Forest carbon sink dynamic closed-loop regulation and control method and system based on multi-source monitoring data

Abstract

The invention discloses a forest carbon sink dynamic closed-loop regulation and control method and system based on multi-source monitoring data, and relates to the technical field of forest ecological monitoring and carbon sink management, and the system comprises a data collection layer which is used for obtaining multi-source monitoring data including carbon flux data, plant physiological data, microclimate data, soil flux data and sample plot survey data; the intelligent fusion layer is used for processing and feature extraction of multi-source monitoring data and outputting a standardized carbon sink driving factor; the dynamic prediction layer predicts the carbon sink trend through an integrated model by using the standardized carbon sink driving factor; and the regulation and control decision-making layer is respectively connected with the dynamic prediction layer and the forest ecosystem, comprises a regulation and control rule base and a feedback optimization unit, and is used for generating a quantitative regulation and control strategy and optimizing a model and a rule. Therefore, by adopting the forest carbon sink dynamic closed-loop regulation and control method and system based on the multi-source monitoring data, the precision and efficiency of forest carbon sink management can be improved, and real-time and accurate closed-loop regulation and control are realized.

Description

Technical Field

[0001] This invention relates to the field of forest ecological monitoring and carbon sequestration management technology, and in particular to a method and system for dynamic closed-loop regulation of forest carbon sequestration based on multi-source monitoring data. Background Technology

[0002] Current forest carbon sequestration management technologies face three core problems that are difficult to overcome with existing technologies: (1) Severe data fragmentation: Multi-source monitoring data (such as carbon flux data at 30-minute resolution and sample plot survey data at quarterly resolution) have large differences in spatiotemporal scales (spanning from hours to quarters), and due to the lack of standardized monitoring equipment (such as differences in the accuracy of sensors from different manufacturers) and heterogeneous data formats (text, binary, images, etc.), there is a lack of standardized fusion mechanisms, resulting in a data utilization rate of less than 50%. For example, in traditional systems, carbon flux data and soil factor data are often stored independently due to mismatched time scales, making it impossible to analyze the carbon sink driving relationship in a collaborative manner.

[0003] (2) Limited prediction accuracy: Traditional methods often use a single model, resulting in insufficient model performance and large prediction errors. For example, pure LSTM or empirical formulas: Although LSTM can capture time series features, it is insufficient in modeling the nonlinear interaction relationship of environmental factors such as light and temperature; empirical formulas (such as regression equations based on sample land data) ignore the long-term time series correlation of carbon sink dynamics, resulting in prediction errors often exceeding 15%.

[0004] (3) Blindness in regulation: Management strategies rely on experience and lack quantitative threshold triggering mechanisms. For example, the intensity of thinning is mostly based on the experience value of "retaining 200-300 trees per hectare", without considering the coupling relationship between the growth stage of trees and soil moisture; the timing of water replenishment depends on manual inspection, with a lag of 30-90 days, resulting in low efficiency of carbon sequestration.

[0005] To address the problems existing in current technologies, there is an urgent need to build a closed-loop system that integrates multi-source data fusion, accurate prediction, and dynamic regulation to achieve real-time response and feedback optimization of regulation strategies. Summary of the Invention

[0006] The purpose of this invention is to provide a dynamic closed-loop regulation method and system for forest carbon sinks based on multi-source monitoring data, which overcomes the problems of low efficiency of multi-source data fusion, insufficient accuracy of carbon sink prediction, and lack of dynamic optimization of regulation strategies in the prior art, thereby improving the accuracy and efficiency of forest carbon sink management. It is applicable to evergreen broad-leaved forests, coniferous forests, and deciduous broad-leaved forests.

[0007] To achieve the above objectives, this invention provides a dynamic closed-loop control system for forest carbon sequestration based on multi-source monitoring data, comprising: The data acquisition layer is used to acquire multi-source monitoring data, including carbon flux data, plant physiological data, microclimate data, soil flux data, and sample plot survey data. The intelligent fusion layer, connected to the data acquisition layer, is used to perform spatiotemporal alignment, noise filtering, and feature extraction on multi-source monitoring data, and output a standardized carbon sink driving factor. The dynamic prediction layer, connected to the intelligent fusion layer, contains an integrated model that includes an improved LSTM sub-model and a random forest sub-model, used to predict carbon sink trends based on standardized carbon sink driving factors. The regulation decision layer, which connects the dynamic prediction layer and the forest ecosystem, includes a regulation rule base and a feedback optimization unit, used to generate quantitative regulation strategies and optimize models and rules based on post-regulation data.

[0008] Furthermore, the data acquisition layer includes: The carbon flux monitoring module is equipped with an eddy covariance system to collect net ecosystem exchange, gross primary productivity, and ecosystem respiration at a 30-minute resolution. The plant physiological monitoring module is equipped with a chlorophyll fluorometer and an online plant physiological and ecological monitoring instrument, which are used to collect plant physiological data including Fv / Fm, tree growth, tree runoff, canopy temperature and leaf humidity. The microclimate monitoring module is equipped with online monitoring equipment for forest microclimate, which is used to collect forest microclimate data including temperature, relative humidity, meteorological rainfall, wind speed, and photosynthetically active radiation. The soil flux monitoring module deploys soil respiration carbon flux monitoring equipment to collect data on soil respiration, soil temperature, and soil moisture content. The sample plot survey module is used to collect ecological survey data, including stand density and dominant species growth status, litter, and soil carbon pool.

[0009] Furthermore, the intelligent fusion layer includes: The spatiotemporal alignment unit uses a sliding window interpolation method to unify multi-source data to a 10m×10m grid / hour level; The calibration and quality control unit removes vibration noise from carbon flux data using wavelet transform and corrects temperature drift in microclimate data using Kalman filtering. The feature extraction unit generates standardized carbon sink driving factors, including the interaction factor of total primary productivity-temperature-photosynthetically active radiation and the coupling index of soil respiration-soil temperature-soil moisture content.

[0010] Furthermore, in the dynamic prediction layer, the improved LSTM sub-model introduces an attention mechanism to assign higher weights to the latest data and learn the temporal correlation of carbon sink dynamics; the random forest sub-model captures the nonlinear characteristics between environmental factors and carbon sink data by constructing a decision tree; and the improved LSTM sub-model and the random forest sub-model are fused according to dynamic weights to output the carbon sink trend.

[0011] Furthermore, the regulation rule base includes built-in quantitative rules related to carbon sinks, including photosynthetic efficiency thresholds, light energy utilization rules, and high temperature stress rules, and supports custom expansion.

[0012] This invention also provides a method for dynamic closed-loop regulation of forest carbon sinks based on multi-source monitoring data, employing a dynamic closed-loop regulation system for forest carbon sinks based on multi-source monitoring data, comprising the following steps: S1. Through the data acquisition layer, multi-source monitoring data is collected at a preset frequency, and based on 3 The principle is to remove outliers; S2. Based on the intelligent fusion layer, the sliding window interpolation method is used to unify the data to the 10m×10m grid / hour level. After wavelet transform and Kalman filtering to remove noise, the standardized carbon sink driving factor is extracted. The standardized carbon sink driving factors include the total primary productivity-temperature-photosynthetically active radiation interaction factor and the soil respiration-soil temperature-soil moisture content coupling index. S3. In the dynamic prediction layer, standardized carbon sink driving factors are used as input. An integrated model including an improved LSTM sub-model and a random forest sub-model is used to predict future carbon sink trends and identify trough periods and potential areas. S4. Based on the regulation rule base, generate a quantitative regulation strategy with GPS boundaries according to the prediction results; S5. Collect carbon sequestration data after regulation and compare it with the predicted carbon sequestration trend to optimize model parameters and regulation rule base, forming a closed-loop regulation.

[0013] Furthermore, in S4, the quantitative regulation strategy includes the thinning range determined by GPS positioning, the thinning intensity dynamically determined based on stand density and dominant species diameter at breast height, the timing of water replenishment based on the dual thresholds of humidity of trees and litter layer, and the fertilization threshold with clear type, amount and range constraints, and each indicator is linked to the synergistic needs of vegetation-litter-soil multi-carbon pool.

[0014] Furthermore, in S5, when the error between the actual effect and the expected effect after regulation is ≥5%, the feedback optimization unit updates the model weights and rule thresholds through gradient descent, with an update cycle of ≤7 days.

[0015] Therefore, the present invention employs the above-mentioned dynamic closed-loop regulation method and system for forest carbon sequestration based on multi-source monitoring data, and has the following technical effects: (1) This invention improves the utilization rate of multi-source data from ≤50% to 92.3% by using spatiotemporal alignment and noise filtering, thus solving the data fragmentation problem and significantly improving the utilization rate of multi-source data.

[0016] (2) The integrated model designed in this invention controls the prediction error (RMSE) to 7.4±0.4%, which is 51.3% lower than that of a single LSTM (error 15.2%), and can accurately identify the carbon sink trough period (such as predicting the GPP decline caused by high temperature one month in advance).

[0017] (3) The present invention also shortens the control response delay from 30-90 days to real time through a quantitative rule base with GPS boundaries, improves carbon sink management efficiency by 20-30%, achieves targeted improvement in control efficiency, improves water resource utilization efficiency, and reduces management costs, which has outstanding features.

[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a dynamic closed-loop regulation system for forest carbon sequestration based on multi-source monitoring data. Figure 2 This is a schematic diagram of a 10m×10m grid division in an embodiment of a dynamic closed-loop regulation method and system for forest carbon sinks based on multi-source monitoring data. Figure 3 This is a comparison chart of the prediction accuracy of different models in the embodiments of the dynamic closed-loop regulation method and system for forest carbon sinks based on multi-source monitoring data. Detailed Implementation

[0020] The present invention will be explained in more detail through the following embodiments. The purpose of disclosing the present invention is to protect all changes and modifications within the scope of the present invention. The present invention is not limited to the following embodiments.

[0021] In existing technologies, forest carbon sequestration monitoring systems only collect and store multi-source data, without involving dynamic prediction and closed-loop regulation. Furthermore, carbon sequestration prediction models only incorporate climatic factors, failing to integrate plant physiological data (such as chlorophyll fluorescence) and soil respiration data, and lacking feedback optimization mechanisms. This results in low efficiency of multi-source data fusion, insufficient accuracy in carbon sequestration prediction, and a lack of dynamic optimization of regulation strategies. Therefore, this invention provides a method and system for dynamic closed-loop regulation of forest carbon sequestration based on multi-source monitoring data. Through standardized data fusion, intelligent model integration prediction, and feedback-based regulation optimization, it improves the accuracy and efficiency of forest carbon sequestration management.

[0022] Example 1 like Figure 1As shown, this invention provides a dynamic closed-loop regulation system for forest carbon sequestration based on multi-source monitoring data, comprising the following four layers: (1) Data acquisition layer, which integrates 5 types of monitoring modules: The carbon flux monitoring module is equipped with a LI-7500DS eddy covariance system with a 30-minute resolution, which collects data on net ecosystem exchange (NEE), gross primary productivity (GPP), and ecosystem respiration (Re).

[0023] The plant physiological monitoring module includes the AS-SpecFluo chlorophyll fluorescence meter (measuring Fv / Fm every half hour) and the TH710 plant physiological and ecological online monitoring instrument (tree growth (diameter at breast height), tree runoff, canopy temperature, and leaf humidity).

[0024] The microclimate monitoring module is equipped with an AQM85 integrated online microclimate and ecology monitoring instrument, with a half-hour resolution, which collects data such as temperature (T), relative humidity (RH), rainfall (mm), wind speed (m / s), and photosynthetically active radiation (PAR).

[0025] The soil flux monitoring module is equipped with a LI-8100A soil carbon flux automatic measuring instrument, and includes a main unit (analyzer unit) and a measuring chamber. It has a half-hour resolution and collects data such as soil respiration (Rs) and soil temperature (Ts).

[0026] The sample plot survey module includes annual surveys of stand density, dominant species growth status, litter, and soil carbon pool.

[0027] (2) Intelligent fusion layer, which realizes data standardization through a three-level processing unit including a spatiotemporal alignment unit, a noise filtering unit and a feature extraction unit.

[0028] Spatiotemporal alignment unit: A sliding window interpolation method is used (the window size is dynamically adjusted according to the data type; for example, quarterly data is interpolated to the hourly level using a 2-hour window) to unify multi-source data to a 10m×10m grid / hour spatiotemporal scale. Figure 2 As shown.

[0029] Correction and quality control unit: First, high-frequency vibration noise in carbon flux data is removed by wavelet transform (db4 wavelet basis, decomposition level 3), and then temperature drift error in microclimate data is corrected by Kalman filtering.

[0030] Feature extraction unit: Based on the processed multi-source data, it generates standardized carbon sink driving factors, such as “GPP-T-PAR interaction factor” (formula: GPP=0.6×PAR+0.3×T-0.1×RH) and “Rs-Ts-SWC coupling index” (formula: Rs=0.5×Ts+0.3×SWC).

[0031] (3) Dynamic prediction layer: An integrated model including an improved LSTM sub-model and a random forest sub-model is used to predict future carbon sink trends.

[0032] Improved LSTM sub-model: Introduce an attention mechanism to assign higher weights to data from the past 72 hours, and learn the temporal correlations of carbon sink dynamics, such as the rhythmic differences between the growing season and the non-growing season.

[0033] Random Forest Sub-model: Construct 200 decision trees to capture the nonlinear characteristics of environmental factors, such as the sharp drop in GPP under high temperature stress.

[0034] Weighted fusion unit: The weights are dynamically adjusted according to the growth stage (e.g., during the growing season, the ratio of LSTM:Random Forest is improved to 6:4; during the non-growing season, the ratio of LSTM:Random Forest is 5:5) to output the carbon sink trend for 1-6 months, such as changes in GPP and carbon storage.

[0035] (4) The regulatory decision-making layer includes a rule base and a feedback optimization unit, which executes a closed-loop regulatory strategy.

[0036] Regulation rule library: Built-in quantitative rules such as photosynthetic efficiency threshold (e.g., triggering fertilization when Fv / Fm<0.7), light energy utilization rules (e.g., prioritizing the retention of tall canopy trees when PAR<500μmol / m² / s), and high temperature stress rules (e.g., starting irrigation when Ts>28℃ and SWC<15%).

[0037] Feedback optimization unit: Based on monitoring data after regulation (such as changes in GPP after thinning), the model weights (such as the LSTM learning rate) and rule thresholds (such as adjusting the SWC trigger threshold to 16%) are iteratively optimized using the gradient descent method to form a closed-loop regulation.

[0038] Example 2 This invention provides a dynamic closed-loop regulation system for forest carbon sequestration based on multi-source monitoring data. The specific steps are as follows: S1. Multi-source data acquisition and preprocessing: Data is acquired at a preset frequency (30 minutes / 1 hour / quarter), and outliers (such as carbon flux data that exceed the mean ± 3 times the standard deviation) are removed using the 3σ principle, retaining valid samples.

[0039] S2. Data Fusion and Feature Extraction: Data is unified to a 10m×10m grid / hour level using sliding window interpolation; noise is removed by wavelet transform and Kalman filtering; standardized driving factors such as GPP-T-PAR interaction factor are extracted.

[0040] S3. Carbon Sequestration Dynamic Prediction: Standardized driving factors are input into the integrated model (improved LSTM + random forest) to predict carbon sequestration trends over 1-6 months, identifying trough periods (such as GPP decline during high-temperature periods) and potential areas (such as carbon sequestration improvement potential during the growing season).

[0041] S4. Quantitative regulation strategy generation: Based on the prediction results, call the rule base to generate management strategies with GPS boundaries, such as "For trees with a thinning density of ≥200 trees / hm² in grid number A01-A05, retain dominant species with a diameter at breast height of ≥15cm".

[0042] In other embodiments, the quantitative regulation strategy also includes the thinning range located by GPS, the thinning intensity dynamically determined based on stand density and dominant species diameter at breast height, the timing of water replenishment based on dual thresholds of humidity for associated trees and litter layer, and fertilization thresholds with clear type, amount and range constraints, and each indicator is connected to the synergistic needs of multiple carbon pools of "vegetation-litter-soil".

[0043] S5. Feedback optimization and closed-loop iteration: Collect data after regulation (such as GPP 2 weeks after thinning), compare with expected results (optimization is triggered when the error is ≥5%), update model parameters and rule base, and form a closed loop of "collection-fusion-prediction-regulation-feedback".

[0044] Example 3 The method and system for dynamic closed-loop regulation of forest carbon sinks based on multi-source monitoring data provided by this invention were applied to the evergreen broad-leaved forest of Guifeng Mountain, collecting data from 2023 to 2025.

[0045] (1) System deployment and data acquisition: Carbon flux monitoring: The LI-7500DS eddy covariance system was deployed in the core area at an altitude of 300m, recording NEE (net ecosystem exchange), GPP (gross primary productivity), and Re (ecosystem respiration) every 30 minutes.

[0046] Plant physiological monitoring: Fv / Fm (maximum quantum yield of photosystem II) was measured hourly using an AS-SpecFluo chlorophyll fluorometer, and diameter at breast height (DBH) was measured monthly using a growth cone (accuracy ±0.1 cm).

[0047] Microclimate monitoring: An AQM85 integrated online microclimate and ecology monitoring instrument is installed to collect data such as temperature (T), relative humidity (RH), rainfall (mm), wind speed (m / s), and photosynthetically active radiation (PAR), recording data every half hour.

[0048] Soil flux monitoring: The system is equipped with an LI-8100A soil carbon flux automatic measuring instrument, consisting of a main unit (analyzer unit) and a measuring chamber. It has a half-hour resolution and collects data such as soil respiration (Rs) and soil temperature (Ts).

[0049] Plot survey: 9 quadrats of 20m×20m were used to investigate the stand density of dominant species such as Schima superba, Castanopsis fargesii, Litsea cubeba, and Acacia confusa (initially 300 trees / hm²).

[0050] (2) Intelligent fusion layer processing: Spatiotemporal alignment: The quarterly stand density data were interpolated to the hour level using the 2-hour sliding window interpolation method and spatially mapped to a 10m×10m grid (GPS coordinates: N22°30′-22°35′, E113°50′-113°55′).

[0051] Noise filtering: db4 wavelet basis decomposition in 3 layers removes high-frequency noise from carbon flux, and Kalman filtering corrects temperature drift in microclimate data (error rate reduced from 15% to 4.2%).

[0052] Feature extraction: Generate GPP=0.6×PAR+0.3×T-0.1×RH, Rs=0.5×Ts+0.3×SWC.

[0053] Through the above process, the rate of collaborative analysis of plot survey data and carbon flux data was increased by 80%.

[0054] (3) Training and output of the dynamic prediction layer: Training data: 2023-2024 data (70% training set, 30% validation set), loss function is MSE, optimizer is Adam.

[0055] Model parameters: LSTM (learning rate 0.001, hidden layers 64), random forest (200 decision trees, maximum depth 15).

[0056] 2025 forecast: GPP will decrease by 12-15% during the high-temperature period in July-August (low point), and increase by 20% during the growing season in March-April (potential area).

[0057] To verify the effectiveness of the ensemble model, comparative experiments were also conducted, such as... Figure 3 As shown.

[0058] (4) Regulation decision-making and feedback: Control strategy: In July, thin out grids A01-A05 (stand density ≥400 trees / hm²), retaining Castanopsis chinensis with a diameter at breast height ≥15cm.

[0059] Results: After thinning, GPP increased by 11.7±3.5% (t=4.32, p<0.01), and carbon sequestration increased by 3.2 tCO2e / hm², which is 91.8% higher than the traditional threshold method (6.1±2.3%).

[0060] Optimization: Updated LSTM weights to 0.65 and added a rule to "start irrigation when Ts>28℃ and SWC<15%".

[0061] Example 4 The method and system for dynamic closed-loop regulation of forest carbon sinks based on multi-source monitoring data provided by this invention were applied to the Northeast Korean pine forest, collecting data from 2023 to 2024.

[0062] System adaptation: Adjust the microclimate monitoring module (add snow depth sensor), and add "reduce Rs threshold by 10% during snow cover period (December-February)" to the rule base; Results: Prediction error was 7.2%, carbon sink increased by 25% after adjustment, and water resource utilization efficiency reached 2.6 gC / mm, verifying the universality of the method of this invention.

[0063] Example 5 The forest carbon sink dynamic closed-loop regulation method based on multi-source monitoring data provided by this invention was also compared with the traditional threshold method, as shown in Table 1.

[0064] Table 1 Comparative Analysis Results ;

[0065] In summary, the method proposed in this application is applicable to various forest types, including evergreen broad-leaved forests and coniferous forests (in the pilot project of Northeast Korean pine forest, the carbon sink efficiency was improved by 25%); the carbon sink increment reached 3.2 tCO2e / hm², which, calculated at a carbon trading price of 6.15 / t, resulted in an annual increase in income of 18.6 per hectare. At the same time, water resource utilization efficiency was improved from 1.2 gC / mm to 2.8 gC / mm, reducing management costs.

[0066] Therefore, the present invention adopts the above-mentioned dynamic closed-loop regulation method and system for forest carbon sinks based on multi-source monitoring data, and improves the accuracy and efficiency of forest carbon sink management through standardized data fusion, intelligent model integration prediction and feedback regulation optimization.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A forest carbon sequestration dynamic closed-loop control system based on multi-source monitoring data, characterized in that, include: The data acquisition layer is used to acquire multi-source monitoring data, including carbon flux data, plant physiological data, forest microclimate data, soil flux data, and sample plot survey data. The intelligent fusion layer, connected to the data acquisition layer, is used to perform spatiotemporal alignment, noise filtering, and feature extraction on multi-source monitoring data, and output a standardized carbon sink driving factor. The dynamic prediction layer, connected to the intelligent fusion layer, contains an integrated model that includes an improved LSTM sub-model and a random forest sub-model, used to predict carbon sink trends based on standardized carbon sink driving factors. The improved LSTM sub-model introduces an attention mechanism, assigning higher weights to the most recent data to learn the temporal correlation of carbon sink dynamics; the random forest sub-model captures the nonlinear characteristics between environmental factors and carbon sink data by constructing a decision tree; and the improved LSTM sub-model and random forest sub-model are fused according to dynamic weights to output the carbon sink trend. The regulation decision layer, which connects the dynamic prediction layer and the forest ecosystem, includes a regulation rule base and a feedback optimization unit, used to generate quantitative regulation strategies and optimize models and rules based on post-regulation data.

2. The forest carbon sequestration dynamic closed-loop control system based on multi-source monitoring data according to claim 1, characterized in that, The data acquisition layer includes: The carbon flux monitoring module deploys an eddy covariance system to collect data on net ecosystem exchange, gross primary productivity, and ecosystem respiration. The plant physiological monitoring module is equipped with a chlorophyll fluorometer and an online plant physiological and ecological monitoring instrument, which are used to collect plant physiological data including Fv / Fm, tree growth, tree runoff, canopy temperature and leaf humidity. The microclimate monitoring module is equipped with online monitoring equipment for the forest microclimate, which is used to collect forest microclimate data including temperature, relative humidity, meteorological rainfall, and photosynthetically active radiation. The soil flux monitoring module is equipped with soil respiration carbon flux monitoring equipment to collect data on soil respiration, soil temperature, and soil moisture content. The sample plot survey module is used to collect ecological survey data, including stand density and dominant species growth status, litter, and soil carbon pool.

3. The forest carbon sequestration dynamic closed-loop control system based on multi-source monitoring data according to claim 1, characterized in that, The intelligent fusion layer includes: The spatiotemporal alignment unit uses a sliding window interpolation method to unify multi-source data to a 10m×10m grid / hour level; The calibration and quality control unit removes vibration noise from carbon flux data using wavelet transform and corrects temperature drift in microclimate data using Kalman filtering. The feature extraction unit generates the interaction factor between total primary productivity, temperature, and photosynthetically active radiation and the coupling index between soil respiration, soil temperature, and soil moisture content.

4. The forest carbon sequestration dynamic closed-loop control system based on multi-source monitoring data according to claim 1, characterized in that, The regulation rule base contains built-in quantitative rules related to carbon sinks, including photosynthetic efficiency thresholds, light energy utilization rules, and high temperature stress rules, and supports custom extensions.

5. A method for dynamic closed-loop regulation of forest carbon sinks based on multi-source monitoring data, employing the dynamic closed-loop regulation system for forest carbon sinks based on multi-source monitoring data as described in any one of 1-4, characterized in that, Includes the following steps: S1. Through the data acquisition layer, multi-source monitoring data is collected at a preset frequency, and based on... The principle is to remove outliers; S2. Based on the intelligent fusion layer, the sliding window interpolation method is used to unify the data to the 10m×10m grid / hour level. After wavelet transform and Kalman filtering to remove noise, the standardized carbon sink driving factor is extracted. The standardized carbon sink driving factors include the total primary productivity-temperature-photosynthetically active radiation interaction factor and the soil respiration-soil temperature-soil moisture content coupling index. S3. In the dynamic prediction layer, standardized carbon sink driving factors are used as input. An integrated model including an improved LSTM sub-model and a random forest sub-model is used to predict future carbon sink trends and identify trough periods and potential areas. S4. Based on the regulation rule base, generate a quantitative regulation strategy with GPS boundaries according to the prediction results; S5. Collect carbon sequestration data after regulation and compare it with the predicted carbon sequestration trend to optimize model parameters and regulation rule base, forming a closed-loop regulation.

6. The method for dynamic closed-loop regulation of forest carbon sequestration based on multi-source monitoring data according to claim 5, characterized in that, In S4, the quantitative regulation strategy includes the thinning range located by GPS, the thinning intensity dynamically determined based on stand density and dominant species diameter at breast height, the timing of water replenishment based on the dual thresholds of humidity of trees and litter layer, and the fertilization threshold with clear type, amount and range constraints. Moreover, each indicator is linked to the synergistic needs of vegetation-litter-soil multi-carbon pool.

7. The method for dynamic closed-loop regulation of forest carbon sinks based on multi-source monitoring data according to claim 5, characterized in that, In S5, when the error between the actual effect and the expected effect after regulation is ≥5%, the feedback optimization unit updates the model weights and rule thresholds through gradient descent, with an update cycle of ≤7 days.

Images