Forest aboveground carbon storage promotion method and system based on forest structure diversity threshold regulation

Abstract

The application provides a forest above-ground carbon storage improvement method and system based on forest structure diversity threshold regulation, belonging to the field of ecology and computer technology. The method first obtains the diameter at breast height data of individual trees in the sample plot, calculates the forest structure diversity index, and determines the structure target threshold T based on the nonlinear relationship between the index and the above-ground carbon storage. If the structure diversity index of the current sample plot is less than T, the individual with the smallest diameter at breast height value is deleted according to the preset screening ratio, the thinning is carried out according to the preset number of trees ratio, and the small-diameter individual of the missing tree species is added. After recalculating the index, the processing rule corresponding to T is output. The method can change the structure target threshold from a static constant to a dynamic variable that drifts synchronously with forest succession, realize the continuous closed-loop evolution of the management strategy, and significantly improve the precision and adaptability of forest carbon sink management at different succession stages.

Description

Technical Field

[0001] This invention relates to the fields of ecology and computer technology, and more specifically, to a method and system for enhancing forest aboveground carbon storage based on threshold regulation of stand structure diversity. Background Technology

[0002] Forest aboveground carbon storage is a crucial component of the terrestrial carbon cycle, and its formation is influenced by a combination of factors, including stand diameter class structure, vertical stratification, and tree species composition. In forest carbon sequestration management, existing methods typically focus on single indicators such as stock volume, afforestation density, or tree species diversity, and formulate management plans based on these indicators. However, these methods rarely consider whether the complexity of the stand structure reaches a critical level as a criterion for determining carbon sequestration capacity, resulting in a lack of ecological threshold support for management objectives. Furthermore, existing methods often lack the ability to continuously track the dynamic evolution of stand structure after management measures are formulated, making it difficult to adaptively adjust management strategies according to changes in stand status. Their applicability is limited across different successional stages and spatial locations, failing to fully respond to the dynamic changes in forest ecosystems. Summary of the Invention

[0003] In view of the above problems, the present invention provides a method for improving forest aboveground carbon storage based on stand structure diversity threshold regulation, the method comprising: Obtain the diameter at breast height (DBH) data of individual trees within the target forest quadrat to form a DBH dataset; Based on the diameter at breast height (DBH) values ​​of each individual tree in the aforementioned DBH dataset, the stand structure diversity index is calculated. Obtain the nonlinear relationship between the forest stand structure diversity index and the aboveground carbon storage of the quadrat, and determine the structural target threshold T based on the nonlinear relationship; Compare the forest stand structure diversity index with the structure target threshold T; If the forest stand structure diversity index is less than the structure target threshold T, then the following processing is performed on the diameter at breast height (DBH) dataset: According to the first preset screening ratio, individuals with the smallest chest diameter value and / or a competitive pressure index greater than a preset threshold are selected from the chest diameter dataset and deleted. According to the second preset screening ratio, some individuals are further screened out from the diameter at breast height (DBH) dataset so that the number of individuals retained in the DBH dataset reaches the preset plant number ratio. Add at least one set of data for individuals of a preset tree species whose diameter at breast height (DBH) is less than a preset threshold to the DBH dataset; The stand structure diversity index was recalculated based on the processed diameter at breast height (DBH) dataset. Determine whether the recalculated stand structure diversity index reaches or exceeds the structure target threshold T; Record the first preset screening ratio, the second preset screening ratio, and / or the number of additional preset tree species and individuals corresponding to the recalculated stand structure diversity index reaching or exceeding the structure target threshold T, and output them as recommended processing rules.

[0004] In addition, the present invention provides a forest aboveground carbon storage enhancement system based on stand structure diversity threshold regulation, which is used to store and execute the above-described method.

[0005] Based on the above, the method and system provided by this invention transform the threshold effect between the stand structure diversity index and aboveground carbon storage into an operable management standard. This upgrades forest carbon sink management from an extensive model relying on traditional indicators such as stock volume or tree species diversity to an indicator-based control model guided by the structural threshold T. This method only requires diameter at breast height (DBH) data to diagnose stand structure status and screen management plans, significantly reducing the data acquisition cost and technical threshold for pre-management assessment compared to directly measuring aboveground carbon storage.

[0006] The dynamic evolution control steps of this invention enable the system to continuously track the effects of management measures by recording the deviation between the actual structural diversity index and the predicted value and generating a deviation time series. Deviations are no longer simply viewed as errors or noise, but are classified into three different types: natural drift deviation, external disturbance deviation, and phase change precursor deviation. These drive differentiated responses such as threshold adaptive updating, screening step size switching, and parallel maintenance of multiple hypotheses and rolling replanning. This mechanism reconstructs prediction deviations from "defects that need to be eliminated" in traditional methods into "information sources driving model evolution," transforming the structural target threshold T from a static constant into a dynamic, active variable that can drift synchronously with stand succession. The parallel maintenance of multiple hypotheses and Bayesian confidence update mechanism enable the system to have continuous response capabilities in the face of uncertainty, rather than discrete binary judgments of achievement or non-achievement. This effectively solves the shortcomings of traditional management methods, where once a plan is formulated, it operates in an open-loop manner for a long time and cannot be corrected based on intermediate state feedback.

[0007] The spatial co-evolutionary regulation of this invention, through the construction of a spatial association network of quadrats and the iterative updating of strategy propagation, variation and adoption weights, upgrades the management decisions of each quadrat in the region from isolated formulation to group co-evolution. Successful strategies can be propagated to neighboring quadrats through the spatial network, enabling local optimization experience to be promoted and reused at the regional scale, significantly improving the efficiency of multi-stand co-management.

[0008] The ultra-fine control mode switching of this invention automatically refines the screening step size and expands the candidate rule space after identifying pre-phase transition deviations. In the critical region where the structural diversity index approaches the threshold, a finer control granularity is employed, effectively avoiding over- or under-intervention in critical quadrats due to excessively large control step sizes. This improves the accuracy of operational measures in the phase transition sensitive region. The deviation multi-timescale decomposition mechanism breaks down the deviation signal into short-term fluctuations and long-term trends, respectively driving differentiated threshold update strategies under transient disturbance, structural drift, and mixed disturbance modes. This enables the system to accurately distinguish between random noise and systemic structural changes, effectively suppressing the interference of short-term random fluctuations on the threshold drift direction and improving the accuracy and robustness of adaptive threshold updates.

[0009] By combining the synergistic effects of the above-mentioned multiple levels, this invention can significantly improve the scientificity and accuracy of forest aboveground carbon storage management. It can achieve adaptive optimization of management strategies under different succession stages, different spatial locations, and different disturbance conditions, providing a quantifiable, operable, and scalable intelligent decision-making tool for enhancing forest carbon sequestration. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the system provided by the present invention.

[0011] Figure 3 This is a schematic diagram comparing the predicted values ​​with the measured values ​​of this invention.

[0012] Figure 4 This is a schematic diagram of the structural exponential evolution and threshold adaptive drift process of the present invention. Detailed Implementation

[0013] like Figures 1-4 As shown.

[0014] This technical solution provides a method and system for enhancing forest aboveground carbon storage based on threshold regulation of stand structure diversity. The technical details are explained in detail below using a specific application scenario.

[0015] In this embodiment, data from fixed sample plots within the temperate forest region of Changbai Mountain are used as the basis for illustration. This region encompasses three successional stages: secondary poplar-birch forest, secondary mixed coniferous and broadleaf forest, and broadleaf Korean pine forest. It covers the main transitional types from early secondary forests to climax communities, making it suitable as a research and validation area for forest carbon sequestration management methods.

[0016] Step S100: Obtain the diameter at breast height (DBH) data of individual trees within the target forest quadrat to form a DBH dataset.

[0017] Fixed sample plots were established within the target forest area, and each sample plot was divided into multiple consecutive small quadrats. In this embodiment, each quadrat was 20 meters by 20 meters. Within each quadrat, all woody plant individuals with a diameter at breast height (DBH) meeting the survey criteria were measured individually, and the quadrat number, tree species name, DBH value, tree height, and successional stage or forest type information of each individual were recorded.

[0018] Taking the temperate forests of Changbai Mountain as an example, the diameter at breast height (DBH) of each individual tree was measured using a DBH measuring tape at a height of 1.3 meters above the ground, with an accuracy of 0.1 centimeters. The DBH data for all individuals within each quadrat were organized according to the quadrat number. The total DBH values ​​of all individual trees in each quadrat constituted the DBH dataset for that quadrat. Each data entry in the DBH dataset contained at least three fields: quadrat number, tree species name, and DBH value. Tree height and succession stage information could also be included as needed.

[0019] Step S200: Calculate the stand structure diversity index based on the diameter at breast height (DBH) values ​​of each individual tree in the DBH dataset.

[0020] In this step, the preferred stand structure diversity index is the diameter-at-breast-width (DBH) Gini index. The DBH Gini index is used to characterize the unevenness of the distribution of DBH values ​​among individual trees within a quadrat. The closer the index value is to 0, the more uniform the DBH of the trees within the quadrat and the weaker the structural differentiation; the larger the index value, the more uneven the DBH distribution, the more obvious the differences between small-diameter, medium-diameter, and large-diameter trees, and the higher the complexity of the stand structure.

[0021] The Gini index for chest diameter is calculated as follows.

[0022] First, extract the diameter at breast height (DBH) values ​​of all individual trees within the quadrat from the DBH dataset, and remove missing values ​​and obvious outliers.

[0023] Then arrange the chest diameter values ​​in ascending order.

[0024] The heterogeneity of the sternal diameter structure at the quadrat scale was then calculated using the Gini index formula. The formula is:

[0025] Where n is the number of individual trees in the quadrat participating in the calculation, x k This represents the diameter at breast height (DBH) of the kth tree after sorting by DBH from smallest to largest, where k is the sorted sequence number.

[0026] The calculated Gini coefficient for diameter at breast height (DBH) was used as the stand structure diversity index for the quadrat. This index effectively reflects the degree of diameter class differentiation within the quadrat. When the index is low, the DBH distribution within the quadrat is concentrated and the structural hierarchy is simple; when the index is medium, there is a certain degree of diameter class difference within the quadrat, and the stand structure begins to differentiate; when the index is high, small-diameter, medium-diameter, and large-diameter trees coexist within the quadrat and their proportions are relatively balanced, indicating a high degree of stand structure complexity.

[0027] Step S300: Obtain the nonlinear relationship between the stand structure diversity index and the aboveground carbon storage of the quadrat, and determine the structural target threshold T based on this nonlinear relationship.

[0028] This step consists of two sub-stages: the first stage is to obtain or construct the response relationship between structural diversity and carbon storage, and the second stage is to determine the specific value of the threshold T based on this relationship.

[0029] In the first phase, the stand structure diversity index *x* and the corresponding aboveground carbon storage index *y* of multiple quadrats within the target forest area were obtained. The aboveground carbon storage index was obtained as follows: the diameter at breast height (DBH) and tree height of each individual tree were converted into individual tree biomass using the allometric growth equations for each tree species; this biomass was then multiplied by the carbon content conversion factor for the corresponding tree species to obtain the individual tree aboveground carbon storage; finally, the aboveground carbon storage of all individual trees was summed to obtain the total aboveground carbon storage of the quadrat. Quadrats with missing data were removed, and the remaining quadrats were sorted in ascending order of their structure diversity index *x* to form the modeling dataset.

[0030] In the second stage, a multi-model comparison method is used to determine the threshold T. The specific implementation method is as follows.

[0031] Linear, quadratic, and generalized additive models were fitted respectively as control models for the threshold-free relationship. The linear model was used to test whether there is a simple linear relationship between structural diversity and carbon storage, the quadratic model was used to test whether there is a monotonically increasing or decreasing nonlinear relationship, and the generalized additive model was used to test whether there are more complex nonlinear forms.

[0032] We fitted three threshold models: a step-type threshold model, a piecewise threshold model, and a smooth piecewise threshold model. The step-type threshold model assumes a sudden jump in the relationship between structural diversity and carbon storage at a certain point of change. The piecewise threshold model assumes that the relationships before and after the point of change follow different linear or nonlinear relationships. The smooth piecewise threshold model assumes that the transition near the point of change is smooth rather than abrupt. Each threshold model outputs a candidate point of change during the fitting process, which represents the potential turning point in the relationship between structural diversity and carbon storage.

[0033] Calculate the AIC value (Akaike Information Criterion value), overall R² value, and P-value for all models, including the control model and the threshold model. For the threshold model, calculate the R² and P-values ​​before and after the threshold to verify the accuracy and significance of the fit on both sides of the threshold. Sort the comparison results of each model according to the AIC value from smallest to largest. Select the threshold model with the smallest AIC value from all successfully fitted threshold models as the optimal threshold model. Determine the change point output by this optimal threshold model as the structural target threshold T.

[0034] Based on this, using a generalized additive model and the Bootstrap resampling method, the overall response curve of structural diversity and aboveground carbon storage and its confidence interval are calculated. The Bootstrap resampling method generates multiple resampled samples by repeatedly sampling the original data with replacement. The response curve is then repeatedly fitted on each resampled sample to obtain the statistical confidence interval of the response curve, which is used to evaluate the identification stability of the threshold T.

[0035] Step S400: Compare the stand structure diversity index with the structure target threshold T.

[0036] Obtain the stand structure diversity index value of the current quadrat from step S200, and compare it with the structure target threshold T determined in step S300. The comparison result is used to determine the stand structure status of the current quadrat.

[0037] Specifically, the comparison operation outputs a comparison value or a comparison result identifier, indicating whether the current structural diversity index is lower than T, close to T, or has already reached T. This comparison result provides a basis for subsequent classification processing.

[0038] Step S500: If the stand structure diversity index is less than the structure target threshold T, then perform processing operations on the diameter at breast height (DBH) dataset.

[0039] When the comparison result of step S400 shows that the stand structure diversity index of the current quadrat is less than the structure target threshold T, it indicates that the structural complexity of the quadrat has not yet reached the high carbon sink response state, and the diameter structure and density structure of the quadrat need to be actively adjusted. This processing operation consists of three sub-steps executed sequentially.

[0040] Sub-step S510: According to the first preset screening ratio, without deleting the target tree, large-diameter healthy tree and structural support tree, select and delete the individual with the smallest diameter at breast height (DBH) value and the individual with the competitive pressure index greater than the preset threshold from the DBH dataset.

[0041] The first preset screening ratio is used to control logging intensity and is expressed as the percentage of the vertical area at breast height. The first preset screening ratio has four levels: no logging (0%), light logging (10%), moderate logging (20%), and heavy logging (30%).

[0042] During this sub-step, individuals with the smallest sternum diameter (TBC) value are prioritized from the quadrat TBC dataset. When TBC values ​​are the same, the competitive pressure index of the individuals is further examined. The competitive pressure index can be the number of neighboring individuals within a certain radius around the individual, the ratio of the individual's TBC to the average TBC of the surrounding individuals, or a competition index calculated based on spatial coordinates. When the competitive pressure index is greater than a preset threshold, the individual is also included in the deletion list. For individuals with the smallest TBC value and a competitive pressure index greater than the threshold, they are deleted from the TBC dataset, while large-diameter individuals and structurally supportive individuals are retained. The deletion operation controls the proportion of the total cross-sectional area deleted to the total cross-sectional area of ​​the quadrat before processing according to the first preset screening ratio.

[0043] Sub-step S520: According to the second preset screening ratio, further screen out some individuals from the diameter at breast height (DBH) dataset so that the number of individuals retained in the DBH dataset reaches the preset plant number ratio.

[0044] The second preset screening ratio is used to control the thinning intensity and is expressed as the percentage of plants retained. The second preset screening ratio has four levels: no thinning (100% of plants retained), light thinning (90% of plants retained), moderate thinning (75% of plants retained), and heavy thinning (60% of plants retained).

[0045] In this sub-step, based on the deletion operation in sub-step S510, some individuals are further screened out from the remaining diameter at breast height (DBH) dataset. Screening prioritizes individuals in areas with excessively high local density and individuals of the same DBH class that are too densely packed. By removing some individuals, the local density of the quadrat is reduced, freeing up growth space and reducing excessive competition among individuals of the same DBH class. After screening, the number of individuals remaining in the DBH dataset reaches the preset plant count ratio.

[0046] Sub-step S530: Add at least one individual data point of a preset tree species whose diameter at breast height (DBH) is less than a preset threshold to the DBH dataset.

[0047] This sub-step is performed after harvesting and thinning are completed. The additional individuals are small-diameter specimens of tree species missing from the quadrat. The predefined tree species are native or valuable species determined based on zonal vegetation patterns and succession directions. The added individuals have a diameter at breast height (DBH) less than a predefined threshold and are typically seedlings or saplings, used to form a regeneration layer.

[0048] The number of individuals of the pre-defined tree species to be added corresponds to four levels of richness: no replanting (0 species), low richness (1 to 5 species), medium richness (6 to 10 species), and high richness (11 or more species). The addition operation adds the diameter at breast height (DBH) value, tree species name, and quadrat number of the new individuals to the DBH dataset, so that the updated DBH dataset includes both the original retained individuals and the added small-diameter individuals.

[0049] In a specific embodiment, the first preset screening ratio, the second preset screening ratio, and the richness level of the replanted tree species used in the three sub-steps can be selected independently. For example, a combination of moderate logging (20% of the basal area), moderate thinning (retaining 75% of the trees), and high richness replanting (11 or more tree species) can be selected; alternatively, a combination of weak logging (10% of the basal area), moderate thinning (retaining 75% of the trees), and high richness replanting can be selected. Different combinations correspond to different management strategies and are applicable to quadrats with different initial structural states.

[0050] Step S600: Recalculate the stand structure diversity index based on the processed diameter at breast height (DBH) dataset.

[0051] After completing the three sub-steps of step S500, the processed diameter at breast height (DBH) dataset is used as input, and the stand structure diversity index is recalculated using the same method as in step S200. The recalculated index value reflects the change in quadrat structure complexity after applying specific first preset screening ratios, second preset screening ratios, and replanting combinations. This recalculated index value is used to determine whether the processing operation successfully brought the quadrat to the target structural threshold T.

[0052] Step S700: Determine whether the recalculated stand structure diversity index has reached or exceeded the structure target threshold T.

[0053] The forest stand structure diversity index recalculated in step S600 is compared with the structure target threshold T determined in step S300. If the recalculated index value is greater than or equal to T, the treatment operation is deemed to have successfully enabled the quadrat to achieve a high carbon sink response structure state; if the recalculated index value is less than T, the treatment operation is deemed to have failed to meet the target.

[0054] Step S800: Record the first preset screening ratio, the second preset screening ratio, and the number of additional preset tree species and individuals corresponding to the recalculated stand structure diversity index reaching or exceeding the structure target threshold T, and output them as recommended processing rules.

[0055] For processing operations deemed compliant in step S700, the first preset screening ratio, the second preset screening ratio, and the additional preset tree species types and individual numbers used in the processing operation are extracted, and these parameters are combined into a recommended processing rule. This recommended processing rule corresponds to a combination of management measures that enable the quadrat to cross the structural target threshold T.

[0056] In practice, multiple treatment operations with different parameter combinations can be tried on the same initial quadrat. Each time, different combinations of the first preset screening ratio, the second preset screening ratio, and the replanting level are used, and steps S500 to S700 are executed respectively. All parameter combinations that cause the recalculated stand structure diversity index to reach or exceed the threshold T are recorded. In the output stage, one or more recommended treatment rules with the highest interrupted area retention rate or the best overall performance can be output according to a certain priority.

[0057] In addition, after completing the above steps, the basal area retention rate is calculated and output. The basal area retention rate is calculated as follows: the sum of the basal area of ​​the retained individuals in the processed diameter at breast height (DBH) dataset is divided by the sum of the basal area of ​​the DBH of all individuals in the unprocessed DBH dataset; the resulting ratio is the basal area retention rate. When outputting the recommended treatment rule, the corresponding basal area retention rate is also output, allowing management decision-makers to assess the degree to which the scheme preserves forest stand volume.

[0058] When applied to temperate forests in Changbai Mountain, this method can be further differentiated by succession stage. Secondary poplar-birch forests in the succession stage have low initial structural diversity in their quadrats and exhibit a more pronounced response to harvesting and replanting. A combination of moderate harvesting, moderate or light thinning, and high replanting targets is suitable. Secondary coniferous-broadleaf mixed forests in the succession stage have low initial target achievement rates but possess a certain structural foundation. A robust combination of light or moderate harvesting and moderate thinning is suitable, focusing on cultivating medium to large diameter trees. Broadleaf Korean pine forests in the succession stage have high initial target achievement rates, with a management focus on maintenance. A combination of thinning without harvesting is suitable to avoid structural degradation caused by strong disturbance.

[0059] The above describes in detail the complete operation process of the method of the present invention in the temperate forest of Changbai Mountain. The parameter selection in each step can be adjusted according to the actual survey data and management objectives of the target forest area, and is not limited to the specific values ​​given in the embodiment. The determination of the structural target threshold T adopts a data-driven multi-model comparison method, which can adapt to the actual conditions of different forest types and regions, and has good scalability and flexibility.

[0060] After step S400, the comparison results between the forest stand structure diversity index and the structure target threshold T are further refined into three categories of state determination.

[0061] If the current stand structure diversity index is less than 0.95T, the quadrat is marked as a Class I non-compliant quadrat. These quadrats typically have a simple structure and insufficient diameter differentiation, requiring structural regulation to promote the coexistence of small-diameter, medium-diameter, and large-diameter trees and increase the complexity of the stand structure.

[0062] If the current stand structure diversity index is greater than or equal to 0.95T and less than T, then the quadrat is marked as a Class II critical quadrat. These quadrats already possess a certain structural foundation and are close to a high carbon sink response structure. Excessive disturbance should be avoided during management, and mild to moderate control measures are suitable to help them stably reach the target threshold.

[0063] If the current stand structure diversity index is greater than or equal to T, then the quadrat is marked as a third-category qualified quadrat. This type of quadrat already has a high degree of structural complexity, and the management focus is not on forcibly increasing the index, but on maintaining multi-diameter structure, protecting large-diameter individuals, and preventing structural degradation.

[0064] For the first type of sub-quadrates that did not meet the standards, the first preset screening ratio used in subsequent processing should be greater than that used for the second type of borderline quadrats, in order to provide more sufficient structural adjustment. For the third type of sub-quadrates that met the standards, there is no need to delete the diameter at breast height (DBH) dataset; the current state can be maintained directly.

[0065] In step S500, the processing operation is performed using a single tree as the smallest operating unit. The specific process is as follows.

[0066] First, the original individual tree diameter at breast height (DBH) data of the quadrat were read, and the stand structure diversity index before treatment was calculated as the baseline value.

[0067] Then, the individuals within the quadrat were processed according to the harvesting rules. The harvesting rules approximate the harvesting intensity using the ratio of height-to-diameter (HDB) to cross-sectional area, prioritizing the selection of individuals with the smallest HDB value, while also considering competition pressure indicators. Individuals with competition pressure indicators exceeding a preset threshold were also prioritized for processing. During the processing, target trees, large-diameter individuals, and structurally supporting individuals were protected and not deleted.

[0068] After harvesting, the remaining individuals are processed according to the thinning rules. The thinning rules express the thinning intensity as the percentage of plants retained, and further screening out some individuals to reduce local density, free up growth space, and reduce excessive competition between individuals of the same diameter class.

[0069] After thinning is completed, replanting is performed. The replanting rules are implemented after harvesting and thinning, appending the small diameter (MD) data of missing tree species from the quadrats to the diameter at breast height (DBH) dataset. These appended individuals are then used in the subsequent recalculation of the structural diversity index.

[0070] After completing the above three operations, a database of individual trees after management is generated, and the stand structure diversity index is recalculated based on this database. Then, it is determined whether the recalculated index reaches or exceeds the structure target threshold T. Simultaneously, evaluation indicators such as structure improvement, target achievement rate, basal area retention rate, tree number retention rate, and number of replanted trees are calculated.

[0071] When outputting the recommended processing rules in step S800, multiple candidate combinations are pre-constructed for enumeration and comparison.

[0072] These candidate combinations were established with reference to the relevant principles of the Forest Tending Regulations and the Changbai Mountain Forest Area Forest Tending Technical Regulations. Specifically, the candidate combinations include the following nine groups.

[0073] The first group is a blank control combination with no logging, no thinning, and no replanting. The second group is a density-controlled combination with no logging and intensive thinning. The third group is a combination with no logging, intensive thinning, and high replanting targets, combined with density control and replanting. The fourth group is a low-disturbance combined combination with weak logging, weak thinning, and high replanting targets. The fifth group is a low-disturbance optimized combination with weak logging, moderate thinning, and high replanting targets. The sixth group is a robust optimized combination with moderate logging, weak thinning, and high replanting targets. The seventh group is a comprehensive optimized combination with moderate logging, moderate thinning, and high replanting targets. The eighth group is a replanting intensity candidate combination with moderate logging, moderate thinning, and moderate replanting targets. The ninth group is a reinforcement modification candidate combination with intensive logging, no thinning, and high replanting targets.

[0074] The nine combinations mentioned above are not a simple enumeration of all theoretical combinations, but rather representative combinations selected based on principles such as moderate intervention, protection of preserved trees, maintenance of continuous stand cover, promotion of natural regeneration, and formation of multi-path structure in near-natural management. Extreme combinations involving simultaneous heavy logging and heavy thinning, even if achieving high simulation compliance rates, are not considered standard recommended solutions.

[0075] By simulating and comparing the recalculated stand structure diversity index and basal area retention rate of the above nine combinations, the combination that achieves the index target and has a better basal area retention rate is selected as the recommended processing rule.

[0076] After step S800 is completed, differentiated recommendation schemes for different succession stages are further output.

[0077] Simulation results show that during the succession stage of secondary poplar and birch forests, the initial compliance rate is low, but the response to management and control is significant. A comprehensive optimization combination of moderate logging, moderate thinning, and high replanting targets is suitable, as is a robust optimization combination of moderate logging, weak thinning, and high replanting targets. Management should focus on proactive structural optimization and long-term multi-pathway structure cultivation.

[0078] For secondary mixed coniferous and broadleaf forests in their succession stage, the initial compliance rate is the lowest, and the improvement after a single management operation is limited, making it difficult to significantly exceed the structural target threshold in the short term. A low-disturbance optimization combination of light harvesting, moderate thinning, and high replanting targets is suitable, or a robust optimization combination of moderate harvesting, light thinning, and high replanting targets. Management should focus on long-term retention of dominant trees, cultivation of medium and large-diameter trees, and control of competition from excessively dense small-diameter trees; strong disturbance methods to pursue short-term compliance are not advisable.

[0079] For broadleaf Korean pine forests in the succession stage, the initial compliance rate is already relatively high, and the stands themselves have significant structural complexity. Suitable regeneration combinations include no logging combined with intensive thinning and high replanting targets, or no logging combined with intensive thinning without replanting for density control. The management focus should shift from enhancement to maintenance, avoiding structural degradation caused by intensive logging and thinning, and prioritizing structural maintenance, protection of large-diameter trees, and moderate thinning.

[0080] Based on all simulation results, the optimal combination of moderate logging, moderate thinning, and high replanting targets shows the most balanced performance between structural compliance rate and basal area retention rate, and can be considered the primary scheme. The low-disturbance optimal combination of light logging, moderate thinning, and high replanting targets has a high basal area retention rate and is suitable for stands with high ecological protection requirements and an existing structural foundation. The regeneration combination of no logging, strong thinning, and high replanting targets is suitable for plots where logging adjustments are not advisable but where there is excessive stand density or insufficient regeneration layer.

[0081] In a specific embodiment of the temperate forests of Changbai Mountain, a fixed quadrat of secondary coniferous and broad-leaved mixed forest with an area of ​​20 meters by 20 meters was selected as the treatment object. The initial stand structure diversity index of this quadrat was calculated to be 0.52. Using the multi-model comparison method described in step S300, based on the structure diversity and aboveground carbon storage data of 390 quadrats in this area, the target threshold T for the structure of this area was determined to be 0.65. Since the current structure diversity index of 0.52 is less than the threshold of 0.65, this quadrat was judged to be a substandard quadrat and required processing.

[0082] After processing the diameter at breast height (DBH) dataset of the sample plot and outputting recommended processing rules, the dynamic evolution control step is initiated.

[0083] First, the deviation recording operation is performed. The specific meaning of the deviation value is defined in this embodiment as follows: the deviation value equals the actually observed stand structure diversity index minus the predicted stand structure diversity index corresponding to the recommended treatment rule. The predicted stand structure diversity index here refers to the expected structure diversity index achieved on the diameter at breast height (DBH) dataset after implementing the recommended treatment rule. This predicted value comes from the result recalculated based on the processed DBH dataset after processing; that is, the index value calculated from the processed DBH dataset serves as the predicted value corresponding to the recommended treatment rule.

[0084] In this embodiment, the observation period is defined as the time interval between two chest diameter data collections, set to 2 years. The end time of each observation period is the observation time, and the interval between adjacent observation times is equal to the length of the observation period, which are equidistant time points.

[0085] The initial treatment of this quadrat employed a comprehensive optimized combination of moderate logging, moderate thinning, and high replanting targets. After treatment, the stand structure diversity index was recalculated based on the treated diameter at breast height (DBH) dataset, yielding a predicted value of 0.68. However, at the end of the actual observation period, the actual stand structure diversity index, recalculated based on newly collected DBH data from the field, was 0.62. The bias of this observation was 0.62 minus 0.68, equaling -0.06.

[0086] Over three consecutive observation periods, the deviation value after each processing step was recorded. The deviation value for the first observation period was -0.06, for the second -0.09, and for the third -0.13. These three deviation values ​​were arranged chronologically according to the observation periods to generate the deviation time series for the quadrat. The deviation time series is the core data foundation for the dynamic evolution control steps, recording the trajectory of the deviation between the predicted and actual values ​​of the quadrat over multiple observation periods.

[0087] After the deviation time series is generated, a pattern recognition operation is performed. In this embodiment, the following three deviation patterns are defined.

[0088] The first type is natural drift bias. When the bias value in the bias time series shows a consistent trend of change in the same direction for more than three consecutive observation periods, this trend is identified as natural drift bias. A consistent trend of change in the same direction means that the bias value continuously increases or decreases between adjacent observation periods. The criterion is that the sign of the difference between the bias values ​​in adjacent periods remains consistent. In this example, the bias value changes from -0.06 to -0.09 and then to -0.13, with adjacent differences of -0.03 and -0.04, both negative values, and the sign remains consistent, therefore it is identified as natural drift bias. This pattern indicates that the stand structure of the quadrat is undergoing a systematic directional change, possibly due to the advancement of natural stand succession or the cumulative effect of long-term management measures, causing a systematic deviation in the original prediction model's estimation of the structural diversity index.

[0089] The second type is external disturbance bias. When the change in the bias value within a single observation period exceeds a preset step threshold, it is identified as external disturbance bias. In this embodiment, the preset step threshold is set to 0.15. For example, if within a certain observation period, the bias value suddenly drops from -0.05 in the previous period to -0.25, a change of 0.20, exceeding the step threshold of 0.15, then this bias is identified as external disturbance bias. External disturbance bias is usually caused by unrecorded external interference events, including but not limited to pest and disease outbreaks, extreme weather events, and human activity interference.

[0090] The third type is the pre-phase change bias. When the variance of the bias value continuously increases over multiple consecutive periods, and the change in the mean of the bias value is less than a preset change threshold, it is identified as a pre-phase change bias. In this embodiment, the variance is calculated using the sample variance method, with the calculation window being the most recent 5 observation periods. Continuous increase means that the variance value increases period by period for at least 3 consecutive periods, and the increase in each period is not less than the preset minimum change threshold. The preset change threshold is set to 0.02. When the variance of the bias value increases from 0.01 to 0.03 and then to 0.06, while the mean of the bias value only changes from -0.08 to -0.09, with a change of only 0.01, which is less than the preset change threshold of 0.02, this pattern is identified as a pre-phase change bias. The pre-phase change bias indicates that the stand structure diversity index of the sample plot may be approaching the critical region of the structure target threshold T, and the system's sensitivity to disturbances is increasing.

[0091] After pattern recognition is completed, the subsequent processing strategy is adjusted based on the identified deviation patterns.

[0092] When natural drift bias or external disturbance bias is identified, the prediction error of candidate processing rules is corrected using a bias time series. Specifically, the cumulative trend direction and cumulative magnitude of the bias time series are used as error compensation parameters and superimposed on the evolution prediction values ​​of subsequent candidate processing rules. In this embodiment, the bias time series exhibits a continuous negative drift trend with a cumulative trend slope of -0.035. This slope value is used as the error compensation parameter, and when generating the evolution prediction values ​​of the next round of candidate processing rules, this error compensation is uniformly applied to the prediction values ​​of each candidate rule at different observation times. Actual observation values ​​are still acquired independently and used to calculate the new round of bias values ​​and update the confidence levels of each candidate rule. This error compensation mechanism does not adjust the value of the structural target threshold T itself, but rather corrects the model's prediction bias of the operating effect, so that subsequent rule confidence updates are based on a more accurate prediction benchmark. The structural target threshold T remains unchanged at 0.65.

[0093] When a phase change precursor deviation is detected, a screening step size switching operation is performed. In this embodiment, the first step size corresponds to the first preset screening ratio in claim 5, which is divided into four levels with 10% intervals: 0%, 10%, 20%, and 30%. The second preset screening ratio is divided into four levels with plant retention ratios: 100%, 90%, 75%, and 60%. The second step size is smaller than the first step size. Specifically, it is implemented by inserting one intermediate level between each two adjacent levels, expanding the number of levels from four to seven. After switching from the first step size to the second step size, the screening granularity in subsequent processing steps is significantly refined, allowing for more precise adjustments to management measures. This switching only takes effect within a limited period after the phase change precursor deviation is identified. In this embodiment, the limited period is set to three consecutive observation periods.

[0094] While performing the aforementioned error compensation and screening step size switching, the system continuously runs the confidence update mechanism for candidate processing rules. After acquiring new diameter at breast height (DBH) data and recalculating the structural diversity index in each round, the system calculates the matching probability between the predicted value and the actual observed value of each candidate processing rule at the observation time, and updates the confidence of each candidate processing rule based on the matching probability. The error compensation operation under the natural drift bias mode makes the predicted values ​​of each candidate rule closer to the actual observed values, thereby improving the screening accuracy of high-confidence rules.

[0095] When newly acquired quadrat data containing the forest stand structure diversity index and corresponding aboveground carbon storage are obtained, the system re-executes the determination step of the structure target threshold T based on all quadrat data, i.e., refitting the nonlinear threshold model and identifying change points. In this embodiment, after accumulating monitoring data for 8 consecutive observation periods, i.e., 16 years, the threshold model is refitted based on the updated data of all 390 quadrats, and the resulting change point is 0.64. The deviation of this value from the currently used structure target threshold of 0.65 is -0.01. The preset threshold update tolerance is set to 0.015. Since the absolute value of the actual deviation of -0.01 is less than the preset threshold update tolerance of 0.015, it is determined that this change belongs to the normal fluctuation in the threshold identification process and is not a systematic change, and the system maintains the current structure target threshold of 0.65 unchanged. If the deviation between the re-determined structure target threshold and the current structure target threshold exceeds the preset threshold update tolerance, for example, if the absolute value of the deviation is greater than 0.015, then the current structure target threshold is updated to the re-determined structure target threshold.

[0096] When a phase change precursor deviation is detected, a screening step size switching operation is performed. In this embodiment, the first step size corresponds to a first preset screening ratio with 10% intervals for level division, namely 0%, 10%, 20%, and 30%; and a second preset screening ratio with plant retention ratios for level division, namely 100%, 90%, 75%, and 60%. The second step size is smaller than the first step size. Specifically, it is implemented by inserting one intermediate level between each two adjacent levels, expanding the number of level divisions from 4 to 7. After switching from the first step size to the second step size, the screening granularity in subsequent processing steps is significantly refined, allowing for more precise adjustments to management measures. This switching only takes effect within a limited period after the phase change precursor deviation is identified. In this embodiment, the limited period is set to 3 consecutive observation periods.

[0097] Before each processing step, the system generates in parallel the predicted evolutionary values ​​of the stand structure diversity index for multiple subsequent observation times corresponding to multiple candidate processing rules. In this embodiment, the number of candidate processing rules is equal to the Cartesian product of the number of first preset screening ratio levels, the number of second preset screening ratio levels, and the number of replanting levels. Under the first step, the number of levels is 4 times 4 times 3, which equals 48 candidate rules. Each candidate processing rule corresponds to a combination of the first preset screening ratio, the second preset screening ratio, and preset tree species. The evolutionary prediction values ​​are generated by trend extrapolation of the currently observed deviation time series, specifically using a linear extrapolation method to predict the expected structure diversity index for the next three observation times.

[0098] After acquiring new diameter at breast height (DBH) data and recalculating the stand structure diversity index, the matching probability between the predicted and actual observed values ​​for each candidate treatment rule at the observation time is calculated. The matching probability is calculated as follows: based on the assumption that the prediction error follows a normal distribution, the probability value is determined by calculating the position of the actual observed value within the confidence interval of the predicted value. In this embodiment, the confidence interval width of the predicted value is set to the predicted value plus or minus 0.05.

[0099] The confidence score of each candidate processing rule is updated based on the matching probability. The initial confidence score is set to 1 divided by the total number of candidate rules, i.e., 1 divided by 48. The confidence score is updated using a Bayesian update method: the posterior confidence score equals the prior confidence score multiplied by the matching probability, and then divided by the sum of the matching probabilities of all candidate rules. This update method ensures that the sum of the confidence scores of all candidate rules is always equal to 1.

[0100] Candidate processing rules with confidence levels below a preset elimination threshold are eliminated. In this embodiment, the preset elimination threshold is set to 0.01. When the confidence level of a candidate rule is below 0.01, the rule is determined to be invalid and will no longer participate in subsequent candidate processing rule enumeration. Candidate processing rules with confidence levels above a preset retention threshold are retained as an active candidate rule set. In this embodiment, the preset retention threshold is set to 0.05. Candidate rules with confidence levels above 0.05 are retained in the active candidate rule set. The active candidate rule set is a continuously maintained subset of rules; the confidence level is recalculated and the set is updated after each round of new observation data arrives.

[0101] In the next round of processing, the candidate processing rule with the highest confidence in the active candidate rule set is output as the recommended processing rule, while the remaining active candidate processing rules are retained as alternative rules. When the deviation between the actual observed value and the predicted value of the candidate processing rule with the highest confidence in the next round exceeds the preset deviation tolerance band, the current recommended processing rule is switched to the alternative rule with the highest confidence among the alternative rules. In this embodiment, the preset deviation tolerance band is a bilaterally symmetrical interval, with an upper limit of the positive deviation allowable range of the structural target threshold T, i.e., +0.05, and a lower limit of the negative deviation allowable range, i.e., -0.05. Assuming that the actual observed value in the next round is 0.58, while the predicted value of the candidate rule with the highest confidence is 0.65, the deviation is -0.07, which exceeds the lower limit of the tolerance band of -0.05, then the rule switching operation is triggered, and the recommended processing rule is switched to the rule with the highest confidence among the alternative rules.

[0102] Through the aforementioned dynamic evolutionary control steps, the structural target threshold T of the quadrat gradually shifted from an initial 0.65 to 0.6325 to adapt to the systematic changes brought about by natural stand succession. Simultaneously, the active candidate rule set always maintained multiple alternative schemes. When the actual observed value deviated significantly from the predicted value of the master rule, the system could automatically switch to the alternative rule best suited to the current state, without waiting for the complete observation period to end before re-formulating a scheme. This mechanism upgrades forest carbon sink management from discrete, periodic decision-making to continuous, rolling decision-making, significantly improving the response speed of management strategies to dynamic changes in the stand.

[0103] Building upon the secondary coniferous and broadleaf mixed forest quadrats described above, this embodiment further extends to multiple quadrats across the entire target forest area. A total of 390 consecutive quadrats were established within this area, each independently undergoing dynamic evolutionary regulation as described above. These quadrats are not completely isolated but rather form a co-evolving whole through a spatial network.

[0104] In this embodiment, the spatial association network is constructed as follows: 390 quadrats within the target forest area are used as nodes in the network. The edge weights between nodes are determined by two factors: the geographical distance between quadrats and the similarity of forest stand types. The geographical distance is calculated using the spatial coordinates of the quadrat's center point. The similarity of forest stand types is calculated based on the tree species composition vector of each quadrat, which records the importance values ​​of each tree species within the quadrat. The specific calculation method for edge weights is as follows: first, the geographical distance between any two quadrats is calculated, and then the cosine similarity of the tree species composition vectors is calculated. The similarity is multiplied by an exponential decay factor of the geographical distance to obtain the final edge weight value. This exponential decay factor controls the boundary of the spatial influence range. When the geographical distance between two quadrats is greater than a preset distance threshold, no edge connection is established between them. In this embodiment, the preset distance threshold is 200 meters. This distance threshold is determined based on the spatial autocorrelation scale of the temperate forests of Changbai Mountain, ensuring that the edge connection can reflect the actual spatial transmission range of the forest stand structure characteristics.

[0105] After the boundary weights are determined, the strategy propagation operation is performed. The goal of strategy propagation is to spread successful strategies from high-performing quadrats to neighboring quadrats. In this embodiment, a strategy is considered successful only if it meets the following two conditions simultaneously: First, the actual stand structure diversity index of the quadrat has reached or exceeded the structure target threshold T, meaning the quadrat is in a high carbon sink response structure state. Second, the deviation between the actual stand structure diversity index and the predicted stand structure diversity index corresponding to the recommended treatment rule falls within the preset deviation tolerance zone for two consecutive observation periods, indicating that the prediction effect of the strategy has high stability.

[0106] In this embodiment, the preset deviation tolerance zone is a bilaterally symmetrical interval with a lower limit of -0.05 and an upper limit of +0.05. The current actual stand structure diversity index of this quadrat is 0.66, and the target threshold T is 0.65, meaning the actual index has reached and exceeded the threshold. Simultaneously, the deviation values ​​for two consecutive observation periods are -0.02 and -0.01, respectively, both falling within the tolerance zone of -0.05 to +0.05. The simultaneous satisfaction of these two conditions indicates that the quadrat has not only reached a high carbon sink response structure state, but also that the strategy's prediction of the structure diversity index maintains consistency with the actual observation results. Therefore, the screening ratio and replanting combination currently used in this quadrat are deemed a successful strategy combination.

[0107] If the actual stand structure diversity index still does not reach the structure target threshold T, even if the deviation value falls within the tolerance zone, it indicates that although the strategy is accurate in its prediction, its regulatory strength is insufficient to make the stand cross the threshold, and it does not meet the criteria for a successful strategy. Therefore, this strategy combination does not trigger strategy propagation operations.

[0108] The successful strategy combination propagates to neighboring sample squares through the edges of the spatial association network. The propagation strength is controlled by the edge weights. Specifically, the larger the edge weight, the higher the propagation strength. In this embodiment, the propagation strength is equal to the edge weight multiplied by a global propagation attenuation coefficient, which is set to 0.8. A neighboring sample square is defined as a sample square directly connected by an edge in the spatial association network, i.e., a first-order neighbor node. When a successful strategy combination propagates from the source sample square to a first-order neighbor sample square, its propagation strength is the edge weight multiplied by 0.8; when it continues to propagate to a second-order neighbor sample square, the propagation strength is multiplied by 0.8 again, attenuating hop by hop.

[0109] Simultaneously with the strategy propagation operation, a strategy mutation operation is performed. Strategy mutation applies a random perturbation to the received successful strategy combinations. The purpose of this perturbation is to explore combinations in the strategy space that may be superior to the original strategy. In this embodiment, the random perturbation uses a discrete uniform distribution, randomly selecting a perturbation value from a preset perturbation set. The perturbation set consists of negative 1, 0, and positive 1, corresponding to adjustments in the grade number. For example, the first preset screening ratio for a successful strategy combination is a moderate logging grade, corresponding to a grade index of 2. If the applied random perturbation is positive 1, then the grade index becomes 3, corresponding to a strong logging grade. The perturbation amplitude monotonically decreases with the distance between the current stand structure diversity index and the structure target threshold of the sample plot. The closer the distance, the smaller the perturbation amplitude. Specifically, this is achieved by multiplying the maximum perturbation amplitude by an exponential decay function, where the independent variable of the exponential decay function is the ratio of distance to a distance scale parameter. In this embodiment, the distance scale parameter is set to 0.1. This means that when the structural diversity index of a quadrat is far from the threshold of 0.65, the variation amplitude is large, and it retains a strong exploratory ability; when the distance approaches the threshold, the variation amplitude approaches zero, and it only retains the fine-tuning ability.

[0110] After the same observation period ends, the stand structure diversity index achievement rate of each quadrat after adopting different strategy combinations is statistically analyzed. The achievement rate is defined as the number of quadrats in which the recalculated stand structure diversity index reaches or exceeds the structure target threshold using a certain strategy combination, divided by the total number of quadrats using that strategy combination. Using this achievement rate as fitness, the adoption weight of each strategy combination in the quadrat population is iteratively updated. The adoption weight is updated using the reproducer dynamic equation. The core logic of this update method is: the adoption weight of strategy combinations with high fitness increases proportionally, while the adoption weight of strategy combinations with low fitness decreases proportionally. Specifically, the rate of change of the adoption weight of a strategy combination is equal to the current weight of that strategy combination multiplied by the fitness of that strategy combination minus the difference in the average fitness of all strategy combinations. In this embodiment, the initial adoption weight is set to be evenly distributed among the strategy combinations, i.e., each combination has an equal weight. After one iteration, the combination of moderate logging combined with moderate thinning and high replanting targets has a higher achievement rate, and its adoption weight increases from 1 / 48 to 0.09. The combination of intensive logging and intensive thinning had a low compliance rate and an excessively low sectional area retention rate, so its adoption weight decreased from 1 / 48 to 0.008.

[0111] When the adoption weight of any strategy combination is lower than a preset elimination threshold, the strategy combination is marked as an elimination strategy combination. In this embodiment, the preset elimination threshold is 0.005. When the adoption weight is lower than 0.005, the combination is determined to be a failed strategy and will no longer participate in the subsequent candidate processing rule enumeration. This mechanism effectively controls the size of the candidate rule space and avoids the waste of computing resources.

[0112] After the aforementioned strategy propagation, strategy mutation, and weighted iterative updates, the surviving strategy combination with the highest weight is selected as the recommended strategy set for the target forest area at the end of the current observation period. In this embodiment, the retained strategy combination with the highest weight is the combination of moderate logging combined with moderate thinning and high replanting targets, with a weight of 0.09. Each strategy combination in this recommended strategy set is associated with a spatial applicability range. The spatial applicability range is determined by the spatial distribution boundary of the sample plots using the strategy combination in the spatial association network. In this embodiment, the spatial distribution boundary is calculated as follows: extract the spatial coordinates of all sample plots using the same strategy combination, construct the convex hull polygon of these coordinate points, and the convex hull is the spatial applicability range of the strategy combination. According to this calculation method, the spatial applicability range of the combination of moderate logging combined with moderate thinning and high replanting targets covers approximately 120 sample plots in the central area of ​​the region, encompassing most of the secondary coniferous and broad-leaved mixed forest sample plots.

[0113] The above describes the specific implementation of spatial collaborative evolutionary regulation. This mechanism enables the management strategies of each sample group within the region to no longer be formulated in isolation, but rather to achieve the sharing of successful experiences through strategy propagation, the exploration of new combinations through strategy variation, and the emergence of collective wisdom through iterative weighting, ultimately forming a set of regional recommended strategies with spatial applicability.

[0114] Building upon spatial collaborative evolutionary regulation, this embodiment also introduces a feedforward compensation step for structural target thresholds. The threshold update in the aforementioned dynamic evolutionary regulation step is a feedback update, meaning the deviation is observed first and then the threshold is corrected. Feedforward compensation, on the other hand, pre-calculates the threshold drift based on long-term succession patterns before processing the circumference-of-the-horsepower dataset, achieving predictive compensation.

[0115] In this embodiment, long-term stand succession monitoring data for the temperate forest region of Changbai Mountain were first acquired. This data originated from continuous monitoring records of fixed plots over the past 30 years, covering the entire succession sequence from the early stages of secondary poplar-birch forests to the climax community of broadleaf Korean pine forests. From this monitoring data, time series of structural diversity indices for multiple historical quadrats at different stand ages were extracted. The structural diversity index values ​​for each historical quadrat were recorded at 5, 10, 15, 20, 25, and 30 years of stand age.

[0116] Logistic growth curves were fitted to the time series of each structural diversity index. Logistic growth curves are suitable for describing the S-shaped growth process of stand structural diversity as it gradually approaches its saturation capacity with increasing stand age. After fitting, two parameters were extracted from each growth curve: the natural growth rate parameter and the saturation capacity parameter. The natural growth rate parameter controls the steepness of the growth curve; a larger value indicates a faster rate of structural diversity growth with increasing stand age. The saturation capacity parameter represents the theoretical maximum value that the structural diversity index can reach under undisturbed conditions, i.e., the structural diversity level under climax community conditions.

[0117] Using the current stand age of the current quadrat as the query index (in this example, the current stand age is 25 years), the set of natural growth rate parameters and the set of saturation capacity parameters corresponding to a 25-year stand age are retrieved from the growth curves of all historical quadrats. Each set contains parameter values ​​from multiple historical quadrats. The median of each set is calculated as a representative value. In this example, the median of the natural growth rate parameter set is 0.015 per year, and the median of the saturation capacity parameter set is 0.72.

[0118] The representative value of the natural growth rate parameter, 0.015, is used as the succession driving input of the feedforward compensator. The representative value of the saturation capacity parameter, 0.72, is divided by the structural target threshold, 0.65, yielding a ratio of approximately 1.11, which serves as the saturation constraint coefficient of the feedforward compensator. The feedforward compensator calculates the feedforward compensation amount based on the succession driving input, the saturation constraint coefficient, and the time span since the last update of the structural target threshold. In this embodiment, the time span since the last threshold update is 2 years. The core logic of the feedforward compensation amount is: the succession driving input multiplied by the saturation constraint coefficient, and then multiplied by the time span. Specifically, the value is 0.015 multiplied by 1.11 and then multiplied by 2, equaling 0.0333.

[0119] The feedforward compensation of 0.0333 is added to the current structural target threshold of 0.65, resulting in a drifted structural target threshold of 0.6833. This drifted threshold replaces the original threshold and is used for subsequent index-threshold comparisons and processing steps. Due to the progression of succession, the structural diversity of 25-year-old plots should theoretically approach the saturation capacity of 0.72. Therefore, the threshold is appropriately increased to 0.6833 to avoid misjudging that the target has been met due to an excessively low threshold.

[0120] After processing and recalculating the stand structure diversity index, the actual deviation between the observed structure diversity index and the drifted target threshold is calculated. In this embodiment, the observed value is 0.67, the drifted threshold is 0.6833, and the actual deviation is -0.0133. This actual deviation is fed back to the feedforward compensator. The feedforward compensator adjusts the calculated gain of the feedforward compensation amount based on the cumulative direction and magnitude of the actual deviation. The gain is a proportional coefficient in the feedforward compensator, which amplifies or reduces the response magnitude of the feedforward compensation amount to the actual deviation. In this embodiment, the initial value of the gain is 1.0. When the actual deviation is negative, it indicates that the actual value is continuously lower than the predicted value, and the feedforward compensation amount needs to be reduced, thus lowering the gain. The specific adjustment method is: the gain equals the baseline gain of 1.0 minus the product of the absolute value of the cumulative deviation magnitude and the integral coefficient. The integral coefficient is set to 0.5. After this feedback, the gain is adjusted to 1.0 minus 0.0133 multiplied by 0.5, which equals 0.9934. The gain is limited between a preset minimum of 0.5 and a maximum of 2.0 to prevent overcompensation or undercompensation. The adjusted gain is used to calculate the feedforward compensation amount for the next round, allowing the compensation amount to adaptively change as the actual deviation accumulates.

[0121] Through the aforementioned feedforward compensation steps, the structural target threshold T can be predictively adjusted based on long-term succession patterns before processing operations are performed, and then continuously calibrated in the direction of threshold drift through feedback correction. This combined feedforward and feedback mechanism transforms the threshold T from a fixed static scale into a dynamic, active variable capable of simultaneously responding to long-term succession trends and short-term observation biases.

[0122] During the operation of the dynamic evolution control steps described above, the deviation time series of this secondary coniferous and broad-leaved mixed forest quadrats was identified as a precursor deviation to phase change in the fifth observation period. This identification was triggered by the following conditions: the variance of the deviation value increased from 0.012 to 0.035 and then to 0.068 over three consecutive periods, with increases of 0.023 and 0.033 respectively, both exceeding the preset minimum change threshold of 0.02. Simultaneously, the mean of the deviation value only changed from -0.07 to -0.09, a change of 0.02, less than the preset change threshold of 0.05. The satisfaction of these conditions indicates that the stand structure diversity index of this quadrats is approaching the critical region of the structure target threshold, and the system's sensitivity to disturbances is increasing, necessitating a switch to a highly refined control mode for the phase change sensitive region.

[0123] The first step in switching to the ultra-fine control mode is switching the screening step size. In this embodiment, the screening step size is a composite concept, encompassing the level intervals of both the first preset screening ratio and the second preset screening ratio. Under the first step size, the first preset screening ratio has four levels: 0%, 10%, 20%, and 30%, with intervals of 10%. The second preset screening ratio has four levels: 100%, 90%, 75%, and 60%, with intervals of plant retention. When a phase transition precursor deviation is detected, the screening step size is switched from the first step size to the second step size.

[0124] The second step length is shorter than the first step length. Specifically, it involves inserting an intermediate level between each of the two adjacent levels of the first step length. The value of this intermediate level is the arithmetic mean of the values ​​of the two adjacent levels. Taking the first preset screening ratio as an example, 5% is inserted between 0% and 10%, 15% between 10% and 20%, and 25% between 20% and 30%, expanding the number of levels from 4 to 7. Taking the second preset screening ratio as an example, 95% is inserted between 100% and 90%, 82.5% between 90% and 75%, and 67.5% between 75% and 60%, again expanding the number of levels from 4 to 7. The replanting target levels remain unchanged: no replanting, 1 to 5 types, 6 to 10 types, and 11 or more types (4 levels in total). After the switch, the candidate processing rule space expands from 4 x 4 x 4 = 64 candidate rules under the first step length to 7 x 7 x 4 = 196 candidate rules under the second step length. The expanded candidate processing rule space includes all combinations of existing levels under the first step length and newly added intermediate levels under the second step length, which significantly refines the granularity of operational measures adjustments.

[0125] In the second step, ultra-fine screening is performed on the diameter at breast height (DBH) dataset of the sample plot. This process is executed in the same way as the previous process, but with a denser candidate rule space and a finer screening step. During the ultra-fine screening process, the distance between the recalculated stand structure diversity index and the structure target threshold for each candidate processing rule is recorded. The distance value is defined as the absolute value of the arithmetic difference in this embodiment, i.e., the absolute value of the recalculated stand structure diversity index minus the absolute value of the structure target threshold T. For example, a candidate rule has a recalculated index of 0.64, a threshold T of 0.65, and a distance value of 0.01; another candidate rule has a recalculated index of 0.67 and a distance value of 0.02. The set of candidate rules with the smallest distance value is selected as the transitional recommended processing rule for the current sample plot. If multiple candidate rules have the same distance value, their sectional area retention rates are further compared, and the set with the highest sectional area retention rate is selected. In this embodiment, the minimum distance value is 0.005, and the corresponding candidate rule is a combination of a first preset screening ratio of 8%, a second preset screening ratio of 83%, and a replanting target of 6 to 10 species with medium richness, with a cross-sectional area retention rate of 86.3%.

[0126] The transitional recommendation processing rules are only effective for a limited period after the phase change precursor deviation is identified. The limited period is defined as three consecutive observation periods in this embodiment, starting from the moment the phase change precursor deviation is identified. Within this limited period, the quadrat uses the transitional recommendation processing rule with the minimum distance value and the highest basal area retention rate to perform subsequent processing operations. When the limited period ends, if the stand structure diversity index of the quadrat has reached or exceeded the structure target threshold T, the screening step size is switched back to the first step size, i.e., it reverts to the four-level classification method of 0%, 10%, 20%, 30% and retention of 100%, 90%, 75%, 60%.

[0127] Simultaneously, a set of rules from the transitional recommended processing rules that cause the forest stand structure diversity index to reach or exceed the threshold are included in the candidate processing rule base. The candidate processing rule base is a historical database storing validated recommended processing rules, stored in the non-volatile memory of the computer system, used to retain rule combinations that have been validated and effective in various scenarios. After rules are entered into the base, they are arranged in reverse chronological order of entry time, with a preset maximum storage capacity of 1000 rules. When the capacity is exceeded, the oldest rule is removed. Rules included in the candidate processing rule base are given priority in the subsequent candidate rule enumeration process, that is, as one of the default candidate rules, they are selected first during the traversal enumeration. In this embodiment, after the limited period ends, the structure diversity index of the sample plot is 0.66, which exceeds the threshold of 0.65. The screening step size is switched back to the first step size. The combination of the first preset screening ratio of 8%, the second preset screening ratio of 83% retention, and the richness in the replanting target is included in the candidate rule base and marked as the default candidate rule.

[0128] After identifying and finely controlling the phase transition precursor bias, the quadrat re-entered the regular dynamic evolution control process. In subsequent observation periods, the quadrat encountered another situation: the short-term and long-term components of the bias time series exhibited inconsistent trends. Therefore, this embodiment further performed multi-timescale decomposition and differential driving operations on the bias.

[0129] The core of multi-timescale decomposition lies in splitting the deviation time series into windows of different lengths to separate short-term fluctuation components and long-term trend components. In this embodiment, a short time window is defined as one observation period, i.e., 2 years. A long time window is defined as 5 observation periods, i.e., 10 years. The deviation time series is divided into multiple short-term deviation segments according to the short time windows. Each segment contains deviation values ​​within one observation period. The average deviation value within each segment is calculated, and the average values ​​of all segments are arranged in chronological order to generate the short-term deviation component sequence. Similarly, the deviation time series is divided into multiple long-term deviation segments according to the long time windows. Each segment contains deviation values ​​within 5 observation periods. The average deviation value within each segment is calculated, and the average values ​​of all segments are arranged in chronological order to generate the long-term deviation component sequence.

[0130] The short-term deviation component sequence reflects the fluctuation of deviations in the short term, while the long-term deviation component sequence reflects the trend of deviations in the long term. The short-term volatility variance of the short-term deviation component sequence is calculated; this variance is the variance between the means of each short-term deviation segment in the sequence, reflecting the magnitude of change in the mean deviation across different short-term windows. The long-term trend slope of the long-term deviation component sequence is calculated; this slope is obtained by performing a linear regression fit on the long-term deviation component sequence. The sign of the slope indicates the long-term direction of deviation change, and the absolute value of the slope represents the rate of change.

[0131] In this embodiment, the short-term volatility variance is calculated to be 0.018, and the long-term trend slope is calculated to be -0.008 per year. The preset short-term volatility threshold is 0.015, and the preset absolute value of the long-term trend threshold is 0.005. Since the short-term volatility variance of 0.018 is greater than the short-term volatility threshold of 0.015, and the absolute value of the long-term trend slope of 0.008 is greater than the long-term trend threshold of 0.005, the system determines that the current deviation mode is a mixed interference mode.

[0132] The mixed interference mode is defined as follows: the short-term volatility variance is greater than the short-term volatility threshold, while the absolute value of the long-term trend slope is greater than or equal to the long-term trend threshold. This mode indicates that the sample plot is simultaneously affected by the superposition of random environmental fluctuations and systemic structural evolution. Under this mode, the drift direction of the structural target threshold is determined by the sign of the long-term trend slope. A negative long-term trend slope indicates a downward drift. The drift step size is determined by the function value of the ratio of the short-term volatility variance to the long-term trend slope. In this embodiment, this ratio is 0.018 divided by 0.008, which equals 2.25.

[0133] The drift step size in the mixed interference mode is calculated as follows: The drift step size s is equal to the baseline step size s0 multiplied by an exponential decay function, i.e., s = s0·exp(-r / λ), where r is the ratio of the absolute value of the short-term volatility variance to the slope of the long-term trend, and λ is the scaling parameter. In this embodiment, the baseline step size s0 is 0.02, the ratio r is 2.25, and the scaling parameter λ is set to 3.0. Substituting the above values ​​into the formula, s = 0.02 × exp(-2.25 / 3.0) = 0.02 × exp(-0.75). exp(-0.75) is approximately equal to 0.472, and the drift step size s is calculated to be approximately 0.02 × 0.472 = 0.0094. This formula uses an exponential decay form, making the drift step size s monotonically decrease as the ratio r increases, thereby suppressing the interference of random fluctuations on the threshold drift direction and ensuring that the drift direction is mainly dominated by the long-term trend.

[0134] In contrast, if the short-term volatility variance is less than or equal to the short-term volatility threshold, and the absolute value of the long-term trend slope is greater than or equal to the long-term trend threshold, it is determined to be a structural drift mode. In this mode, the sign of the long-term trend slope determines the drift direction, and the magnitude of the long-term trend slope multiplied by the time step coefficient determines the drift step size. A continuous unidirectional drift correction is applied to the structural target threshold, and the adaptively updated gain state is switched to a high-gain state. In the high-gain state, the maximum single drift of the structural target threshold increases to three times the maximum drift in the transient disturbance mode. The maximum drift in the transient disturbance mode is 0.01, and the maximum single drift in the high-gain state is 0.03.

[0135] If the short-term volatility variance is greater than the short-term volatility threshold, and the absolute value of the long-term trend slope is less than the long-term trend threshold, it is determined to be a transient disturbance mode. In this mode, the current structural target threshold remains unchanged, and only the magnitude of the adaptive threshold update is subject to short-term suppression. Short-term suppression refers to multiplying the maximum adjustment amount of a single threshold by a decay factor of 0.3. Since the deviation in the transient disturbance mode mainly comes from short-term random factors rather than systemic structural changes, the threshold should not drift; therefore, its adjustment amount is compressed to 30% of the normal level through suppression. The suppression duration is the entire period during which the current deviation mode is determined to be a transient disturbance mode; once the mode switches, the suppression is lifted.

[0136] If the short-term volatility variance is less than or equal to the short-term volatility threshold, and the absolute value of the long-term trend slope is less than the long-term trend threshold, then it is determined to be a mode with no significant bias. In this mode, the biased time series has neither a clear trend nor significant volatility, and the system does not perform any threshold update operations, maintaining the current state.

[0137] Through multi-timescale decomposition and differentiated driving operations of the aforementioned biases, this quadrat can accurately distinguish different types of bias causes. In the mixed disturbance mode, the threshold drifts along the long-term trend, but the drift step size is suppressed by short-term fluctuations, avoiding interference from random fluctuations on the drift direction. In the structural drift mode, the threshold responds quickly to systemic structural changes with a larger step size. In the transient disturbance mode, the threshold remains unchanged, avoiding erroneous adjustments caused by short-term random fluctuations. These three modes cover various possible changes in the bias time series, providing clear judgment criteria and response rules for the updating behavior of the structural target threshold T in different scenarios, effectively improving the accuracy and robustness of adaptive threshold updates.

[0138] This specific implementation also includes an example of a forest aboveground carbon storage enhancement system based on stand structure diversity threshold regulation. The system uses a computer as its operating platform and includes a central processing unit (CPU), memory, non-volatile storage, and a network communication interface. The CPU is used to execute computational tasks. Memory is used for temporary storage of running programs and data. Non-volatile storage is used for persistent storage of diameter at breast height (DBH) datasets, structural target thresholds, and a candidate processing rule base. The network communication interface is used for data interaction with field observation equipment.

[0139] The data acquisition module connects to the field data acquisition terminal via a wired interface. Field observers use an electronic diameter gauge to measure the diameter at breast height (DBH) of each individual tree within a fixed quadrat, with a measurement accuracy of 0.1 cm. The measurement data is transmitted to the data acquisition terminal via Bluetooth, where the terminal records the quadrat number, tree species name, and DBH value for each individual tree. The data acquisition terminal then uploads the data to the data acquisition module via a wireless network. Upon receiving the data, the data acquisition module verifies the raw data, removing missing and obvious outliers. The verified data is then organized according to the quadrat number to form a DBH dataset. The DBH dataset is stored in tabular form in non-volatile memory, with each data entry containing three fields: quadrat number, tree species name, and DBH value.

[0140] The diversity index calculation module reads the diameter at breast height (DBH) dataset from non-volatile memory. This module first extracts the DBH values ​​of all individual trees within the quadrat and arranges them in ascending order. Then, it calculates the heterogeneity of the DBH structure at the quadrat scale using the DBH Gini index formula. The DBH Gini index calculation is performed in the central processing unit (CPU) and is a floating-point numerical operation. The result is the stand structure diversity index for that quadrat, stored in memory for subsequent modules to access. The index ranges from 0 to 1; a higher value indicates a more uneven distribution of DBH and a higher complexity of the stand structure.

[0141] The threshold determination module reads the stand structure diversity index and corresponding aboveground carbon storage index of each quadrat from memory. The aboveground carbon storage index is pre-calculated using the allometric growth equation. This module sequentially fits linear, quadratic, and generalized additive models as threshold-free control models, and fits step-threshold, piecewise, and smooth piecewise threshold models as threshold models. The fitting calculations for each model are executed in the central processing unit, using the least squares method to solve for the model parameters. The AIC value of each model is calculated, and the change point output by the threshold model with the smallest AIC value is determined as the structural target threshold T. After the threshold is determined, the T value is stored in non-volatile memory for subsequent modules to access.

[0142] The comparison module reads the stand structure diversity index of the current quadrat from memory and the structure target threshold T from non-volatile memory, then performs a numerical comparison operation. The comparison result is categorized into three cases: index less than T, index greater than or equal to T but less than 0.95T, and index greater than or equal to T. The comparison module outputs the comparison result as an enumerated set of values ​​for subsequent modules to use in their judgment. If the index is greater than or equal to T, the system directly outputs a suggestion to maintain the status quo, without needing to call the processing module.

[0143] When the comparison module outputs an exponent less than T, the processing module is triggered. The processing module filters out individuals with the smallest diameter at breast height (DBH) value from the DBH dataset and calculates the competition pressure index for each individual. The competition pressure index is calculated as the ratio of the number of neighboring individuals within a 5-meter radius of the individual to the average DBH value of the sample plot. The processing module deletes individuals with the smallest DBH value and competition pressure indices greater than a preset threshold from the DBH dataset according to a first preset screening ratio. The first preset screening ratio has four levels: 0%, 10%, 20%, and 30%. After deletion, the processing module further filters out some individuals according to a second preset screening ratio. The second preset screening ratio has four levels: retaining 100% of the trees, retaining 90% of the trees, retaining 75% of the trees, and retaining 60% of the trees. Screening prioritizes individuals in areas with excessively high local density and individuals with excessively dense clusters of the same diameter class. After screening, the processing module adds data on small-diameter individuals of a preset tree species. The added individuals have a DBH value less than 5 cm. The replanting levels are divided into four categories: 0 species, 1 to 5 species, 6 to 10 species, and 11 or more species. The processing module saves the processed diameter at breast height (DBH) dataset in memory, overwriting the original data.

[0144] The recalculation module reads the processed diameter at breast height (DBH) dataset from memory and recalculates the stand structure diversity index using the same method as the diversity index calculation module. The calculation result is a floating-point number and stored in memory.

[0145] The judgment module reads the recalculated forest stand structure diversity index and structure target threshold T from memory and performs a numerical comparison operation. If the recalculated index value is greater than or equal to T, the processing of this group is considered successful; otherwise, it is considered unsuccessful. The judgment result is stored in memory as a Boolean value.

[0146] The output module receives the judgment result from the judgment module. For processing operations judged as successful, the output module extracts the first preset screening ratio, the second preset screening ratio, the species and number of replanted trees used in the processing operation, and combines these parameters into a recommended processing rule. The output module also calculates the sectional area retention rate. The sectional area retention rate is calculated as follows: the sum of the sectional areas of the retained individuals in the processed diameter at breast height (DBH) dataset divided by the sum of the sectional areas of the DBH of all individuals in the pre-processed DBH dataset. The output module presents the recommended processing rule and its corresponding sectional area retention rate in the form of a data report on the display device for management decision-makers' reference. The output format is a table containing five fields: processing rule number, first preset screening ratio, second preset screening ratio, replanting level, and sectional area retention rate.

[0147] The modules of this system interact with each other via memory data transfer and function calls. The output of the data acquisition module serves as input to the diversity index calculation and comparison modules. The output of the comparison module determines whether to trigger the processing module. The output of the processing module serves as input to the recalculation module. The output of the recalculation module serves as input to the judgment module. The output of the judgment module serves as input to the output module. The data flow is unidirectional, and each module executes sequentially, forming a complete processing chain.

[0148] The system may further include a dynamic evolution control module. This module obtains recommended processing rules and corresponding predicted values ​​from the output module, and actual observed values ​​from the measured data. It calculates the deviation value and generates a deviation time series. Pattern recognition is performed on the deviation time series, and the threshold is adaptively updated based on the identified deviation patterns. Before processing, multiple candidate rules' evolution prediction values ​​are generated in parallel, the confidence scores of each rule are updated, and an active candidate rule set is maintained. In the next round of execution, the rule with the highest confidence score is output as the recommended processing rule. The specific workflow of the dynamic evolution control module is completely consistent with the dynamic evolution control steps in the aforementioned method embodiments, and will not be repeated here.

[0149] The system may further include a spatial co-evolution module. This module constructs a spatial association network of sample plots, performs policy propagation and policy mutation operations, iteratively updates the adoption weights of policy combinations, and generates a set of regional recommendation policies. The specific workflow of the spatial co-evolution module is completely consistent with the spatial co-evolution control steps in the aforementioned method embodiments.

[0150] The system may further include a feedforward compensation module. This module acquires long-term stand succession monitoring data, fits growth curves, extracts natural growth rate parameters and saturation capacity parameters, calculates the feedforward compensation amount, and applies it to the structural target threshold. The specific workflow of the feedforward compensation module is completely consistent with the feedforward compensation steps in the aforementioned method embodiments.

[0151] The system may further include an ultra-fine control module. Upon identifying a phase transition precursor deviation, the ultra-fine control module switches the screening step size, performs ultra-fine screening processing, selects transition recommendation rules, and adds the rules to the database after a limited period. The specific workflow of the ultra-fine control module is completely consistent with the ultra-fine control mode switching steps in the aforementioned method embodiments.

[0152] The system may further include a deviation multi-scale decomposition module. This module decomposes the deviation time series into short and long time windows, calculates the short-term volatility variance and long-term trend slope, and performs differentiated threshold-driven operations based on the judgment results for transient disturbance modes, structural drift modes, or mixed disturbance modes. The specific workflow of the deviation multi-scale decomposition module is completely consistent with the deviation multi-timescale decomposition and differentiated driving steps in the aforementioned method embodiments.

[0153] Through the coordinated operation of the above modules, the entire process from data collection to recommendation rule output is automated.

[0154] Through the coordinated operation of the above steps and modules, this embodiment fully demonstrates the closed-loop process from diameter at breast height (DBH) data collection to the generation of recommended processing rules. The process begins with the data acquisition module collecting and verifying field DBH data. This is followed by structural quantification by the diversity index calculation module, critical point identification by the threshold determination module, and state judgment by the comparison module. When structural diversity does not meet the standard, the processing module performs screening and replanting operations. Finally, the output module generates recommended rules including the basal area retention rate. Based on this, the dynamic evolution control module uses deviation time series to drive adaptive threshold updates, achieving continuous tracking of stand structure changes; the spatial co-evolution module promotes successful experiences at the regional scale through strategy propagation and variation among quadrats; the feedforward compensation module implements predictive compensation before threshold adjustment based on long-term succession patterns; the ultra-fine control module refines the control granularity in phase transition sensitive areas to avoid over- or under-intervention in critical quadrats; and the deviation multi-scale decomposition module decomposes the deviation signal into short-term fluctuations and long-term trends, driving differentiated threshold update strategies respectively. The above-mentioned mechanisms form an organic whole with the deviation signal as the link, so that the structural target threshold T is transformed from a static constant into a dynamic variable that can follow the co-evolution of the forest ecosystem.

[0155] This method and system are not dependent on specific regions or forest types. The determination of the structural target threshold T adopts a data-driven multi-model comparison method, which can be recalculated based on historical monitoring data from different forest types and regions. This method is also applicable to other forest types such as warm-temperate deciduous broad-leaved forests, subtropical evergreen broad-leaved forests, plantations, and natural secondary forests. The system's modules adopt a general computing architecture and can be deployed on existing forest management platforms, exhibiting good portability and scalability.

[0156] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art can make various modifications, equivalent substitutions, or improvements without departing from the principles of the present invention, and these modifications, equivalent substitutions, or improvements should also be considered as included within the scope of protection of the present invention.

Claims

1. A method for enhancing forest aboveground carbon storage based on stand structure diversity threshold regulation, characterized in that, The method includes: Obtain the diameter at breast height (DBH) data of individual trees within the target forest quadrat to form a DBH dataset; Based on the diameter at breast height (DBH) values ​​of each individual tree in the aforementioned DBH dataset, the stand structure diversity index is calculated. Obtain the nonlinear relationship between the forest stand structure diversity index and the aboveground carbon storage of the quadrat, and determine the structural target threshold T based on the nonlinear relationship; Compare the forest stand structure diversity index with the structure target threshold T; If the forest stand structure diversity index is less than the structure target threshold T, then the following processing is performed on the diameter at breast height (DBH) dataset: According to the first preset screening ratio, individuals with the smallest chest diameter value and / or a competitive pressure index greater than a preset threshold are selected from the chest diameter dataset and deleted. According to the second preset screening ratio, some individuals are further screened out from the diameter at breast height (DBH) dataset so that the number of individuals retained in the DBH dataset reaches the preset plant number ratio. Add at least one set of data for individuals of a preset tree species whose diameter at breast height (DBH) is less than a preset threshold to the DBH dataset; The stand structure diversity index was recalculated based on the processed diameter at breast height (DBH) dataset. Determine whether the recalculated stand structure diversity index reaches or exceeds the structure target threshold T; Record the first preset screening ratio, the second preset screening ratio, and / or the number of additional preset tree species and individuals corresponding to the recalculated stand structure diversity index reaching or exceeding the structure target threshold T, and output them as recommended processing rules.

2. The method according to claim 1, characterized in that, The stand structure diversity index is the diameter at breast height (DBH) Gini index, which characterizes the degree of unevenness in the distribution of DBH values ​​of individual trees within a quadrat.

3. The method according to claim 1, characterized in that, The method for determining the structural target threshold T includes: Obtain the stand structure diversity index x and aboveground carbon storage index y of multiple quadrats, and sort them in ascending order of x; Linear, quadratic, and generalized additive models were fitted respectively as threshold-free control models. Fit the step threshold model, the piecewise threshold model, and the smooth piecewise threshold model respectively; Calculate the AIC value for each model, and select the threshold model with the smallest AIC value from the successfully fitted threshold models; The change point output by the threshold model with the smallest AIC value is determined as the structural target threshold T.

4. The method according to claim 1, characterized in that, After comparing the forest stand structure diversity index with the structure target threshold T, the method further includes: If the current stand structure diversity index is less than 0.95T, the quadrat will be marked as the first type of state; If the current stand structure diversity index is greater than or equal to 0.95T and less than T, then the quadrat is marked as the second type of state; If the current stand structure diversity index is greater than or equal to T, then the quadrat is marked as the third type of state; Specifically, for the first type of state, the first preset screening ratio is greater than the first preset screening ratio used for the second type of state; for the third type of state, no deletion process is performed on the caliper diameter dataset.

5. The method according to claim 1, characterized in that, The first preset screening ratio includes four levels: 0%, 10%, 20%, and 30%, expressed as the percentage of the cross-sectional area at breast height; the second preset screening ratio includes four levels: retaining 100% of the trees, retaining 90% of the trees, retaining 75% of the trees, and retaining 60% of the trees; the richness corresponding to the number of individuals of the added preset tree species includes four levels: 0 species, 1 to 5 species, 6 to 10 species, and more than 11 species.

6. The method according to claim 1, characterized in that, Also includes: The ratio of the sum of the chest height cross-sectional areas of the individuals retained in the processed chest diameter dataset to the sum of the chest height cross-sectional areas of all individuals in the unprocessed chest diameter dataset is used as the cross-sectional area retention rate. When outputting the recommended processing rule, the cross-sectional area retention rate corresponding to the recommended processing rule is also output.

7. The method according to claim 1, characterized in that, After executing the processing and outputting the recommended processing rules, a dynamic evolution control step is also included: Record the deviation between the recalculated stand structure diversity index after each execution of the above processing and the predicted stand structure diversity index corresponding to the recommended processing rule, and generate a deviation time series. Pattern recognition is performed on the deviation time series: when the deviation value shows a trend of change in the same direction for three or more consecutive periods, the trend is identified as natural drift deviation; when the change amplitude of the deviation value in a single observation period exceeds a preset step threshold, it is identified as external disturbance deviation; when the variance of the deviation value continues to increase in multiple consecutive periods and the change in the mean of the deviation value is less than a preset change threshold, it is identified as a phase transition precursor deviation. The subsequent processing strategy is adjusted based on the identified deviation patterns: when the natural drift deviation or the external disturbance deviation is identified, the prediction error of the candidate processing rule is corrected using the deviation time series, and the cumulative trend direction and cumulative amplitude of the deviation time series are used as error compensation parameters and superimposed on the evolution prediction values ​​of each subsequent candidate processing rule without adjusting the structural target threshold T; when the phase transition precursor deviation is identified, the screening step size in the subsequent processing steps of the current sample plot is switched from the first step size to the second step size, where the second step size is smaller than the first step size; Before each execution of the processing, multiple candidate processing rules are generated in parallel to predict the evolution of the forest stand structure diversity index at multiple subsequent observation times. Each candidate processing rule corresponds to a combination of the first preset screening ratio, the second preset screening ratio, and the preset tree species. After acquiring new diameter at breast height (DBH) data and recalculating the stand structure diversity index, the matching probability between the predicted value and the actual observed value of each candidate processing rule at the observation time is calculated. The confidence level of each candidate processing rule is updated based on the matching probability. Candidate processing rules with confidence levels lower than a preset elimination threshold are eliminated, and candidate processing rules with confidence levels higher than a preset retention threshold are retained as an active candidate rule set. In the next round of processing, the candidate processing rule with the highest confidence in the active candidate rule set is used as the recommended processing rule for output, while the remaining active candidate processing rules are retained as alternative rules. When the deviation between the actual observed value in the next round and the predicted value of the candidate processing rule with the highest confidence exceeds the preset deviation tolerance band, the current recommended processing rule is switched to the alternative rule with the highest confidence among the alternative rules. When new sample plot data containing forest stand structure diversity index and corresponding aboveground carbon storage is obtained, the determination step of the structure target threshold T is re-executed based on all sample plot data. If the deviation between the re-determined structure target threshold and the current structure target threshold exceeds the preset threshold update tolerance, the current structure target threshold is updated to the re-determined structure target threshold.

8. The method according to claim 7, characterized in that, The dynamic evolution control steps also include spatial coordinated evolution control: Multiple quadrats within the target forest area are constructed into a quadrat spatial association network, where each quadrat is a node and the edge weights between nodes are determined by the geographical distance between quadrats and the similarity of forest stand types. A strategy propagation operation is performed on the quadrat spatial association network: when the deviation between the actual stand structure diversity index of any quadrat and the predicted stand structure diversity index corresponding to the recommended processing rule is less than the lower limit of the preset deviation tolerance zone for two consecutive observation periods, the combination of the first preset screening ratio, the second preset screening ratio and the preset tree species currently used in the quadrat is determined to be a successful strategy combination, and the successful strategy combination is propagated to neighboring quadrat through the edges of the quadrat spatial association network, with the propagation intensity controlled by the edge weights; Simultaneously with the execution of the strategy propagation operation, a strategy mutation operation is performed: for the sample plots that receive the successful strategy combination, a random perturbation is applied to the first preset screening ratio or the second preset screening ratio in the successful strategy combination with a preset mutation probability. The magnitude of the random perturbation decreases monotonically with the distance between the current forest stand structure diversity index of the sample plot and the structure target threshold, thereby generating a mutated strategy combination. After the same observation period ends, the forest stand structure diversity index compliance rate of each quadrat after adopting different strategy combinations is calculated. The compliance rate is used as fitness to iteratively update the adoption weight of each strategy combination in the quadrat population. When the adoption weight of any strategy combination is lower than the preset elimination threshold, the strategy combination is marked as an elimination strategy combination and will no longer participate in the enumeration of subsequent candidate processing rules. The strategy combination that survives the strategy propagation operation, the strategy mutation operation, and the weighted iterative update and has the highest weight is used as the regional recommended strategy set for the target forest area after the end of the current observation period. Each strategy combination in the regional recommended strategy set is associated with a spatial applicable range, which is determined by the spatial distribution boundary of the sample plots using that strategy combination in the sample plot spatial association network.

9. The method according to claim 1, characterized in that, Before performing the processing on the diameter at breast height (DBH) dataset, a feedforward compensation step for the structural target threshold is also included: Long-term stand succession monitoring data of the area where the target forest quadrat is located is obtained. The structural diversity index time series of multiple historical quadrats at different age stages is extracted from the long-term stand succession monitoring data. A growth curve is fitted to each structural diversity index time series to obtain the natural growth rate parameter and saturation capacity parameter of the structural diversity index of the historical quadrat. Using the current age of the target forest quadrat as a query index, retrieve the set of natural growth rate parameters and the set of saturation capacity parameters of the structural diversity index corresponding to the current age from the growth curves of all historical quadrats, and calculate the representative values ​​of the set of natural growth rate parameters and the set of saturation capacity parameters of the structural diversity index respectively. The representative value of the natural growth rate parameter of the structural diversity index is used as the succession driving input of the feedforward compensator, and the ratio of the representative value of the saturation capacity parameter of the structural diversity index to the structural target threshold is used as the saturation constraint coefficient of the feedforward compensator. The feedforward compensator calculates the feedforward compensation amount of the structural target threshold based on the succession driving input, the saturation constraint coefficient and the time span since the last update of the structural target threshold. The feedforward compensation is superimposed on the current structural target threshold to obtain the drifted structural target threshold. The original structural target threshold is replaced by the drifted structural target threshold. This is used in the step of comparing the forest stand structure diversity index with the structural target threshold and subsequent steps. After performing the aforementioned processing and recalculating the stand structure diversity index, the actual deviation between the observed stand structure diversity index and the drifted structure target threshold is calculated. The actual deviation is then fed back to the feedforward compensator. The feedforward compensator adjusts the calculation gain of the feedforward compensation amount according to the cumulative direction and magnitude of the actual deviation, so that the gain adapts to the accumulation of the actual deviation.

10. A forest aboveground carbon storage enhancement system based on stand structure diversity threshold regulation, the system being used to store and execute the method described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire the diameter at breast height (DBH) data of individual trees within the target forest quadrat, forming a DBH dataset. The diversity index calculation module is used to calculate the stand structure diversity index based on the diameter at breast height (DBH) values ​​of each individual tree in the DBH dataset. The threshold determination module is used to obtain the nonlinear relationship between the forest stand structure diversity index and the aboveground carbon storage of the quadrat, and to determine the structural target threshold T based on the nonlinear relationship; The comparison module is used to compare the forest stand structure diversity index with the structure target threshold T; The processing module is used to perform the following processing on the diameter at breast height (DBH) dataset when the stand structure diversity index is less than the structure target threshold T: according to a first preset screening ratio, individuals with the smallest DBH value and / or a competition pressure index greater than a preset threshold are screened out from the DBH dataset and deleted; according to a second preset screening ratio, some individuals are further screened out from the DBH dataset so that the number of individuals retained in the DBH dataset reaches a preset number of trees. Add at least one set of data for individuals of a preset tree species whose diameter at breast height (DBH) is less than a preset threshold to the DBH dataset; The recalculation module is used to recalculate the stand structure diversity index based on the processed diameter at breast height (DBH) dataset. The judgment module is used to determine whether the recalculated stand structure diversity index reaches or exceeds the structure target threshold T; The output module is used to record the first preset screening ratio, the second preset screening ratio, and / or the number of additional preset tree species and individuals corresponding to the recalculated forest stand structure diversity index reaching or exceeding the structure target threshold T, as the recommended processing rules output.

Images