Mangrove forest sediment carbon sink threshold value metering method

Abstract

The invention discloses a mangrove forest sediment carbon sequestration threshold value metering method, and belongs to the field of ecological environment monitoring and carbon sequestration capacity evaluation. Aiming at a multi-type mangrove forest ecological system, an environment-microorganism-carbon sink multi-dimensional data set is established by means of fusing remote sensing and unmanned aerial vehicle surveying and mapping, field environment sample collection, multi-modal feature fusion, microbiological omics analysis, climate simulation and the like. A self-adaptive hierarchical attention mechanism and an evolutionary threshold detection network are adopted to dynamically identify carbon sink efficiency characteristic inflection points under different habitats, and a carbon sink quantity dynamic measurement and risk prediction model is constructed through deep causal inference and high-dimensional discriminant clustering. According to the method, real-time sensing and prospective forecasting of the mangrove forest carbon sequestration capacity under different habitats are achieved, scientificity, accuracy and intelligent management level of carbon sequestration quantity metering are improved, and a typing carbon sequestration threshold value evaluation standard can be formulated according to the scientificity, accuracy and intelligent management level of carbon sequestration quantity metering.

Description

Technical Field

[0001] This invention belongs to the field of ecological environment monitoring and carbon sequestration capacity assessment, and more specifically relates to a method for measuring the carbon sequestration threshold of mangrove sediments. Background Technology

[0002] Blue carbon ecosystems such as mangroves, as important natural resources in coastal areas, play a crucial role not only in maintaining biodiversity, purifying water quality, preventing wind erosion and sandstorms, and safeguarding ecological security, but also serve as a vital ecological barrier for addressing global climate change due to their exceptional carbon sequestration capacity. Mangroves fix atmospheric carbon dioxide in vegetation and soil through photosynthesis, effectively mitigating the climate risks posed by rising greenhouse gas concentrations. According to authoritative international research, the carbon sequestration capacity per unit area of ​​blue carbon ecosystems such as mangroves is far higher than that of terrestrial forests, making them veritable carbon sinks. However, affected by multiple factors such as climate change, sea-level rise, human disturbance, and extreme weather, the carbon sequestration function of mangroves faces severe challenges. Maintaining the stability of their carbon pool and the long-term preservation of their carbon sequestration effect has become a research hotspot in the global field of ecology and carbon management.

[0003] A scientific, intelligent, and efficient carbon sequestration capacity assessment system is the foundation for realizing the assetization, marketization, and high-quality management decisions of blue carbon. Existing blue carbon measurement and monitoring technologies mainly include ground-based plot surveys, remote sensing monitoring, ecological model estimation, and biogeochemical analysis. While these methods support carbon sequestration assessment and certification to some extent, they face bottlenecks when dealing with ecosystems such as mangroves, which are characterized by multi-habitat patchy structures, strong multi-factor interactions, and significant spatiotemporal variability. First, traditional plot surveys and single-point sampling methods are easily affected by sample representativeness, seasonal variations, and human interference, resulting in limited data stability and accuracy. This makes it difficult to reveal the multi-level dynamic changes in carbon sequestration capacity from the microscale (microhabitat, microbiome, etc.) to the macroscale (region, watershed).

[0004] Furthermore, the impacts of typhoons, floods, droughts, and human disturbances on mangrove carbon sequestration capacity exhibit highly nonlinear characteristics. Problems such as identifying carbon sequestration thresholds and their abrupt change points, forecasting carbon sequestration loss risks, and assessing vulnerability are becoming increasingly prominent. Existing methods mainly rely on empirical thresholds or static models, lacking real-time, dynamic, and intelligent risk perception and response capabilities. Especially in application scenarios such as carbon trading and standard certification, errors in carbon sequestration capacity assessment and delayed responses can easily lead to a series of management and economic problems, such as unfair ecological compensation and inaccurate carbon asset assessments. This places more stringent demands on the high precision, intelligence, and standardization of carbon sequestration measurement systems.

[0005] Currently, intelligent carbon sequestration measurement systems for blue carbon ecosystems such as mangroves are still in their early stages both domestically and internationally. There is a lack of integrated models and related standard systems for collaborative identification and dynamic tracking across multiple habitats, factors, and scales. Some studies have explored carbon sequestration capacity prediction based on remote sensing, machine learning, and microbiome data, but these have shortcomings in model generalization ability, accurate threshold discrimination, dynamic risk response, and continuous standard optimization. Typical technical challenges include: 1) accurate identification of carbon sequestration thresholds and attribution classification of heterogeneous habitats; 2) efficient fusion of multi-source heterogeneous data and causal mechanism analysis; 3) dynamic adaptive adjustment of carbon sequestration discrimination standards and construction of a certification system; and 4) rapid forecasting of carbon sequestration risks and management decision support under different habitats.

[0006] Therefore, there is an urgent need to develop a novel carbon sequestration capacity measurement and evaluation system based on multi-source heterogeneous data fusion modeling, joint analysis of microbiome and environmental factors, deep causal inference, and dynamic risk perception. This system should be able to achieve high-precision, multi-factor, and multi-scale carbon sequestration threshold discrimination for blue carbon ecosystems such as mangroves, and also possess intelligent functions such as contextualized comparison, standardized certification, dynamic feedback, and adaptive optimization. Through a fully visualized and interactive management platform, it should provide scientific, intuitive, and quantitative decision-making support for carbon sequestration project managers, ecological compensation, and carbon trading stakeholders, addressing current challenges such as inconsistent carbon sequestration standards, large discrimination errors, and weak adaptability, thereby comprehensively improving the scientific management and sustainable development capabilities of blue carbon ecosystems. Summary of the Invention

[0007] This invention aims to address the technical challenges in assessing the carbon sequestration capacity of blue carbon ecosystems such as mangroves, including difficulties in multi-source data fusion, insufficient accuracy in determining carbon sequestration thresholds, unintelligent attribution classification and risk forecasting, and an incomplete standard system. It overcomes the assessment errors caused by single data and static standards, and achieves high-precision, multi-factor, dynamic intelligent identification and standardized certification of mangrove carbon sequestration capacity, providing reliable support for the scientific management, ecological compensation, and carbon asset trading of blue carbon ecosystems.

[0008] To achieve the above objectives, the present invention employs the following technical solution: the method comprises: Data stratification and habitat gradient construction: Based on the typical estuarine, bay and island habitat distribution of mangroves, remote sensing image interpretation and UAV topographic mapping were integrated to determine multi-scale habitat zoning and sediment samples were collected in stratification according to environmental gradient. Deep characterization of microbial-environment synergy: Using a multimodal feature fusion algorithm (MFEM), microbial community structure, metagenomics, sedimentary physicochemical properties, and climate factor data are transformed into a high-dimensional embedding space. Through an adaptive hierarchical attention mechanism, the response threshold of microbial function to environmental stress and key features of dominant carbon sink processes are dynamically captured. The inflection point of carbon sequestration efficiency is mined and identified. Based on the evolutionary threshold detection network (Evo-TDN), an evolutionary learning strategy and a dynamic threshold activation function are introduced to couple and model the carbon burial efficiency and key microbial metabolic pathways of multi-habitat samples. Through global sensitivity analysis and local variation detection, the critical signals of abrupt changes in carbon sequestration capacity under different habitat conditions are automatically identified. Carbon sink measurement and risk prediction: Based on deep causal inference and high-dimensional discriminative clustering, a multi-scenario attribution model of carbon sink function is constructed. Multi-habitat data and climate simulation results are integrated, a carbon sink threshold-environment multi-factor response map is developed, a dynamic measurement algorithm for sediment carbon sink is formed, an adaptive risk perception module is embedded, and changes in mangrove carbon sink function under typhoon and extreme rainfall climate stress are predicted to achieve real-time updates of carbon sink threshold. The application and evaluation criteria are formulated by combining model output with actual observation results to compare carbon sink thresholds under different habitats and form a carbon sink threshold measurement standard based on habitat type and microbiome characteristics.

[0009] In one scheme, the data stratification and habitat gradient construction include: first, using high-resolution remote sensing images and UAV mapping to accurately delineate the boundaries of mangrove forests and shorelines, combining DEM micro-topographic unit division, and integrating salinity, hydrodynamic, and matrix type parameters from field investigations to quantitatively stratify the environmental gradient of the habitat. A stratified random sampling method was used to collect sediment samples at different depths in each soil layer, and intelligent devices were deployed simultaneously to acquire multidimensional hydrological and climate real-time data. Microbial samples were sequenced using high-throughput sequencing to obtain microbial community structure and metagenomic information. All raw data were integrated with carbon storage analysis results to form a high-resolution, fully traceable database.

[0010] In one approach, the deep characterization of the microbial-environment synergistic features includes: using a multimodal feature fusion algorithm (MFEM) to extract and embed features from microbial community structure features, metagenomic features, sedimentary physicochemical properties, and climate and environmental factor data to obtain intermediate characterization vectors for each modality. Based on a multimodal fusion transformer, multi-channel vectors are sequentially spliced ​​according to the hierarchical structure of the spatial environment, and an adaptive hierarchical attention mechanism is used to assign dynamic weights to different modal components to achieve synergistic enhancement of high-dimensional embedded spatial features. Supervised training enables the identification and quantification of the response thresholds of microbial functions to environmental stress and carbon sequestration potential under different habitats and environmental gradients, providing collaborative embedding features for subsequent threshold discrimination and prediction model training.

[0011] In one scheme, the inflection point mining and discrimination of carbon sink efficiency features includes: an evolutionary threshold detection network (Evo-TDN), which takes a multi-habitat hierarchical high-dimensional collaborative feature representation as input and jointly models the carbon burial efficiency of the sample with the contribution of key microbial metabolic pathways; Evo-TDN dynamically optimizes the network structure and parameters through an evolutionary learning strategy, adaptively adjusts the threshold activation function to respond to sudden changes in carbon sequestration capacity caused by different habitat variables, and combines a fitness function to select the optimal structure.

[0012] In one scheme, the carbon sink measurement and risk prediction includes: based on multi-source heterogeneous data and causal mechanisms, using a deep causal inference algorithm to input the high-dimensional feature set of microorganism-environment-carbon storage synergy, climate simulation results and carbon burial response variables into a structural causal model, and establishing the causal influence weights of environmental and biological factors on carbon sink function. By using a discriminative high-dimensional clustering method, we can identify the carbon sink dynamics mechanism in multiple scenarios. Combined with the attribution results of multiple scenarios and climate scenario simulation, we can draw a threshold-environment multi-factor response map. By integrating attribution weights and response functions, a dynamic carbon sink prediction model is established, and a temporal convolutional network and attention mechanism are embedded to achieve risk range prediction and real-time forecasting of carbon sink function under different habitats. The threshold is updated and the forecast signal is output, thereby improving the intelligent estimation and management capabilities of carbon sinks.

[0013] In one scheme, the formulation of application and evaluation standards includes: taking the output of the fusion model as the core, combining field observations and actual monitoring data of multiple habitats, conducting contextualized comparison and scientific evaluation of carbon sink thresholds, matching the thresholds calculated by the model with field observation data, performing difference analysis and error assessment, summarizing the carbon sink capacity threshold ranges corresponding to different habitats and their microbiome characteristics, and formulating categorized carbon sink threshold measurement standards.

[0014] In one scheme, the evolutionary threshold detection network uses global sensitivity analysis to calculate the influence rate of key variables and performs statistical tests on the mean and variance mutations of the output mutation signal through sliding window local variation detection, thereby realizing the automatic identification and quantitative judgment of the inflection point of carbon sink efficiency characteristics.

[0015] Beneficial effects of this invention: This invention proposes a method for measuring carbon sink thresholds in mangrove sediments, integrating multiple cutting-edge technologies such as remote sensing interpretation, UAV mapping, high-dimensional multimodal feature analysis, deep causal inference, and intelligent risk forecasting to comprehensively improve the accuracy and dynamic adaptability of carbon sink threshold identification. Through meticulous data stratification and multi-scale habitat gradient construction, it can systematically reflect the carbon burial process in various mangrove ecosystems, enabling comprehensive attribution and sensitivity analysis of different environmental and biological factors. It innovatively combines microbial-environment synergistic features, multimodal fusion, and evolutionary threshold detection networks to dynamically mine inflection points and functional mutation thresholds of carbon sink capacity, significantly improving the speed and accuracy of identifying nonlinear changes in carbon sink capacity.

[0016] Furthermore, this method effectively integrates climate simulation and environmental monitoring data through high-dimensional discriminative clustering and deep causal inference, significantly enhancing the real-time perception and forward-looking forecasting capabilities of carbon sink risks under different habitats, providing a scientific basis for carbon sink management and regulation. The method also supports the development of carbon sink measurement standards based on habitat type and microbiome characteristics, enabling dynamic updates and scientific assessments of carbon sink thresholds under different scenarios, greatly promoting the quantitative research and intelligent management of mangrove carbon sink functions. Overall, this invention significantly improves the scientific rigor, accuracy, and practicality of quantifying and managing the carbon sink function of mangrove sediments, possessing broad application value in ecological protection, carbon trading, and climate adaptation. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0019] Unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. To facilitate understanding, the invention will now be described more fully with reference to the accompanying drawings. Typical embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to make the disclosure of the invention more thorough and complete.

[0020] Figure 1As shown, a method for measuring the carbon sink threshold of mangrove sediments includes the following specific steps: Step 1: Data Stratification and Habitat Gradient Construction. Based on the typical estuarine, bay, and island habitat distribution of mangroves, multi-scale habitat zoning was determined by integrating remote sensing image interpretation and UAV topographic mapping. Sediment samples were collected in stratification according to environmental gradients (salinity, hydrodynamics, matrix type). Simultaneously, multidimensional hydrological and climatic data were acquired in situ, and high-throughput microbiome and functional gene sequencing were performed. Through the construction of an integrated environmental-biological-carbon pool original database, the comprehensiveness and fine granularity of the data were ensured.

[0021] First, multi-scale spatial ecological zoning needs to be carried out around typical estuarine, bay, and island mangrove distribution areas in Guangdong Province. Through high-resolution remote sensing image interpretation and UAV low-altitude topographic mapping, the complex shorelines and mangrove distribution boundaries are accurately located, and micro-topographic units are defined using a DEM (Digital Elevation Model). Based on this, environmental parameters such as salinity, hydrodynamic conditions (flow velocity, tidal location), and matrix type obtained from field investigations are used to quantitatively stratify environmental gradients in different habitat areas. For sampling, a stratified random sampling method is adopted, selecting representative sampling points within each habitat and environmental gradient layer to drill surface and sediment samples at different depths. Simultaneously, intelligent monitoring equipment is deployed at each sampling point to acquire real-time multi-dimensional hydro-climatic dynamic data such as temperature, humidity, pH, salinity, dissolved oxygen, and tidal level, forming a spatiotemporally continuous environmental information collection network. High-throughput microbial community structure and metagenomic sequencing are performed on microbial samples to analyze microbial diversity and metabolic functions under different habitats and environmental gradients. The obtained raw data were correlated with the carbon storage analysis results to construct a high-resolution, all-element integrated database encompassing remote sensing spatial information, in-situ environmental parameters, omics, and carbon pool determination. Throughout the process, data standardization and complete metadata collection were consistently implemented, providing a sufficient, accurate, and traceable data foundation for subsequent innovative algorithm analysis.

[0022] Step 2: Deep Characterization of Microbial-Environment Synergistic Features. A customized multi-modal feature fusion algorithm (MFEM) is used to transform microbial community structure, metagenomic data, sedimentary physicochemical properties, and climatic factors into a high-dimensional embedding space. Through an adaptive hierarchical attention mechanism, the algorithm dynamically captures the response threshold of microbial function to environmental stress and key features of dominant carbon sink processes, achieving a deep characterization of nonlinear relationships between cross-scale, multi-dimensional variables.

[0023] Based on the completion of high-resolution data acquisition and database construction across multiple environmental layers, a multi-modal feature fusion algorithm (MFEM, Multi-modal Feature Embedding and Mapping) is applied to achieve high-dimensional unified analysis of multi-source data. Specifically, the microbial community structure features are first... Metagenomic features , sedimentary physicochemical properties and climate and environmental factor data After being embedded into corresponding intermediate representation vectors by their respective feature extraction networks, that is... , , , ,in The nonlinear feature encoding function represents the data of each modality. Subsequently, a multimodal fusion transformer is used to sequentially concatenate the above multi-channel vectors according to the hierarchical structure of the spatial environment, and an adaptive hierarchical attention mechanism is used to assign dynamic weights to each modal component to achieve synergistic enhancement of high-dimensional embedded spatial features.

[0024] The mathematical expression of this mechanism is as follows: For the multimodal vector aggregation corresponding to the l-th layer, it can be denoted as: ,in This represents the transformation weights of different modes in the hierarchical structure. An adaptive attention allocation function dynamically adjusts the contribution of different data sources to the overall feature representation. Through supervised or semi-supervised training, the model can autonomously capture the response thresholds of microbial function to environmental stress and carbon sequestration potential under different habitats and environmental gradients, achieving the identification and quantification of key features dominating carbon sequestration processes. Ultimately, the entire algorithm framework not only deeply reveals the complex nonlinear coupling relationships across scales and dimensions among microorganisms, the environment, and carbon sequestration, but also provides highly representative and analytically resolvable co-embedded features for subsequent threshold discrimination and prediction model training.

[0025] Step 3: Threshold-Driven Inflection Point Mining and Identification of Carbon Sequestration Efficiency Characteristics. An Evolutionary Threshold Detection Network (Evo-TDN) was developed, incorporating evolutionary learning strategies and dynamic threshold activation functions to couple and model the carbon burial efficiency and key microbial metabolic pathways of multi-habitat samples. Through global sensitivity analysis and local variation detection, critical signals indicating abrupt changes in carbon sequestration capacity under different habitat conditions were automatically identified, enabling quantitative mining of threshold inflection points in the habitat-microbe-carbon storage relationship.

[0026] After obtaining multi-habitat stratification data and multimodal synergistic features of microorganisms and the environment, the next step is to mine and identify inflection points of carbon sequestration efficiency features driven by thresholds. This invention first introduces an evolutionary threshold detection network (Evo-TDN) to represent the high-dimensional synergistic features obtained in step 2. As input, the carbon burial efficiency of the sample and contribution of key microbial metabolic pathways Joint modeling is performed. The core mechanism of Evo-TDN is to dynamically optimize the network structure and parameters through an evolutionary learning strategy and adaptively adjust the threshold activation function to sensitively respond to abrupt changes in carbon sequestration capacity caused by different habitat variables. Specifically, Evo-TDN contains an evolutionary threshold unit, the mathematical expression of which is:

[0027] in, The signal indicating the intensity of abrupt changes in carbon sequestration capacity. It is a dynamic threshold activation function that evolves with the training iteration period t. and These are the weights and bias parameters, respectively. To fuse feature vectors, This is an evolutionary adaptive threshold parameter.

[0028] By setting the fitness function

[0029] By combining genetic algorithms or other evolutionary learning methods, Evo-TDN can continuously screen for optimal structures and parameters, enabling the network to accurately identify the inflection points of carbon sequestration capacity under different habitats.

[0030] Furthermore, key environmental or microbial variables are calculated through global sensitivity analysis. Impact rate on network output And using a sliding window for local variation detection, within the interval Perform a statistical test on the abrupt change in the mean and variance: if

[0031] This suggests that a critical abrupt change in carbon sequestration capacity exists under these habitat conditions. The above process enables automatic identification and quantitative judgment of threshold inflection points from multimodal feature fusion, providing a novel model architecture with self-evolutionary adjustment capabilities for determining the carbon sequestration threshold of mangrove sediments.

[0032] Step 4: Intelligent Carbon Sequestration Measurement and Risk Prediction. Based on deep causal inference and high-dimensional discriminative clustering, a multi-scenario attribution model for carbon sequestration function with strong interpretability is constructed. Multi-habitat data and climate simulation results are integrated to develop a carbon sequestration threshold-environmental multi-factor response map, forming a dynamic measurement algorithm for sediment carbon sequestration. An adaptive risk perception module is embedded to intelligently predict changes in mangrove carbon sequestration function under climate stresses such as typhoons and extreme rainfall, enabling real-time updates of carbon sequestration thresholds.

[0033] In the specific implementation process, firstly, based on the deep causal inference algorithm, the high-dimensional feature set of microorganism-environment-carbon storage synergy of multi-habitat stratified samples is generated. Climate simulation results Response variables to carbon burial Input into the structural causal model (SCM) to build a causal directed graph. , where nodes Edge E encodes the causal relationships between variables. Through Bayesian structure learning and gradient optimization, the weights of causal influence strength are obtained. This study quantifies the direct and indirect effects of various environmental and biological factors on carbon sequestration. Subsequently, a high-dimensional discriminative clustering method (discriminative embedding spectral clustering) is employed to classify attribution variables and their response relationships across multiple scenarios, achieving the categorization and identification of carbon sequestration dynamic mechanisms. The mathematical expression of discriminative clustering is: [The text abruptly shifts to a different topic] ...for high-dimensional feature matrices... Construct a discriminative similarity matrix with label Y Solve the objective function ,in The optimal genotyping feature subspace is extracted from the clustering indicator matrix. Furthermore, by integrating multi-scenario attribution results and climate scenario simulations, a carbon sink threshold-environmental multi-factor response map is constructed. ,in This represents a combination of environmental and genetic / microbial variables. The threshold response function is used to characterize the dynamic changes in sediment carbon sink under different habitat and climate driving forces.

[0034] At the carbon sequestration level, a dynamic carbon sequestration prediction model is established based on attribution weights and response functions: ,in For deep nonlinear regression mapping, the system estimates the carbon sink level of each habitat unit in real time at the target spatiotemporal resolution. An adaptive risk perception module is embedded, utilizing a temporal convolutional network (TCN) and attention mechanism to assess carbon sequestration levels under different habitat conditions (typhoon intensity). Extreme rainfall The module predicts the variation range of carbon sequestration function under different conditions. The core formula of this module is:

[0035] in For the future Perceived and predicted values ​​of carbon sequestration risks over a given period For risk perception prediction function, Other environmental and biological variables are also considered. The system automatically updates the carbon sink threshold in real time and outputs a forecast signal when the environmental threshold approaches or exceeds a critical inflection point.

[0036] Step 5: Application and Evaluation Standard Development. Combining model output with actual observation results, carbon sink thresholds under different habitats are compared to establish carbon sink threshold measurement standards and a results visualization system based on habitat type and microbiome characteristics. A dynamic iterative optimization mechanism is established to promote continuous feedback and self-correction between the evaluation model and field observation data, achieving standardization and widespread application of carbon sink threshold measurement, and providing support for the scientific management and precise trading of blue carbon projects.

[0037] Centered on the output of the fusion model, and closely integrated with field observations and actual monitoring data from multiple habitats, this study comprehensively conducts contextualized comparisons and scientific evaluations of carbon sequestration thresholds. The specific implementation process includes: First, matching, analyzing differences, and assessing errors between the carbon sequestration thresholds calculated by the model for different habitat types (tidal flats, estuaries, bays, etc.) and field-observed carbon burial data. Through contextual comparison, the study summarizes the carbon sequestration capacity threshold ranges corresponding to various habitats and their typical microbiome characteristics, and elucidates the quantitative relationships between thresholds and multiple factors such as environmental disturbances and microbiome community function. Based on this, carbon sequestration threshold measurement standards are formulated for different habitats and microbiome configurations. To enhance the intuitive experience and management applicability of the results, a flexible and interactive results visualization system is developed, supporting the dynamic presentation of the spatiotemporal distribution of carbon sequestration thresholds, dominant environmental factors, and sources of model uncertainty, providing visual support for management decisions and stakeholders.

[0038] Based on established standards, the system incorporates a dynamic iterative optimization mechanism. This mechanism periodically compares the latest field observation data with model predictions, automatically identifying residuals and mode shifts between model outputs and actual observations. It then adaptively adjusts model parameters and structure accordingly, gradually converging to a more accurate and representative carbon sink assessment standard. Through a dynamic feedback loop, the accuracy and applicability of carbon sink threshold measurement are continuously optimized, establishing a scientific, fair, and transparent evaluation and certification system. This process provides a solid technical and data foundation for carbon sink measurement, standardized assessment, and precise trading in blue carbon ecosystems such as mangroves, thereby supporting the scientific management and widespread application of blue carbon projects in areas such as ecological compensation, carbon trading, and ecological protection.

[0039] Example: Example: Application of specific threshold measurement of carbon sink habitat in mangrove sediments in Guangdong Province Mangroves are widely distributed in Guangdong Province, including typical estuarine, bay, and island habitats such as the Pearl River Estuary and the Leizhou Peninsula, making it an important area for carbon sequestration in my country. However, most existing carbon sequestration estimation methods fail to fully consider the environmental gradients, microbiome characteristics, and influences of different habitat types, resulting in significant deficiencies in the determination of carbon sequestration thresholds. This embodiment, based on the mangrove sediment carbon sequestration threshold measurement method proposed in this invention, focuses on typical mangrove habitats in Guangdong Province. Through the fusion of big data, multimodal features, and intelligent models, it achieves accurate identification and risk assessment of habitat-specific carbon sequestration thresholds, providing a scientific basis for regional mangrove ecological management and carbon trading.

[0040] I. Data Collection and Habitat Zoning Taking the Pearl River Estuary (estuary type), Huizhou Port (harbor type), and Nan'ao Island (island type) as representative locations, high-resolution remote sensing imagery and UAV topographic mapping technology were first used to accurately delineate the distribution of mangroves and micro-topographic units. Combined with field surveys, sediment samples were collected in layers to analyze environmental parameters including salinity, hydrodynamics, and vegetation cover, and high-throughput sequencing was performed on the microbial diversity and metabolic functions of soil samples.

[0041] Table 1. Results of key environmental and biological parameters collected from mangrove sediments in different habitats.

[0042] II. Microbial-Environment Multimodal Feature Embedding and Key Threshold Feature Extraction A high-dimensional feature set was constructed using the data in Table 1 and climate simulation variables (such as annual average rainfall and frequency of extreme weather events). A multimodal feature fusion algorithm (MFEM) was used to uniformly transform microbial community structure, functional gene abundance, environmental physicochemical parameters, and climate factors into the embedding space. An adaptive hierarchical attention mechanism was employed to highlight the functional responses of microorganisms that dominate carbon sequestration processes under different habitats, automatically screening sensitive threshold factors for changes in carbon burial efficiency.

[0043] III. Automatic Identification of Inflection Points in Carbon Sequestration Efficiency Characteristics Sediment carbon burial rate (gC·m) - ²·a - ¹) Using the response variable, an evolutionary threshold detection network (Evo-TDN) is constructed, introducing a dynamic threshold activation function to couple and model multi-habitat samples, while considering the global sensitivity and local variation of influencing factors. A sliding window method is used to input time-series data to detect abrupt change signals and determine the critical inflection point of carbon sequestration capacity.

[0044] Table 2 Relationship between carbon burial rate and threshold response under different habitats

[0045] IV. Carbon Sequestration Measurement and Risk Prediction Based on the integration of the aforementioned threshold identification results and multi-scenario climate simulations, a multi-factor attribution model for carbon sequestration is established to dynamically measure changes in carbon sequestration capacity and predict risk ranges under different extreme climate scenarios. A temporal convolutional-attention network is employed, combined with typhoon and extreme rainfall scenarios, to calculate the critical transition probability of carbon sequestration function and output forecast signals in real time.

[0046] Table 3. Carbon sequestration and risk prediction results of harbour mangroves under extreme climate influences

[0047] V. Application and Threshold Evaluation Standards Development Based on model output and field observations, scenario-based comparisons of carbon sequestration thresholds across multiple habitats were conducted. For example, in typhoon-prone years, the carbon sequestration rate of harbor-type mangroves fell below the 3.8% organic matter content threshold, resulting in a significant decline in their carbon sequestration function. Therefore, corresponding carbon sequestration threshold ranges were established for different mangrove habitat types in Guangdong Province, and the differences and sources of error were scientifically assessed, providing a refined foundation for ecological compensation and carbon trading.

[0048] Table 4 Carbon Sequestration and Threshold Reference Standards in Typical Mangrove Habitats of Guangdong Province

[0049] This embodiment successfully revealed the critical carbon sink thresholds and their main controlling factors for major mangrove habitat types in Guangdong Province through hierarchical data acquisition and environmental-microbial multimodal fusion; it also innovatively achieved carbon sink functional risk prediction for extreme climate processes. The results show that traditional unified carbon sink estimation methods have significant biases, while this invention can dynamically sense carbon sink changes under different habitats and climate stresses, achieving a more practical categorized carbon sink threshold measurement standard. This method provides solid data support and a technical foundation for the scientific assessment, carbon sink management, and climate adaptation decision-making of mangrove ecosystems in Guangdong Province and even larger regions.

[0050] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0051] It should be understood that the above detailed description of the technical solutions of the present invention with reference to preferred embodiments is illustrative and not restrictive. Those skilled in the art can modify the technical solutions described in the embodiments or make equivalent substitutions for some of the technical features based on reading this specification; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for measuring the carbon sink threshold of mangrove sediments, characterized in that: The method includes: Data stratification and habitat gradient construction: Based on the typical estuarine, bay and island habitat distribution of mangroves, remote sensing image interpretation and UAV topographic mapping were integrated to determine multi-scale habitat zoning and sediment samples were collected in stratification according to environmental gradient. Deep characterization of microbial-environment synergy: Using a multimodal feature fusion algorithm (MFEM), microbial community structure, metagenomics, sedimentary physicochemical properties, and climate factor data are transformed into a high-dimensional embedding space. Through an adaptive hierarchical attention mechanism, the response threshold of microbial function to environmental stress and key features of dominant carbon sink processes are dynamically captured. The inflection point of carbon sequestration efficiency is mined and identified. Based on the evolutionary threshold detection network (Evo-TDN), an evolutionary learning strategy and a dynamic threshold activation function are introduced to couple and model the carbon burial efficiency and key microbial metabolic pathways of multi-habitat samples. Through global sensitivity analysis and local variation detection, the critical signals of abrupt changes in carbon sequestration capacity under different habitat conditions are automatically identified. Carbon sink measurement and risk prediction: Based on deep causal inference and high-dimensional discriminative clustering, a multi-scenario attribution model of carbon sink function is constructed. Multi-habitat data and climate simulation results are integrated, a carbon sink threshold-environment multi-factor response map is developed, a dynamic measurement algorithm for sediment carbon sink is formed, an adaptive risk perception module is embedded, and changes in mangrove carbon sink function under typhoon and extreme rainfall climate stress are predicted to achieve real-time updates of carbon sink threshold. The application and evaluation criteria are formulated by combining model output with actual observation results to compare carbon sink thresholds under different habitats and form a carbon sink threshold measurement standard based on habitat type and microbiome characteristics.

2. The method for measuring the carbon sink threshold of mangrove sediments according to claim 1, characterized in that: The aforementioned data stratification and habitat gradient construction include: firstly, using high-resolution remote sensing images and UAV mapping to accurately delineate mangrove forests and shoreline boundaries, combining DEM micro-topographic unit division, and integrating salinity, hydrodynamic, and matrix type parameters from field investigations to quantitatively stratify the environmental gradient of the habitat; A stratified random sampling method was used to collect sediment samples at different depths in each soil layer, and intelligent devices were deployed simultaneously to acquire multidimensional hydrological and climate real-time data. Microbial samples were sequenced using high-throughput sequencing to obtain microbial community structure and metagenomic information. All raw data were integrated with carbon storage analysis results to form a high-resolution, fully traceable database.

3. The method for measuring the carbon sink threshold of mangrove sediments according to claim 1, characterized in that: The aforementioned deep characterization of microbial-environment synergistic features includes: using a multimodal feature fusion algorithm (MFEM) to extract and embed features from microbial community structure features, metagenomic features, sedimentary physicochemical properties, and climate and environmental factor data to obtain intermediate characterization vectors for each modality; Based on a multimodal fusion transformer, multi-channel vectors are sequentially spliced ​​according to the hierarchical structure of the spatial environment, and an adaptive hierarchical attention mechanism is used to assign dynamic weights to different modal components to achieve synergistic enhancement of high-dimensional embedded spatial features. Supervised training enables the identification and quantification of the response thresholds of microbial functions to environmental stress and carbon sequestration potential under different habitats and environmental gradients, providing collaborative embedding features for subsequent threshold discrimination and prediction model training.

4. The method for measuring the carbon sink threshold of mangrove sediments according to claim 1, characterized in that: The aforementioned inflection point mining and discrimination of carbon sink efficiency features includes: an evolutionary threshold detection network (Evo-TDN), which takes a multi-habitat hierarchical high-dimensional collaborative feature representation as input and jointly models the carbon burial efficiency of samples with the contribution of key microbial metabolic pathways; Evo-TDN dynamically optimizes the network structure and parameters through an evolutionary learning strategy, adaptively adjusts the threshold activation function to respond to sudden changes in carbon sequestration capacity caused by different habitat variables, and combines a fitness function to select the optimal structure.

5. The method for measuring the carbon sink threshold of mangrove sediments according to claim 1, characterized in that: The carbon sink measurement and risk prediction includes: based on multi-source heterogeneous data and causal mechanisms, using a deep causal inference algorithm to input the high-dimensional feature set of microorganism-environment-carbon storage synergy, climate simulation results and carbon burial response variables into a structural causal model, and establishing the causal influence weights of environmental and biological factors on carbon sink function. By using a discriminative high-dimensional clustering method, we can identify the carbon sink dynamics mechanism in multiple scenarios. Combined with the attribution results of multiple scenarios and climate scenario simulation, we can draw a threshold-environment multi-factor response map. By integrating attribution weights and response functions, a dynamic carbon sink prediction model is established, and a temporal convolutional network and attention mechanism are embedded to achieve risk range prediction and real-time forecasting of carbon sink function under different habitats. The threshold is updated and the forecast signal is output, thereby improving the intelligent estimation and management capabilities of carbon sinks.

6. The method for measuring the carbon sink threshold of mangrove sediments according to claim 1, characterized in that: The application and evaluation criteria are formulated as follows: taking the output of the fusion model as the core, combining field observation and actual monitoring data of multiple habitats, conducting contextualized comparison and scientific evaluation of carbon sink thresholds, matching the thresholds calculated by the model with the field observation data, performing difference analysis and error assessment, summarizing the carbon sink capacity threshold ranges corresponding to different habitats and their microbiome characteristics, and formulating categorized carbon sink threshold measurement standards.

7. The method for measuring the carbon sink threshold of mangrove sediments according to claim 4, characterized in that: The aforementioned evolutionary threshold detection network uses global sensitivity analysis to calculate the influence rate of key variables and performs statistical tests on the mean and variance mutations of the output mutation signal through sliding window local variation detection, thereby achieving automatic identification and quantitative judgment of the inflection point of carbon sink efficiency characteristics.

Images