in the output layer is 7, which is used to predict the carbon fixation rate of the 6 organic carbon components and the total organic carbon fixation rate.

[0024] Furthermore, in the multilayer perceptron neural network model, the hidden layer activation function adopts the ReLU function; the output layer activation function adopts the linear activation function; the training adopts the Adam optimization algorithm; K-fold cross-validation is used to evaluate the model performance; a weighted mean square error is constructed as the loss function, wherein the weights of the loss function are determined according to the ecological stability of each organic carbon component; the weight coefficient of organic carbon components with high stability is greater than the weight coefficient of organic carbon components with low stability.

[0025] Furthermore, in step S6, the problem diagnosis and risk level assessment specifically include:

[0026] The predicted future carbon fixation rate of each component is compared with the historical average rate baseline of that component and the expected rate baseline of that developmental stage.

[0027] A comprehensive risk score is automatically calculated based on the degree of deviation, the absolute value, and the ecological importance weight of the component.

[0028] Based on the risk score, the problem is classified into different levels and precisely located to the specific soil layer.

[0029] Furthermore, in step S6, generating decision suggestions based on the multi-objective optimization problem specifically includes:

[0030] The problem of maximizing single carbon sequestration is transformed into a multi-objective optimization problem, which includes: ecological benefit optimization, i.e., maximizing the carbon sequestration rate of stable carbon components; economic benefit optimization, i.e., simulating the timber revenue brought about by different thinning intensities under thinning scenarios; and ecological safety optimization, i.e. ensuring that key indicators such as predicted soil erosion rate and water conservation capacity do not exceed safety thresholds after management measures such as thinning are implemented.

[0031] Transform actionable operational measures into continuous variables that the model can understand;

[0032] A multi-objective optimization algorithm is used to search for the Pareto optimal solution set among the combination of measures;

[0033] For representative schemes on the Pareto front, Monte Carlo simulations are performed using a multilayer perceptron neural network model to give the probability distribution of the expected results.

[0034] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0035] 1. This invention, by constructing a "spatial substitution time" sequence and a carbon sequestration rate calculation model, enables a quantitative assessment of the dynamic change rate of different stable carbon components during the development of Masson pine plantations. It can reveal the dynamic change law of soil carbon sequestration capacity and provide a scientific basis for medium- and long-term carbon sink potential assessment.

[0036] 2. This invention uses a combined physical-chemical grouping technique to decompose the total carbon pool of soil into components with different levels of stability. This allows for precise differentiation between unstable active carbon pools and stable inert carbon pools, thus more accurately reflecting the long-term carbon sequestration potential of the soil. It avoids the risk of misjudging easily decomposable active carbon as a persistent carbon sink, making the assessment results more scientific and reliable.

[0037] 3. This invention introduces machine learning models and multi-objective optimization algorithms, which can not only predict future carbon sequestration trends but also simulate the potential effects of different management measures (such as thinning, replanting, and fertilization). The system can output quantitative and actionable precise management plans and estimate their carbon sequestration and economic benefits, changing the previous extensive management model that relied on experience-based decision-making and realizing intelligent and precise forest management. Attached Figure Description

[0038] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0039] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0040] Figure 2 This is a schematic diagram of a multilayer perceptron neural network model according to an embodiment of the present invention;

[0041] Figure 3 This is a sample plot distribution map according to an embodiment of the present invention;

[0042] Figure 4 This is a graph showing the carbon sequestration rate of various soil organic carbon components in a Pinus massoniana plantation at different developmental stages according to an embodiment of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] like Figure 1 As shown, a method for dynamically assessing the soil carbon sequestration rate of Masson pine plantations includes the following steps:

[0045] Step S1: Select a developmental sequence plot of Masson pine plantation in the target area, which includes five developmental stages: young forest, middle-aged forest, near-mature forest, mature forest and over-mature forest. Collect soil samples at different depths and pre-process them.

[0046] Step S2: Divide the collected soil samples into groups and determine the content of each group component and the total organic carbon content of the soil.

[0047] Step S3: Calculate the density of each organic carbon component in the soil profile by combining soil bulk density and soil layer thickness;

[0048] Step S4: Based on the organic carbon density data of sample plots at different developmental stages, calculate the apparent carbon fixation rate of each organic carbon component between adjacent developmental stages.

[0049] Step S5: Construct a machine learning model, taking forest stand characteristic factors, soil environmental factors and current organic carbon density as inputs, and the carbon sequestration rate of each organic carbon component in the future stage as output, and train and validate the model.

[0050] Step S6: Use the trained model to predict the future carbon sequestration rate, perform problem diagnosis and risk level assessment, and generate decision suggestions based on multi-objective optimization problems.

[0051] like Figure 3 As shown, following the spatial substitution for time method, forest stands with relatively concentrated locations and basically consistent site conditions within the forest farm were selected to establish five developmental stages: young forest, middle-aged forest, near-mature forest, mature forest, and over-mature forest. Three plots were established for each stage, totaling 15 plots. Each plot was 20m × 20m in size and divided into 10 4m × 10m quadrats to investigate vegetation composition. In each 4m × 10m quadrat, all trees were surveyed, and the species, height, diameter at breast height (DBH), branch height, and crown width of each tree were recorded. In each 4m × 10m quadrat, one 2m × 2m quadrat was selected to survey shrubs, and the species, quantity, height, and ground diameter of shrubs were recorded. In each 2m × 2m quadrat, one 1m × 1m quadrat was selected to survey herbs, and the species, abundance, average height, and total canopy of herbs were recorded. For interlayer plants, they were assigned to the corresponding layer based on their height.

[0052] At the center of each plot, a complete soil profile was excavated. Sampling was conducted in 20cm layers, up to a depth of 80cm, for a total of four layers. Approximately 1.5kg of soil from each layer was collected and placed in a resealable bag for chemical property analysis. Simultaneously, a ring sampler was used to extract soil from each layer for bulk density analysis. After bringing the soil samples back to the laboratory, visible gravel and roots were removed. The samples were then allowed to air dry naturally indoors. The dried soil was sieved through a 2mm sieve, bagged, and labeled for testing. The ring sampler was dried in a 60℃ oven to constant weight, and the weight was recorded. The soil bulk density and moisture content were calculated.

[0053] Step S2 specifically includes: physically grouping soil samples to obtain large aggregates, micro aggregates and silty aggregates, grouping by density to obtain light and heavy organic carbon components, and chemically grouping to obtain inert organic carbon components.

[0054] In step S2, the physical grouping adopts the wet sieving method, the density grouping reference adopts the Fourier transform method, and the chemical grouping adopts the acid hydrolysis method.

[0055] The steps for separating soil aggregates are as follows: (1) Weigh 50g of air-dried soil that has passed through a 2mm sieve, accurate to 0.01g. Place the weighed soil onto a 0.25mm sieve and place the sieve in a clean container. (3) Add distilled water to the container until it covers the sieve by 2cm and let it stand for 5min. (4) Gently shake the sieve up and down about 50 times within 2min. Note that the water level should not be exceeded when shaking upwards and should not touch the bottom of the container when shaking downwards. (5) Remove the sieve and place the soil on the sieve into a pre-weighed sample tube to obtain large aggregates (0.25~2mm). (6) Take another clean container. Place a 0.053 mm sieve on it and gently pour the sewage from step (4) onto the sieve; (7) Repeat step (4). After completion, remove the sieve and place the soil on the sieve into a pre-weighed sample tube to obtain micro-aggregates (0.053~0.2534 mm); (8) Pour the remaining sewage into a 1 L centrifuge tube (pre-weighed) and centrifuge at 3220 RFC for 15 min; (9) Discard the supernatant to obtain powder-mucilage aggregates (<0.053 mm); (10) Dry the separated aggregates in a 50 °C oven to constant weight, weigh, bag, and label for testing.

[0056] Soil density grouping was performed using the improved Fu Jiping method: 5.0 g of air-dried soil (passed through a 0.25 mm sieve) was weighed and placed in a pre-weighed 100 mL centrifuge tube. 25 mL of a heavy liquid with a relative density of 1.8 g·cm⁻³ was added to the centrifuge tube. The tube was capped and placed vertically in a trough-type ultrasonic cleaning tank filled with ice water, maintaining the ice water level in the cleaning tank close to the liquid level in the centrifuge tube. The mixture was ultrasonically dispersed at 21.5 kHz and 300 mA for 10 min to ensure thorough dispersion. After ultrasonic dispersion, the soil sample was centrifuged at 3000 rpm for 10 min. At this point, the lighter fractions were suspended above the heavy liquid, while the heavier fractions settled at the bottom. The heavy liquid containing the suspended lighter fractions was then poured into a glass funnel lined with filter paper for filtration. More heavy liquid with a relative density of 1.8 g·cm⁻³ was added to the soil sample in the centrifuge tube, and the above process was repeated until no lighter fractions remained in the centrifuged heavy liquid. Generally, 2-3 separations were sufficient to remove the lighter fractions. The recombinant in the centrifuge tubes was washed three times with 95% ethanol, then twice with water. Finally, the centrifuge tubes containing the recombinant were dried in a forced-air drying oven at a temperature not exceeding 40°C, weighed, and packaged for testing. The weight of the light fraction was calculated using the difference method, and the mass percentages of the light fraction and recombinant were calculated separately. The product of the organic carbon content of the light fraction and recombinant and their respective mass percentages gives the organic carbon content of the light fraction and recombinant in the soil.

[0057] Inert organic carbon in soil was determined by acid hydrolysis. 0.25 g of air-dried soil (passed through a 2 mm sieve) was weighed into a crucible, and 20 ml of 6 mol / L HCl was slowly added. The mixture was digested on a hot plate at 115 °C for 16 h. After cooling, the soil was washed with distilled water until neutral, dried, and milled through an 180 μm sieve. The soil was then bagged, labeled, and ready for analysis.

[0058] In step S3, the specific formula for calculating the density of each organic carbon component in the soil profile is as follows:

[0059]

[0060] in, Indicates organic carbon components. This represents the total density of organic carbon component x in the 0-80cm soil layer. Indicates the number of soil layers. Represents the i-th soil layer. This represents the organic carbon content of organic carbon component x in the i-th soil layer, expressed in g·kg⁻¹. -1 , This represents the soil bulk density of the i-th soil layer, expressed in g·cm³. -3 , This represents the thickness of the i-th soil layer, in cm, 10 -1 This indicates the unit conversion factor, which converts the units to t·hm. -2 The coefficient.

[0061] In step S4, the apparent carbon fixation rate is calculated as follows: the difference between the total density of the same organic carbon component in the later developmental stage and the total density in the previous developmental stage is calculated, and then the difference in total density is divided by the difference between the average age of the stand in the later developmental stage and the average age of the stand in the previous developmental stage. The result is the average apparent carbon fixation rate of the organic carbon component from developmental stage m to stage n.

[0062] The specific formula for the apparent carbon fixation rate is:

[0063]

[0064] in, This represents the apparent carbon fixation rate from developmental stage to stage b carbon component x. and Let x represent the organic carbon density of carbon component x at stages a and b, respectively. and These represent the average age of stands in stages a and b, respectively.

[0065] In step S5, the forest stand characteristic factors include: stand age, stand density, average diameter at breast height (DBH), average tree height, and litter biomass; the soil environmental factors include: soil clay content and soil pH; and the current organic carbon density status includes: macroaggregate organic carbon density, microaggregate organic carbon density, silty aggregate organic carbon density, light aggregate organic carbon density, recombinant organic carbon density, inert organic carbon density, and total organic carbon density.

[0066] like Figure 2 As shown, in step S5, the machine learning model is a multilayer perceptron neural network model, wherein the number of neurons in the input layer of the multilayer perceptron is 14, which is the same as the dimension of the input features; the number of hidden layers is set to 2, with 128 neurons in the first hidden layer and 64 neurons in the second hidden layer; and the number of neurons in the output layer is 7, which predicts the carbon fixation rate of 6 organic carbon components and the total organic carbon fixation rate.

[0067] The densities of the seven organic carbon components include:

[0068] Macro-aggregate organic carbon density (Ma-OC);

[0069] Micro-aggregate organic carbon density (Mi-OC);

[0070] Organic carbon density of powdered agglomerates (Silt-ClayOC Density, SC-OC);

[0071] Light fraction organic carbon density (LFOC);

[0072] Heavy Fractional Organic Carbon Density (HFOC);

[0073] Inert organic carbon density (RIOC);

[0074] Total Organic Carbon Density (TOC) of Soil;

[0075] In the multilayer perceptron neural network model, the ReLU function is used as the activation function in the hidden layer; the linear activation function is used as the activation function in the output layer; the Adam optimization algorithm is used for training; K-fold cross-validation is used to evaluate the model performance; a weighted mean square error is constructed as the loss function, wherein the weights of the loss function are determined according to the ecological stability of each organic carbon component; the weight coefficient of organic carbon components with high stability is greater than the weight coefficient of organic carbon components with low stability.

[0076] Assign appropriate weights to each carbon component based on its ecological stability:

[0077] ω_RIOC (inert organic carbon weight) = highest, for example, set to 3.0;

[0078] ω_HFOC (recombined organic carbon weight) = high, for example, set to 2.0;

[0079] ω_Ma-OC (large aggregate carbon weight) = medium, for example, set to 1.5;

[0080] ω_Mi-OC, ω_SC-OC (micro-aggregates, carbon weights of powder and clay particles) = baseline weight 1.0;

[0081] ω_LFOC (light group organic carbon weight) = relatively low, for example, set to 0.8 (because it has high activity, fast turnover, large short-term changes, and relatively small long-term predictive significance).

[0082] ω_TOC (total organic carbon weight) = 1.0 or derived from other components, and may not require a separate weight setting.

[0083] Learning rate: The initial learning rate is set to 0.001, and a learning rate scheduler can be set to automatically reduce the learning rate when the validation set loss no longer decreases, for fine-tuning.

[0084] Batch size: Depending on the sample size, the batch size can be set to 8 or 16.

[0085] Regularization: Using Dropout in the hidden layer to randomly drop a portion of neurons effectively prevents the model from overfitting.

[0086] Validation and Early Stopping: K-fold cross-validation is used to rigorously evaluate model performance. Early stopping is used during training; training is automatically terminated when the validation set loss no longer decreases for several consecutive epochs (e.g., 10 epochs), and the model parameters that best perform on the validation set are restored.

[0087] The formula for calculating the loss function is as follows:

[0088]

[0089] in, Represents the loss function. This represents the total number of training data. This represents the weighting coefficient assigned to the predicted value of the x-th organic carbon component. This represents the actual measured carbon fixation rate. This indicates that the model predicts the carbon fixation rate.

[0090] In step S6, the problem diagnosis and risk level assessment specifically include:

[0091] The predicted future carbon fixation rate of each component is compared with the historical average rate baseline of that component and the expected rate baseline of that developmental stage.

[0092] A comprehensive risk score is automatically calculated based on the degree of deviation, the absolute value, and the ecological importance weight of the component.

[0093] Based on the risk score, the problem is classified into different levels and precisely located to the specific soil layer.

[0094] Historical average rate baseline: The average carbon sequestration rate of the same component across all plots and all developmental stages. Expected rate baseline for developmental stage: The average carbon sequestration rate of the component at a specific developmental stage (e.g., across all “middle-aged forest” plots).

[0095] The formula for calculating the comprehensive risk score is as follows:

[0096]

[0097] in, This represents the overall risk score for carbon component x. This indicates the percentage deviation of the predicted value from the historical global average. This indicates the percentage deviation of the predicted value from the expected level at its developmental stage. This represents the predicted carbon fixation rate of carbon component x. This indicates an indicator function. The function value is 1 when the condition within the parentheses is true, i.e., the predicted rate becomes negative; otherwise, it is 0. These represent weighting coefficients, used to adjust the relative importance of the three indicators. For example, you can set ( =0.3, =0.4, =0.3), indicating a greater focus on deviations from the stage baseline.

[0098] The system automatically classifies risk levels based on the calculated comprehensive risk score:

[0099] High-risk warning (overall risk score ≥ 2.0): The system highlights it in red and issues an immediate alert.

[0100] Moderate concern (0.5 ≤ Comprehensive risk score < 2.0): Highlighted in yellow by the system, attention is recommended.

[0101] Status is good (overall risk score <0.5): The system is marked green, no action is required.

[0102] In step S6, generating decision recommendations based on the multi-objective optimization problem specifically includes:

[0103] The decision-making recommendation is transformed from a single problem of maximizing carbon sequestration into a multi-objective optimization problem. The multi-objective optimization includes: ecological benefit optimization, that is, maximizing the carbon sequestration rate of stable carbon components; economic benefit optimization, that is, simulating the timber revenue brought about by different thinning intensities under thinning scenarios; and ecological security optimization, that is, ensuring that key indicators such as predicted soil erosion rate and water conservation capacity do not exceed the safety threshold after the implementation of management measures, such as thinning.

[0104] Actionable management measures are transformed into continuous variables that the model can understand (e.g., thinning intensity: 0%, 10%, 20%...; soil conditioner application rate: 0, 1, 2 t·ha). -1 wait);

[0105] A multi-objective optimization algorithm is used to search for the Pareto optimal solution set among the combination of measures;

[0106] For representative schemes on the Pareto front, Monte Carlo simulations are performed using a multilayer perceptron neural network model to give the probability distribution of the expected results.

[0107] Objective 1: Maximize ecological benefits (stable carbon sequestration):

[0108] The function F1 is the weighted sum of the carbon sequestration rates of each organic carbon component predicted by the machine learning model and their corresponding eco-importance weight coefficients. The weights are consistent with the definitions in the loss function and risk scoring, thus ensuring that the optimization direction is to improve the carbon sequestration capacity of stable carbon pools such as RIOC and HFOC.

[0109] Objective 2: Maximize economic benefits (timber revenue):

[0110] The timber revenue of function F2 is calculated based on the parameters of tending and harvesting intensity, current stand volume and timber market price in the management plan. The function aims to quantify and evaluate the direct economic value that the management plan can generate.

[0111] Objective 3: Minimize ecological security risks (constraints):

[0112] Treated as a constraint rather than an optimization objective, business plan X must satisfy a series of ecological security constraints. For example:

[0113] Soil erosion risk ≤0.5 (safe threshold)

[0114] The change in water conservation capacity is ≥-10% (it cannot be more than 10% lower than the current level).

[0115] Use constrained multi-objective optimization algorithms, such as NSGA-II.

[0116] Output: The algorithm will search for a Pareto optimal solution set. Each solution in this set represents an "optimal" business plan that cannot be further improved among multiple objectives (F1, F2) (i.e., improving one objective will inevitably lead to a decline in another objective).

[0117] Monte Carlo simulation: For candidate solutions X on the Pareto solution set, the system performs Monte Carlo simulation. This involves adding random noise (reflecting data uncertainty) to the input features, running the prediction model hundreds of times, and obtaining the probability distributions of F1 and F2 (such as 90% confidence intervals) to assess the risk and robustness of each solution.

[0118] A final solution can be selected from the Pareto optimal solution set based on actual preferences:

[0119] Ecosystem preference: Choose the option that maximizes F1 performance.

[0120] Preferred economy: Choose the option that maximizes F2.

[0121] Compromise solution: Choose the option located at the "inflection point" on the Pareto front, where a greater improvement in ecological benefits can be achieved at a smaller economic cost.

[0122] like Figure 4 As shown, the carbon fixation rates of different components are Ma-OC > TOC > RIOC > HFOC > LFOC > Mi-OC > SC-OC. Overall, the carbon fixation rates of all seven organic carbon components decrease with the forward progression of the developmental stage, and remain relatively stable in the mature to over-mature forest stage.

[0123] The examples described herein are merely preferred embodiments of the invention and are not intended to limit the concept and scope of the invention. Any modifications and improvements made by those skilled in the art to the technical solutions of the invention without departing from the design concept of the invention should fall within the protection scope of the invention.

Claims

1. A method for dynamically evaluating the soil carbon sequestration rate of Pinus massoniana plantation, characterized in that, Comprising the following steps: Step S1, select a development sequence sample site of Pinus massoniana plantation containing five development stages of young forest, middle-aged forest, near-mature forest, mature forest and over-mature forest in the target area, collect soil samples at different depths and pretreat them; Step S2, group the collected soil samples and determine the content of each group component and total organic carbon in the soil; Step S3, combine the soil bulk density and soil thickness to calculate the density of each organic carbon component in the soil profile; Step S4, based on the organic carbon density data of different development stage sample sites, calculate the apparent carbon sequestration rate of each organic carbon component between adjacent development stages; Step S5, construct a machine learning model, take stand characteristic factors, soil environmental factors and current organic carbon density state as inputs, take the carbon sequestration rate of each organic carbon component in the future stage as output, train and verify the model; Step S6, use the trained model to predict the future carbon sequestration rate, conduct problem diagnosis and risk level assessment, and generate decision suggestions based on multi-objective optimization problems; In step S5, the stand characteristic factors include stand age, stand density, average diameter at breast height, average tree height and litter biomass; the soil environmental factors include soil clay content and soil pH value; the current organic carbon density state includes macroaggregates organic carbon density, microaggregates organic carbon density, silt-clay aggregates organic carbon density, light fraction organic carbon density, heavy fraction organic carbon density, inert organic carbon density and total organic carbon density; In step S5, the machine learning model is a multilayer perceptron neural network model, wherein the number of input layer neurons of the multilayer perceptron is 14, which is the same as the input feature dimension; the number of hidden layers is set to 2, the number of neurons in the first hidden layer is 128, and the number of neurons in the second hidden layer is 64; the number of output layer neurons is 7, that is, the carbon sequestration rates of 6 organic carbon components and the total organic carbon carbon sequestration rate are predicted; In the multilayer perceptron neural network model, the hidden layer activation function uses the ReLU function; the output layer activation function uses the linear activation function; the training uses the Adam optimization algorithm; the model performance is evaluated by K-fold cross-validation; the weighted mean square error is constructed as the loss function, wherein the weights of the loss function are determined according to the ecological stability of each organic carbon component; the weight coefficient of the organic carbon component with high stability is greater than that of the organic carbon component with low stability.

2. The method of claim 1, wherein, Step S2 specifically includes: obtaining macroaggregates, microaggregates and silt-clay aggregates components by physically grouping soil samples, obtaining light fraction and heavy fraction organic carbon components by density grouping, and obtaining inert organic carbon component by chemical grouping.

3. The method of claim 2, wherein, In step S2, the physical grouping uses wet sieving method, the density grouping uses the improved method of Follin-Tyurin, and the chemical grouping uses acid hydrolysis method.

4. The method of claim 3, wherein, In step S3, the formula for calculating the density of each organic carbon component in the soil profile is: wherein, denotes the organic carbon component, denotes the total density of the organic carbon component x in the 0-80 cm soil layer, denotes the number of soil layers, denotes the i-th soil layer, denotes the organic carbon content of the organic carbon component x in the i-th soil layer, denotes the soil bulk density of the i-th soil layer, denotes the thickness of the i-th soil layer, 10 -1 denotes the unit conversion factor.

5. The method of claim 4, wherein, In the step S4, the calculation method of the apparent carbon sequestration rate is: calculating the difference between the total density of the same organic carbon component in the next development stage and the total density in the previous development stage, and then dividing the total density difference by the difference between the stand average age in the next development stage and the stand average age in the previous development stage, and the result is the average apparent carbon sequestration rate of the organic carbon component from the development stage m to stage n.

6. The method of claim 5, wherein, In the step S6, the problem diagnosis and risk level assessment specifically include: Comparing the predicted future carbon sequestration rate of each component with the historical average rate baseline of the component and the expected rate baseline of the development stage; According to the deviation degree, the absolute value and the ecological importance weight of the component, an integrated risk score is automatically calculated; According to the risk score, the problem is divided into different levels, and is accurately positioned to the specific soil layer.

7. The method of claim 6, wherein, In the step S6, the decision suggestion based on the multi-objective optimization problem specifically includes: The single carbon sequestration maximization problem is converted into a multi-objective optimization problem, wherein the multi-objective optimization includes: ecological benefit optimization, i.e. maximizing the carbon sequestration rate of stable carbon components; economic benefit optimization, i.e. simulating the timber yield brought by different thinning intensities under the thinning scenario; ecological safety optimization, i.e. ensuring that the key indicators such as the predicted soil erosion rate and water conservation capacity do not exceed the safety threshold after the implementation of management measures such as thinning; The operable management measures are converted into continuous variables understandable by the model; A multi-objective optimization algorithm is used to search for a Pareto optimal solution set in the measure combination; For the representative schemes on the Pareto frontier, a multi-layer perception neural network model is called to perform Monte Carlo simulation to give the probability distribution of the expected effect.

Images