A method for classifying forest ecological quality levels based on statistical ecological dual-constraint thresholds

Abstract

This invention relates to the field of forest ecological quality assessment technology, specifically to a method for classifying forest ecological quality levels based on statistical ecological dual-constraint thresholds. The invention obtains the prediction residual by subtracting the quality index predicted by a dual-task model from the quality index assessed based on actual observation data; it uses a SHAP interpreter to determine the classification SHAP value and regression SHAP value of each ecological feature predicted by the dual-task model; it integrates the analysis results of the SHAP interpreter to find key features for each dynamic trend level; and for each dynamic trend level, it constructs a classification standard for each dynamic trend level based on the feature values ​​of the key features and the prediction residual. This classification standard has both statistical and ecological significance, thus forming an interpretable and reusable hierarchical rule system.

Description

Technical Field

[0001] This invention relates to the field of forest ecological quality technology, specifically to a method for classifying forest ecological quality levels based on statistical ecological dual-constraint thresholds. Background Technology

[0002] The ecological quality of forest ecosystems is directly related to regional ecological security, biodiversity conservation, and carbon sequestration. Therefore, scientifically and accurately assessing forest ecological quality is of great significance for guiding forest protection, restoration, and sustainable management.

[0003] Currently, traditional methods for classifying forest ecological quality employ single-indicator evaluation or comprehensive index methods, which are highly subjective, consider limited information, and are prone to deviating from reality. Existing technologies include methods for training models using historical data to predict ecological quality; however, these predictions are significantly influenced by the historical data being learned. When the evolution of forest ecosystems is dominated by human or natural factors, the model predictions may differ from the actual situation, directly affecting the accuracy of forest ecological quality classification.

[0004] Therefore, there is an urgent need for a technical solution that can accurately classify the quality level of the actual evolution of forest ecosystems. Summary of the Invention

[0005] The purpose of this invention is to provide a method for classifying forest ecological quality levels based on statistical ecological dual-constraint thresholds, so as to solve the technical problems mentioned in the background art.

[0006] This invention provides a method for classifying forest ecological quality levels based on statistical ecological dual-constraint thresholds, which includes: Obtain the ecological characteristics of the forest ecosystems for which forest ecological quality classification is required, and construct a feature matrix based on the ecological characteristics and the pre-obtained environmental characteristics; The feature matrix is ​​input into a pre-trained dual-task model. The classification task of the dual-task model outputs the probability distribution of dynamic trend levels, and the regression task outputs a quality index. The prediction residual is obtained by subtracting the quality index obtained from the evaluation based on real observation data from the quality index predicted by the dual-task model. The SHAP interpreter is used to determine the classification SHAP value and regression SHAP value of each ecological feature in the feature matrix during the prediction of the dual-task model; the classification SHAP value represents the degree of contribution of the ecological feature to each dynamic trend level, and the regression SHAP value represents the magnitude of the influence of the ecological feature on the quality index. The analysis results of the SHAP interpreter are integrated to find key features for each dynamic trend level. For each dynamic trend level, the ecological features with the highest categorical SHAP value and the highest regression SHAP value are identified as the corresponding key features. A classification standard is constructed for each dynamic trend level based on the eigenvalues ​​of key features and the predicted residuals; the actual dynamic trend level of the forest ecosystem is determined based on the classification standard satisfied by the forest ecosystem.

[0007] As an optional embodiment, the dual-task model includes the XGBoost dual-task model.

[0008] As an optional embodiment, the training method of the dual-task model includes: Using the feature matrix as input data, the dynamic trend level as the true label for the classification task, and the quality index as the true label for the regression task, training samples are constructed. A training set is constructed based on the training samples, and the dual-task model is trained using the training set to obtain a trained dual-task model.

[0009] As an optional embodiment, the prediction residual is obtained by subtracting the quality index estimated based on real-world observation data from the quality index predicted by the dual-task model, and is expressed as: in, Indicates the predicted residual. This represents a quality index obtained based on real-world observation data. This represents the quality index of the predictions made by the dual-task model.

[0010] As an optional embodiment, the dynamic trend levels include: significant improvement, slight improvement, relatively stable, slight degradation, and significant degradation; Based on the eigenvalues ​​of key features and the prediction residuals, a classification standard is constructed for each dynamic trend level, including: The criteria for defining significant improvement are as follows: >0、 > Furthermore, the increase in biomass density relative to the biomass density baseline value is greater than twice the standard deviation of the biomass density baseline. Indicates the baseline standard deviation of the residuals; The criteria for classifying the minor improvements are as follows: >0、 ≤ ≤ Furthermore, the increase in net primary productivity relative to the net primary productivity benchmark is greater than the standard deviation of the net primary productivity benchmark; The relatively stable classification criterion is: < Furthermore, the average absolute value of the rate of change of ecological indicators is less than the preset average threshold. The criteria for classifying slight degradation are as follows: <0、 ≤ ≤ Furthermore, the core area fragmentation index is greater than the preset fragmentation index threshold; The criteria for classifying significant degradation are as follows: <0、 > Furthermore, the decrease in biomass density relative to the biomass density benchmark value is greater than twice the standard deviation of the biomass density benchmark.

[0011] As an optional embodiment, it also includes a dominant factor for determining the dynamic trend level based on the predicted residuals, if < If so, then the dominant factor is determined to be natural fluctuation. > If so, then the dominant factor is determined to be human-driven.

[0012] As an optional embodiment, after determining the actual dynamic trend level and dominant factors of the forest ecosystem, the method further includes: An analysis report on quality level classification is generated by utilizing the actual dynamic trend level of the forest ecosystem, the quality index predicted by the dual-task model, the dominant factors of the dynamic trend level, and the characteristic values, classification SHAP values, and regression SHAP values ​​of the corresponding key features of the dynamic trend level.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This invention first acquires the ecological characteristics of the forest ecosystem for which forest ecological quality grading needs to be performed. Based on these ecological characteristics and pre-acquired environmental characteristics, a feature matrix is ​​constructed. This feature matrix is ​​then input into a dual-task model pre-trained using historical data. The dual-task model's classification task outputs the probability distribution of dynamic trend levels, while the regression task outputs a quality index. This invention initially uses the model trained on historical data to make preliminary predictions of forest ecological quality (dynamic trend levels and quality index). However, considering the potential bias in model predictions, this invention further subtracts the quality index predicted by the dual-task model from the quality index obtained based on actual observation data to obtain the prediction residual. A SHAP interpreter is used to determine the classification SHAP value and regression SHAP value of each ecological characteristic in the dual-task model's predicted feature matrix. The classification SHAP value characterizes the contribution of the ecological characteristic to each dynamic trend level, reflecting statistical significance, while the regression SHAP value characterizes the magnitude of the ecological characteristic's influence on the quality index, representing ecological criticality. Subsequently, the analysis results of the SHAP interpreter are integrated to find key features for each dynamic trend level. For each dynamic trend level, ecological features with a pre-preset proportion of classification SHAP values ​​and regression SHAP values ​​greater than a pre-preset threshold are identified as the corresponding key features. Based on the feature values ​​of the key features and the prediction residuals, a classification standard is constructed for each dynamic trend level. The classification standard has both statistical and ecological significance, thereby forming an interpretable and reusable hierarchical rule system for dynamic trend levels to guide forest ecological maintenance strategies. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating a method for classifying forest ecological quality levels based on statistical ecological dual-constraint thresholds, according to an embodiment of the present invention. Detailed Implementation

[0015] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0016] Combination Figure 1 This embodiment provides a method for classifying forest ecological quality levels based on statistical ecological dual-constraint thresholds, which includes: Step S1: Obtain the ecological characteristics of the forest ecosystem for which forest ecological quality grading is required, and construct a feature matrix based on the ecological characteristics and pre-obtained environmental characteristics. The ecological characteristics here can be structural, vitality, functional, and stability features. As an optional example, the structural characteristics include quantity structure indices (absolute values ​​of the rate of change of ecological indicators such as vegetation cover, biomass density, and vegetation cover or biomass density), origin structure indices, community structure indices (core area fragmentation index), and age structure indices; the vitality characteristics include productivity indices (net primary productivity, NPP); the functional characteristics include carbon sequestration indices (carbon sequestration rate, NEP), water conservation indices (water conservation capacity), soil conservation indices (conserved soil quantity), and the biodiversity index (SHDI); the stability characteristics include pest and disease severity levels; and as an example, the environmental characteristics include climate conditions and site factors.

[0017] Step S2: Input the feature matrix into the dual-task model pre-trained using historical data. The classification task of the dual-task model outputs the probability distribution of dynamic trend levels, and the regression task outputs a quality index of 0-100 points. The training method for the dual-task model includes: using the feature matrix as input data, using the dynamic trend level as the true label for the classification task, and using the quality index as the true label for the regression task to construct training samples; using the training samples obtained from historical data to construct a training set; and using the training set to train the dual-task model to obtain a trained dual-task model.

[0018] Specifically, the dual-task model in this embodiment is the XGBoost dual-task model. This embodiment utilizes the multi-task learning capability of the XGBoost framework to simultaneously construct a classifier (objective='multi:softprob') and a regressor (objective='reg:squarederror'). The two tasks share the same feature input and underlying tree structure (base learner). By establishing mapping relationships between the classification task and the regression task in the output layer, the joint optimization of feature representation and information complementarity between tasks are achieved.

[0019] Specifically, the classification task outputs a probability distribution of the forest ecosystem corresponding to the current feature input, indicating whether it belongs to one of five dynamic trend levels (significant improvement, slight improvement, relatively stable, slight degradation, and significant degradation). The dynamic trend level corresponding to the highest probability value is used as the dynamic trend level predicted by the model. For example, the probability vector [0.02, 0.08, 0.15, 0.60, 0.15] corresponds to "slight degradation." The regression task outputs a quality index, such as 76.3 points, as a quantitative expression of the ecosystem's state.

[0020] Step S3: Subtract the forest ecosystem quality index calculated based on real-world observation data from the quality index predicted by the dual-task model under the corresponding environmental background based on the input feature matrix (including ecological and environmental features) to obtain the prediction residual. This residual characterizes the deviation between the actual state and the theoretical state simulated by the model. The expression for the prediction residual is: in, Indicates the predicted residual. Indicating the quality index of actual observations, This represents the quality index of the predictions made by the dual-task model.

[0021] Step S4: Use the SHAP interpreter (SHapleyAdditive exPlanations) to determine the classification SHAP value and regression SHAP value of each ecological feature in the feature matrix during the dual-task model prediction; the classification SHAP value represents the degree of contribution of the ecological feature to each dynamic trend level, and the regression SHAP value represents the magnitude of the influence of the ecological feature on the quality index. This embodiment employs a SHAP interpreter-driven two-dimensional rule parsing method to calculate the classification SHAP value and regression SHAP value of each ecological feature in the dual-task model prediction. The classification SHAP value can quantify the contribution of each ecological feature to the dynamic trend level (such as "significant improvement") and reflect statistical significance. The regression SHAP value reflects the magnitude of the impact of the unit change in the feature value of the ecological feature on the quality index (FQI). For example, for every 1 unit increase in biomass density, the FQI increases by 2.5 points, which can effectively characterize the ecological criticality.

[0022] Step S5: Integrate the analysis results of the SHAP interpreter to find key features for each dynamic trend level. For each dynamic trend level, ecological features with a top-preset proportion of categorical SHAP values ​​and a regression SHAP value greater than a preset threshold are identified as the corresponding key features. This embodiment finds key features by integrating SHAP analysis results and sets thresholds for key features that have both statistical and ecological significance. For example, taking "significant improvement" as an example, for this dynamic trend level, the categorical SHAP value of biomass density ranks in the top 10% of all ecological features, and the regression SHAP value is greater than or equal to 2.0. Therefore, setting a corresponding threshold for biomass density can serve as one of the criteria for classifying "significant improvement," thereby forming an interpretable and reusable hierarchical rule system.

[0023] Step S6: Construct a classification standard for each dynamic trend level based on the eigenvalues ​​of the key features and the predicted residuals; determine the actual dynamic trend level of the forest ecosystem based on the classification standard satisfied by the forest ecosystem.

[0024] In one specific embodiment, the criteria for classifying each dynamic trend level are as follows, and each classification criterion must be statistically significant.

[0025] The criteria for defining significant improvement are as follows: >0、 > Furthermore, the increase in biomass density relative to the biomass density baseline value is greater than twice the standard deviation of the biomass density baseline. Indicates the baseline standard deviation of the residuals; The criteria for classifying the minor improvements are as follows: >0、 ≤ ≤ Furthermore, the increase in net primary productivity relative to the net primary productivity benchmark is greater than the standard deviation of the net primary productivity benchmark; The relatively stable classification criterion is: < Furthermore, the average absolute value of the rate of change of ecological indicators is less than the preset average threshold (for example, it can be set to 5%). The criteria for classifying slight degradation are as follows: <0、 ≤ ≤ Furthermore, the core area fragmentation index is greater than the preset fragmentation index threshold (for example, it can be set to 0.3). The criteria for classifying significant degradation are as follows: <0、 > Furthermore, the decrease in biomass density relative to the biomass density benchmark value is greater than twice the standard deviation of the biomass density benchmark.

[0026] The residual baseline standard deviation, biomass density baseline value, biomass density baseline standard deviation, net primary productivity baseline value, and net primary productivity baseline standard deviation are all calculated using historical statistical data. Furthermore, if the forest ecosystem does not meet any of the classification criteria, the dynamic trend level predicted by the model is directly used as the actual dynamic trend level of the forest ecosystem.

[0027] This embodiment also determines the dominant factors for the dynamic trend level based on the predicted residuals. < If so, then the dominant factor is determined to be natural fluctuation. > If this is the case, then the dominant factor is determined to be human-driven. Furthermore, after determining the actual dynamic trend level and dominant factor of the forest ecosystem, an analysis report on quality level classification is generated using the actual dynamic trend level of the forest ecosystem, the quality index predicted by the dual-task model, the dominant factor of the dynamic trend level, and the characteristic values, classification SHAP values, and regression SHAP values ​​of the corresponding key features of that dynamic trend level.

[0028] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0029] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0030] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0031] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0032] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for classifying forest ecological quality levels based on statistical ecological dual-constraint thresholds, characterized in that, include: Obtain the ecological characteristics of the forest ecosystems for which forest ecological quality classification is required, and construct a feature matrix based on the ecological characteristics and the pre-obtained environmental characteristics; The feature matrix is ​​input into a pre-trained dual-task model. The classification task of the dual-task model outputs the probability distribution of dynamic trend levels, and the regression task outputs a quality index. The prediction residual is obtained by subtracting the quality index obtained from the evaluation based on real observation data from the quality index predicted by the dual-task model. The SHAP interpreter is used to determine the classification SHAP value and regression SHAP value of each ecological feature in the feature matrix during the prediction of the dual-task model; the classification SHAP value represents the degree of contribution of the ecological feature to each dynamic trend level, and the regression SHAP value represents the magnitude of the influence of the ecological feature on the quality index. Integrate the analysis results from the SHAP interpreter to find key features for each dynamic trend level; For each dynamic trend level, ecological features with a pre-defined proportion of classification SHAP values ​​and a regression SHAP value greater than a pre-defined threshold are identified as the corresponding key features. A classification standard is constructed for each dynamic trend level based on the eigenvalues ​​of key features and the predicted residuals; the actual dynamic trend level of the forest ecosystem is determined based on the classification standard satisfied by the forest ecosystem.

2. The forest ecological quality classification method based on statistical ecological dual-constraint thresholds according to claim 1, characterized in that, The dual-task model includes the XGBoost dual-task model.

3. The method for classifying forest ecological quality levels based on statistical ecological dual-constraint thresholds according to claim 1, characterized in that, The training method for the dual-task model includes: Using the feature matrix as input data, the dynamic trend level as the true label for the classification task, and the quality index as the true label for the regression task, training samples are constructed. A training set is constructed based on the training samples, and the dual-task model is trained using the training set to obtain a trained dual-task model.

4. The forest ecological quality classification method based on statistical ecological dual-constraint thresholds according to claim 1, characterized in that, The prediction residual is obtained by subtracting the quality index obtained from the actual observation data from the quality index predicted by the dual-task model, and is expressed as: in, Indicates the predicted residual. This represents a quality index obtained based on real-world observation data. This represents the quality index of the predictions made by the dual-task model.

5. The forest ecological quality classification method based on statistical ecological dual-constraint thresholds according to claim 4, characterized in that, The dynamic trend levels include: significant improvement, slight improvement, relatively stable, slight degradation, and significant degradation; Based on the eigenvalues ​​of key features and the prediction residuals, a classification standard is constructed for each dynamic trend level, including: The criteria for defining significant improvement are as follows: >0、 > Furthermore, the increase in biomass density relative to the biomass density baseline value is greater than twice the standard deviation of the biomass density baseline. Indicates the baseline standard deviation of the residuals; The criteria for classifying the minor improvements are as follows: >0、 ≤ ≤ Furthermore, the increase in net primary productivity relative to the net primary productivity benchmark is greater than the standard deviation of the net primary productivity benchmark; The relatively stable classification criterion is: < Furthermore, the average absolute value of the rate of change of ecological indicators is less than the preset average threshold. The criteria for classifying slight degradation are as follows: <0、 ≤ ≤ Furthermore, the core area fragmentation index is greater than the preset fragmentation index threshold; The criteria for classifying significant degradation are as follows: <0、 > Furthermore, the decrease in biomass density relative to the biomass density benchmark value is greater than twice the standard deviation of the biomass density benchmark.

6. The forest ecological quality classification method based on statistical ecological dual-constraint thresholds according to claim 5, characterized in that, It also includes the dominant factors for determining the dynamic trend level based on the predicted residuals, if < If so, then the dominant factor is determined to be natural fluctuation. > If so, then the dominant factor is determined to be human-driven.

7. The forest ecological quality classification method based on statistical ecological dual-constraint thresholds according to claim 6, characterized in that, After determining the actual dynamic trend levels and dominant factors of the forest ecosystem, the method further includes: An analysis report on quality level classification is generated by utilizing the actual dynamic trend level of the forest ecosystem, the quality index predicted by the dual-task model, the dominant factors of the dynamic trend level, and the characteristic values, classification SHAP values, and regression SHAP values ​​of the corresponding key features of the dynamic trend level.

Images