Dynamic monitoring and accounting system for organic carbon reserves of lake and reservoir sediments

Abstract

The invention discloses a dynamic monitoring and accounting system for organic carbon reserves of lake and reservoir sediments, belongs to the technical field of carbon cycle monitoring, and aims to solve the problems of spatial-temporal resolution contradiction and insufficient multi-source data fusion caused by a single data source in a traditional method. A space-time continuous fusion data product is generated in combination with an adaptive fusion algorithm, the system comprises a multi-source data acquisition module, a preprocessing module, an uncertainty quantification module, a fusion processing module and a verification feedback module, deep fusion of space-based remote sensing, foundation in-situ and experimental analysis data is achieved, and through independent actual measurement data verification and parameter feedback optimization, the space-time continuous fusion data product is obtained. And the system forms a precision improvement closed loop, so that the stability and accuracy of long-term operation are ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of carbon cycle monitoring, in particular to a lake and reservoir sediment organic carbon storage dynamic monitoring and accounting system. BACKGROUND

[0002] As active carbon sinks, the burial flux of organic carbon in the sediments of lakes and reservoirs is usually smaller than the carbon emission flux, but it is an important long-term carbon sink and plays an important role in regional and even global carbon cycle. Compared with the ocean, inland water bodies such as lakes and reservoirs have higher sedimentation rates, lower oxygen supply, and higher proportions of terrestrial organic carbon, showing higher organic carbon burial efficiency. In recent years, with the intensification of human activities such as soil erosion, dam building and eutrophication, the organic carbon burial rate of lakes and reservoirs has shown an increasing trend, and it has become increasingly important to accurately and dynamically monitor and account for its storage. Traditional monitoring methods rely on single field sampling and laboratory analysis, or single remote sensing inversion technology, which cannot meet the needs of dynamic monitoring in a large range, high frequency and high precision.

[0003] At present, the existing related technical solutions focus on the use of specific links or single data sources. For example, the invention patent with publication number CN118707042A "Improved method and system for estimating organic carbon content in lake sediments", which collects undisturbed sediment cores and divides them into multiple layers, measures parameters such as density, moisture content, and median particle size of sediment, and then uses an improved empirical formula to calculate the organic carbon content. This method takes into account the differences in vertical distribution of sediments, improving the accuracy of laboratory-scale estimates, but it still relies heavily on limited field sampling points, making it difficult to achieve continuous coverage in space and dynamic updating in time, and not introducing remote sensing and other large-scale observation data, which cannot solve the fundamental problem of insufficient spatial and temporal resolution in large regional scale.

[0004] Another patented technology involving large-area monitoring (CN118010639A, a remote sensing method for particulate organic carbon storage in lake groups) mainly relies on remote sensing reflectance data to build an inversion model to estimate particulate organic carbon (POC) concentration, and combines lake depth and area to calculate storage. Although this technology has the advantage of spatial coverage, it relies solely on remote sensing data and lacks in-situ measured data for real-time correction and verification of the model, making it difficult to quantify and control the uncertainty of the inversion results. Especially for inland water bodies with complex optical characteristics, the reliability and stability of its estimation results face challenges. At the same time, although this patented technology attempts to apply multi-source data (CN118010639A mentions the use of remote sensing reflectance and ground synchronous data), it mostly stays at the level of building models separately or simple data input, failing to deeply integrate multi-source heterogeneous data at the mechanistic level, and failing to build an effective uncertainty propagation and quantification model, making it difficult to achieve cross-validation of errors and substantial improvement in accuracy.

[0005] To address the aforementioned shortcomings in existing technologies, this invention aims to resolve the core problem in dynamic monitoring of organic carbon storage in lake and reservoir sediments: the spatiotemporal resolution discrepancies and difficulties in cross-validation of errors due to the limitations of single data sources. Specifically, while single remote sensing data offers broad coverage, it suffers from insufficient inversion accuracy and susceptibility to interference; single in-situ monitoring data, while highly accurate, has a limited representative range and is costly; and single experimental analysis data inherently suffers from spatiotemporal discontinuities. Existing technologies have failed to deeply fuse multi-source heterogeneous data and quantify their uncertainties, thus failing to optimally utilize the complementarity of data from different sources. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a dynamic monitoring and accounting system for organic carbon storage in lake and reservoir sediments. By creatively integrating probabilistic graphical models, multi-source heterogeneous data fusion theory, and lake and reservoir sediment carbon monitoring practices across disciplines, a dynamic monitoring and accounting system based on uncertainty quantification and optimal fitting is constructed. By introducing a probabilistic graphical model based on Bayes' theorem to structurally quantify the uncertainty of various data sources, and designing an adaptive fusion algorithm accordingly, the optimal fitting of multi-source data under a unified spatiotemporal benchmark is achieved, thereby generating a synergistic enhancement effect at the technical level and significantly improving the accuracy and reliability of organic carbon storage estimation at a large regional scale.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: On one hand, a dynamic monitoring and accounting system for organic carbon storage in lake and reservoir sediments, comprising:

[0008] The multi-source data acquisition module is used to acquire optical data retrieved from airborne remote sensing, hydrological and water quality parameter data from ground-based in-situ sensors, and physicochemical property data from sediment experiments.

[0009] The data preprocessing and standardization module is used to perform spatiotemporal benchmark unification, quality control and standardization processing on the heterogeneous data acquired by the multi-source data acquisition module, and generate a spatiotemporally aligned standardized dataset.

[0010] The uncertainty quantification module, connected to the data preprocessing and standardization module, is used to quantify the uncertainty of various types of data in the standardized dataset based on a probabilistic graphical model, and output the uncertainty measure of each data point.

[0011] The multi-source data fusion processing module is connected to the uncertainty quantification module. It is used to receive the standardized dataset and its corresponding uncertainty measure, and to use a spatiotemporal adaptive fusion algorithm based on uncertainty weighting to perform optimal fitting on the multi-source data, thereby generating a spatiotemporally continuous fusion data product of lake and reservoir sediment organic carbon storage.

[0012] The system output module is connected to the multi-source data fusion processing module and is used to output and store the fused data products.

[0013] Furthermore, the probability graphical model in the uncertainty quantification module is a hierarchical probability model based on Bayes' theorem. This model quantifies the uncertainty of multi-source heterogeneous data in a structured way by introducing latent variables of data source reliability and spatiotemporal correlation. The mathematical expression of this model is the formula for calculating the posterior probability distribution:

[0014]

[0015] Wherein, P(M|D) represents the posterior probability distribution of the model parameter set M under the condition of obtaining the observed dataset D, the model parameter set M includes the true value of organic carbon content, the systematic bias of the data source and random error, P(D|M) represents the likelihood function, which characterizes the probability of observing dataset D under the model parameter set M. This likelihood function is jointly defined by the observation error distribution model of each data source, P(M) represents the prior probability distribution of the model parameter set M, which is determined based on historical data and domain knowledge and is used to introduce constraints on the reasonable range of parameter values, and P(D) represents the evidence factor, which is regarded as a normalization constant when calculating the posterior probability of a specific parameter;

[0016] The hierarchical probability model dynamically adjusts the weights of different data sources in the prior definition and likelihood function through the latent variable of data source reliability, and transmits uncertainty information in the spatiotemporal dimension through the latent variable of spatiotemporal correlation, thereby realizing the structuring and quantification of uncertainty of all input data points.

[0017] Furthermore, the spatiotemporal adaptive fusion algorithm based on uncertainty weighting used in the multi-source data fusion processing module operates as follows:

[0018] Assign an adaptive weight to each standardized data point that is inversely proportional to its uncertainty measure, and perform a weighted average within a spatiotemporal window, while also considering the spatiotemporal correlation between data points. The mathematical expression of this algorithm is as follows:

[0019]

[0020] in, Z represents the fused estimate of organic carbon storage at spatial location s and time point t, where n represents the total number of data source types, m represents the total number of data points within the defined spatiotemporal neighborhood window, and Z... i,j (s,t) represents the standardized observation of the j-th data point in the i-th data source at location s and time t, where λ i,j (s,t) represents the adaptive weight of the j-th data point in the i-th data source when participating in the fusion calculation, ω i K represents the overall credibility coefficient assigned to the i-th type of data source based on prior knowledge. s (‖ss j ‖) represents the spatial kernel function, used to characterize the spatial distance between the target point and the fusion target point as ‖ss j The weight of a data point decreases with distance according to the law of K. t (‖tt j ‖) represents the time kernel function, used to characterize the time distance between the target point and the fusion point as ‖tt. j The decay pattern of the weight of the data points in the equation with increasing time interval. This represents the uncertainty metric value of the j-th data point in the i-th type of data source provided by the uncertainty quantification module at location s and time t;

[0021] The algorithm uses the adaptive weight λ i,j (s,t) simultaneously weighs the uncertainty of the data point itself, its spatiotemporal proximity to the target point, and the overall credibility of its data source to achieve dynamic adaptive adjustment of the fusion weights.

[0022] Furthermore, the multi-source data acquisition module specifically includes:

[0023] The airborne remote sensing data unit is used to retrieve data on dissolved organic carbon concentration, particulate organic carbon concentration, colored soluble organic matter concentration, water surface temperature, and chlorophyll a concentration of a large area of ​​lake and reservoir surface water by using multispectral and hyperspectral sensors mounted on any platform such as satellites or drones.

[0024] The ground-based in-situ sensing data unit is used to periodically and in real-time collect water level, water temperature, transparency, pH value, conductivity, dissolved oxygen concentration and underwater spectral data of monitoring points through a network of intelligent sensing nodes deployed in the lake and reservoir water.

[0025] The experimental analysis data unit is used to collect sediment column or surface samples from lakes and reservoirs, and then conduct laboratory physicochemical analysis to obtain data on the dry density, water content, grain size distribution, total organic carbon content, and stable isotope ratios of the sediments.

[0026] The multi-source data acquisition module provides area spatial coverage through the airborne remote sensing data unit, provides continuous time series data of key points through the ground-based in-situ sensing data unit, and provides high-precision calibration and verification benchmark data through the experimental analysis data unit. Together, the three constitute a spatiotemporally complementary heterogeneous data acquisition system.

[0027] Furthermore, the execution flow of the data preprocessing and standardization module includes:

[0028] The spatiotemporal reference unification submodule is used to convert all data acquired by the multi-source data acquisition module to a unified spatial coordinate system and time reference, and to use the Kriging spatial interpolation algorithm and time series interpolation algorithm to convert the point in-situ data and experimental data into area data that matches the remote sensing data grid, fill in the missing data periods, and generate a preliminary spatiotemporally aligned dataset.

[0029] The quality control and outlier removal submodule is connected to the spatiotemporal benchmark unification submodule. It is used to perform anomaly detection on each data channel in the preliminary dataset based on the statistical 3σ criterion and knowledge thresholds in the field of lake and reservoir water environment, and remove outlier data points that deviate significantly from the normal range.

[0030] The standardization processing submodule, connected to the quality control and outlier removal submodule, is used to normalize the dataset after quality control, eliminate the differences caused by different physical dimensions, and convert the numerical range of all datasets to the same interval, generating the final standardized dataset input to the uncertainty quantification module.

[0031] Furthermore, the uncertainty quantification module works in conjunction with the multi-source data fusion processing module to ensure the complete transfer of uncertainty information from quantification to fusion.

[0032] The uncertainty quantification module outputs the uncertainty measure. It is directly used as an important input parameter for the fusion algorithm in the multi-source data fusion processing module;

[0033] After completing the fusion calculation, the multi-source data fusion processing module also performs uncertainty propagation calculations on the fusion results. Its output includes not only the fusion estimate of organic carbon storage. Simultaneously output the uncertainty of this estimate. The formula for calculating the uncertainty of the fusion result is derived based on the error propagation law:

[0034]

[0035] in, Indicates the fusion estimate The variance at position s and time t is used to measure its uncertainty; Indicates the fusion estimate Relative to the j-th observation Z in the i-th data source i,j The partial derivative of (s,t) reflects the sensitivity of the contribution of the observation to the fusion result. This represents the uncertainty measure of the j-th observation within the i-th type of data source.

[0036] Through this calculation, the final fused data product provided by the system output module simultaneously includes the best estimate of organic carbon storage and its corresponding uncertainty quantification index.

[0037] Furthermore, it also includes a verification and feedback module that connects to the system output module;

[0038] The verification and feedback module includes:

[0039] An independent verification data import unit is used to import measured organic carbon storage data of lake and reservoir sediments that have not participated in the fusion calculation process of the multi-source data fusion processing module, as a verification benchmark.

[0040] The accuracy evaluation unit, connected to the independent verification data import unit, is used to compare the fused data product generated by the system output module with the verification benchmark at the same spatiotemporal location, and calculate multiple accuracy evaluation indicators such as root mean square error, mean absolute error, and coefficient of determination.

[0041] The model parameter feedback optimization unit is connected to the accuracy evaluation unit and is used to generate an adjustment signal based on the accuracy evaluation result, and feed the adjustment signal back to the probabilistic graphical model in the uncertainty quantification module.

[0042] The prior distribution parameters or likelihood function parameters in the probabilistic graphical model are iteratively optimized to achieve self-improvement and adaptive adjustment of the processing accuracy of the entire system.

[0043] On the other hand, a method for dynamic monitoring and calculation of organic carbon storage in lake and reservoir sediments, the specific steps of which are as follows:

[0044] S100. The multi-source data acquisition module synchronously acquires airborne remote sensing inversion optical data, ground-based in-situ sensor hydrological and water quality parameter data, and sediment experimental analysis physicochemical property data.

[0045] S200. The data preprocessing and standardization module performs a spatiotemporal reference unification operation on the acquired airborne remote sensing inversion optical data, the ground-based in-situ sensor hydrological and water quality parameter data, and the sediment experimental analysis physicochemical property data, converting all data to a unified spatial coordinate system and time reference. Then, a quality control operation is performed to remove abnormal data points. Finally, standardization processing is performed to eliminate dimensional differences and generate a standardized dataset.

[0046] S300. The standardized dataset is received through the uncertainty quantification module, and the hierarchical probability model based on Bayes' theorem is used to analyze various types of data in the standardized dataset, quantify the uncertainty of each data point, and output a dataset with an uncertainty metric.

[0047] S400. The multi-source data fusion processing module receives the dataset with uncertainty measurement and uses a spatiotemporal adaptive fusion algorithm based on uncertainty weighting to perform fusion calculation on the multi-source data, generating a spatiotemporal continuous data product that simultaneously contains the fusion estimate of organic carbon storage and its uncertainty index.

[0048] S500, The spatiotemporal continuous data product is output and stored through the system output module;

[0049] S600. The verification and feedback module uses independent measured data to evaluate the accuracy of the spatiotemporal continuous data product, and generates a feedback signal based on the evaluation results to optimize the model parameters in the uncertainty quantification module.

[0050] Compared with existing technologies, this dynamic monitoring and accounting system for organic carbon storage in lake and reservoir sediments has the following advantages:

[0051] I. This invention addresses the problems of spatiotemporal resolution discrepancies and difficulties in cross-validation of errors caused by single data sources in monitoring organic carbon storage in lake and reservoir sediments by constructing a probabilistic graphical model based on Bayes' theorem and an adaptive fusion algorithm. Specifically, the probabilistic graphical model quantifies the uncertainties of airborne remote sensing, ground-based in-situ, and experimental analysis data, providing a weighting basis for multi-source data fusion; the adaptive fusion algorithm achieves optimal fitting under a unified spatiotemporal benchmark, generating high-precision fused data products, thereby overcoming the technical limitations of traditional single data sources and significantly improving the accuracy and reliability of organic carbon storage estimation at large regional scales.

[0052] Second, this invention solves the problems of insufficient multi-source data fusion and long-term system accuracy optimization by using a spatiotemporal adaptive fusion algorithm based on uncertainty weighting and a verification feedback module. Specifically, uncertainty measurement serves as the basis for weight allocation, ensuring that high-reliability data dominates the fusion result; weighted averaging within the spatiotemporal window takes into account data continuity and improves fusion efficiency; the verification module uses independent measured data to evaluate accuracy and provides feedback to optimize model parameters, forming a closed loop for accuracy improvement. This achieves optimal fusion of multi-source data and system self-optimization, improving data utilization efficiency and long-term operational stability.

[0053] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0055] Figure 1 This is a system architecture diagram of the present invention;

[0056] Figure 2 This is a flowchart of the process of the present invention;

[0057] Figure 3 This is a flowchart illustrating the core data processing and uncertainty propagation process of this invention. Detailed Implementation

[0058] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0059] Example 1

[0060] like Figure 1 and Figure 3As shown, this embodiment addresses the shortcomings of existing lake and reservoir sediment organic carbon storage monitoring technologies, such as insufficient spatiotemporal resolution of single data sources, difficulty in quantifying uncertainty, and insufficient data fusion depth. Taking a typical freshwater lake or reservoir as the monitoring object, it implements a dynamic monitoring and accounting system for lake and reservoir sediment organic carbon storage by constructing a complete technical process of multi-source data acquisition, preprocessing standardization, uncertainty quantification, spatiotemporal adaptive fusion, and verification feedback optimization. In this embodiment, a heterogeneous data acquisition system is constructed through airborne remote sensing, ground-based in-situ sensing, and laboratory analysis. A hierarchical probability model based on Bayes' theorem is used to quantify the uncertainty of each data source. Combined with an uncertainty-weighted spatiotemporal adaptive fusion algorithm, optimal fitting of multi-source data is achieved. The system accuracy is improved through independent verification and parameter feedback optimization, ultimately generating a spatiotemporally continuous and accuracy-controllable organic carbon storage fusion data product, effectively solving the accuracy and reliability problems of dynamic monitoring of organic carbon storage at a large regional scale.

[0061] Background and preparation: This embodiment selects a typical temperate freshwater lake (hereinafter referred to as "target lake") as the monitoring object. The lake has the functions of flood control, water supply and ecological regulation. The change of sediment organic carbon storage has an important impact on the regional carbon cycle. Moreover, its water optical characteristics, hydrological conditions and sediment types are common to lakes and reservoirs, making it suitable as the implementation carrier of this system.

[0062] The monitoring period is set to one complete hydrological year to cover the dynamic changes in organic carbon storage during the high-water, normal-water, and low-water periods of the lake and reservoir. The monitoring spatial range covers the entire water area of ​​the target lake and reservoir and the surrounding nearshore sedimentary areas to ensure coverage of the differences in organic carbon distribution in different sedimentary environments (such as the deep water area in the center of the lake, the shallow water area near the shore, and the estuary sedimentary area).

[0063] Preliminary preparations include: equipment selection and commissioning: for the airborne data acquisition end, satellites equipped with multispectral sensors (resolution meeting the monitoring needs of the lake and reservoir area) and drones equipped with hyperspectral sensors will be selected; for the ground-based data acquisition end, a monitoring network consisting of intelligent sensor nodes will be deployed, with each node equipped with modules for detecting water level, water temperature, pH value, conductivity, dissolved oxygen, and underwater spectroscopy; for the laboratory analysis end, sediment sample processing equipment (such as freeze dryers and particle size analyzers) and organic carbon detection equipment (such as elemental analyzers and stable isotope ratio mass spectrometers) will be prepared.

[0064] Basic data collection: Acquire geographic information data (such as digital elevation models and water boundary vector data) of the target lakes and reservoirs, historical monitoring data (such as measured values ​​of sediment organic carbon content in the past 5 years and time series data of hydrological and water quality parameters), and domain knowledge (such as the normal parameter range of lake and reservoir water environment in the region and the distribution characteristics of sediment types) to provide a basis for subsequent model parameter setting and quality control;

[0065] System software deployment: Deploy data processing and analysis software on the server side, integrating Kriging spatial interpolation algorithm, time series interpolation algorithm, hierarchical probability model calculation module based on Bayes' theorem, and uncertainty weighted spatiotemporal adaptive fusion algorithm module to ensure smooth data interaction between modules and meet real-time or near real-time data processing requirements.

[0066] Implementation of the multi-source data acquisition module: The multi-source data acquisition module achieves synchronous acquisition of heterogeneous data through the collaborative work of airborne remote sensing data units, ground-based in-situ sensing data units, and experimental analysis data units, thus constructing a spatiotemporally complementary data foundation. The specific implementation process is as follows:

[0067] Implementation of the Space-Based Remote Sensing Data Unit: The space-based remote sensing data unit adopts a collaborative observation mode of "satellite + UAV" to acquire optical and environmental parameters related to organic carbon in the surface water of large-scale lakes and reservoirs. Satellite observation selects multispectral satellites passing over target lakes and reservoirs, acquiring remote sensing image data with a revisit cycle of 10-15 days. After radiometric calibration, atmospheric correction, and water extraction preprocessing, data on dissolved organic carbon concentration, particulate organic carbon concentration, colored soluble organic matter concentration, water surface temperature, and chlorophyll a concentration are obtained using a validated inversion model. UAV observation supplements data for areas where satellite data spatial resolution is insufficient or cloud cover leads to missing data. Flight altitude is set according to the monitoring area and resolution requirements, and the flight path uses a grid pattern to ensure no blind spots after image stitching. Higher spatial resolution water organic carbon parameters are obtained through hyperspectral data inversion, complementing the satellite data at a higher scale. The core function of the space-based remote sensing data unit is to provide large-scale, periodic areal data, solving the problem of insufficient spatial coverage in traditional in-situ monitoring.

[0068] Implementation of Ground-Based In-Situ Sensing Data Units: Ground-based in-situ sensing data units acquire data by deploying a network of intelligent sensor nodes in the target lake / reservoir. These nodes are positioned in key monitoring areas such as the lake center, near the shore, and estuary. The node spacing is determined based on the lake / reservoir area and the heterogeneity of the sedimentary environment to ensure coverage of major sedimentary types. Each sensor node employs a dual power supply mode, combining solar power and battery backup, to ensure long-term stable operation. Data transmission utilizes wireless communication technology to upload monitoring data to the server in real-time or near real-time.

[0069] The monitoring modes of the sensor nodes are set according to parameter characteristics: water level, water temperature, pH value, conductivity, and dissolved oxygen concentration are collected periodically at a frequency of 1 hour per acquisition to ensure the capture of temporal changes in hydrological and water quality parameters; underwater spectral data are collected at fixed times each day (such as midday when there are no clouds) to avoid interference from changes in lighting conditions. The collected data is transmitted to the server in real time and temporarily stored in the raw database for subsequent preprocessing. The core function of this unit is to provide continuous time-series data for key areas, compensating for the insufficient temporal resolution and susceptibility to weather interference inherent in remote sensing data.

[0070] Experimental Analysis Data Unit Implementation: The experimental analysis data unit acquires high-precision sedimentary physical and chemical property data through field sampling and laboratory analysis, serving as the calibration and verification benchmark for the system. Sampling is conducted quarterly, synchronized with the acquisition time of airborne and ground-based data to ensure spatiotemporal matching of the data. Sampling points coincide with the locations of ground-based sensor nodes, and additional sampling points are added in typical sedimentary areas where no sensor nodes are deployed to ensure the representativeness of the sampling points.

[0071] Sediment samples were collected using a combination of columnar and surface samplers: the columnar sample collection depth was set according to the sediment thickness of the target lake / reservoir to analyze the vertical distribution characteristics of organic carbon; the surface sample collection depth was 0-10 cm to analyze the horizontal distribution characteristics of organic carbon in surface sediments. After collection, the samples were processed according to standard experimental procedures upon arrival at the laboratory.

[0072] Determination of dry density and moisture content: After the sample is freeze-dried, the dry weight is weighed and the volume is measured to calculate the dry density; the moisture content is calculated based on the difference between the wet weight and the dry weight.

[0073] Particle size distribution determination: The proportions of sand, silt, and clay in the sample were analyzed using a laser particle size analyzer;

[0074] Total organic carbon content determination: The total organic carbon content was determined using an elemental analyzer after removing inorganic carbon by acid hydrolysis.

[0075] Determination of stable isotope ratios of organic carbon: δ¹² was determined using a stable isotope ratio mass spectrometer. 13 C-value. The data obtained from the experimental analysis are organized and stored in the experimental database. The core function of this unit is to provide high-precision and high-reliability benchmark data, providing a basis for uncertainty quantification and fusion result verification.

[0076] Implementation of the data preprocessing and standardization module: The data preprocessing and standardization module, for heterogeneous data acquired by the multi-source data acquisition module, generates a standardized dataset that is spatiotemporally aligned, of acceptable quality, and with consistent dimensions through spatiotemporal benchmark unification, quality control, and standardization. This lays the foundation for subsequent uncertainty quantification. The specific implementation process is as follows:

[0077] Implementation of the Spatiotemporal Reference Unification Submodule: The core task of the spatiotemporal reference unification submodule is to transform three types of data—airborne, ground-based, and experimental—into a unified spatiotemporal framework. Regarding the spatial reference, the CGCS2000 geodetic coordinate system is adopted as the unified spatial coordinate system. Satellite and UAV remote sensing data are transformed to this coordinate system through image georegistration, with registration accuracy controlled within one pixel. The coordinates of ground-based in-situ sensing nodes and experimental sampling points are obtained through GNSS positioning, with errors controlled at the centimeter level. For point-like ground-based in-situ data and experimental data, the Kriging spatial interpolation algorithm is used, with the grid resolution of the remote sensing data as the reference, to interpolate the point data into area data, ensuring matching with the remote sensing data grid.

[0078] Regarding the time reference, Coordinated Universal Time (UTC) is adopted as the unified time reference, and the acquisition time of all data is converted to this reference: the imaging time of satellite data, the flight time of UAV, the acquisition time of in-situ sensors, and the experimental sampling time are all recorded in UTC time, accurate to the second. For periods with missing data (such as satellite cloud obstruction or temporary sensor failure), time series interpolation algorithms (such as linear interpolation or spline interpolation) are used to fill in the missing data. The interpolation period is set according to the length of missing data to ensure that the generated preliminary dataset has no obvious time discontinuities and achieves complete alignment of data in the spatiotemporal dimensions.

[0079] The quality control and outlier removal submodule ensures data validity based on statistical criteria and domain knowledge. First, it performs anomaly detection on the initial dataset using the statistical 3σ criterion: calculating the mean and standard deviation of each data channel (e.g., dissolved organic carbon concentration, water temperature, total organic carbon content), and marking data points exceeding the range of "mean ± 3 times standard deviation" as suspected outliers. Second, it performs a secondary judgment on suspected outliers using domain knowledge thresholds for lake and reservoir water environments: for example, the normal annual temperature range for the target lake / reservoir is 4-30℃, and the normal pH range is 6.5-8.5. If a suspected outlier exceeds these domain thresholds, it is identified as an outlier and removed; if the suspected outlier is within the domain threshold range, it is retained, avoiding the accidental deletion of valid data due to reliance solely on statistical criteria.

[0080] For data gaps after outlier removal, the same time series interpolation algorithm as the unified spatiotemporal benchmark submodule is used to fill them, ensuring the continuity of the dataset. After quality control is completed, a qualified intermediate dataset is output, which can meet the requirements of subsequent standardization processing.

[0081] Standardization Processing Submodule Implementation: The standardization processing submodule is used to eliminate differences in the physical dimensions of different data, ensuring that multi-source data can be fused and calculated. This embodiment uses the min-max normalization method to process the qualified intermediate dataset, uniformly converting the numerical range of all data channels to the [0,1] interval. The normalization formula is as follows:

[0082]

[0083] in, The standardized data values, x represents the original data value. min x is the minimum value of this data channel. max This is the maximum value for this data channel.

[0084] In calculating x min With x max At that time, the target lake / reservoir's historical data and domain knowledge are combined to determine, for example, the x-value of total organic carbon content. min Take the minimum measured value of the lake / reservoir over the past 5 years, x max The smaller of the measured maximum value over the past 5 years and the theoretically reasonable upper limit is selected to avoid the impact of extreme outliers (which have been removed) on the standardization range. After standardization, a standardized dataset is generated, which serves as the input data for the uncertainty quantification module.

[0085] Implementation of the Uncertainty Quantification Module: Based on a hierarchical probability model using Bayes' theorem, the uncertainty quantification module performs structured quantification of the uncertainty in the standardized dataset, outputting an uncertainty measure for each data point. This provides a weighting basis for subsequent data fusion. This module is one of the core innovative components of this system, and its specific implementation process is as follows:

[0086] Hierarchical Probability Model Construction: In this embodiment, the hierarchical probability model is based on Bayes' theorem. By introducing "latent variables of data source reliability" and "latent variables of spatiotemporal correlation," it achieves accurate quantification of the uncertainty of multi-source heterogeneous data. The core mathematical expression of the model is the posterior probability distribution formula:

[0087]

[0088] The definitions and physical meanings of each parameter are as follows:

[0089] D: Observation dataset, also known as standardized dataset, which contains all standardized observations of airborne remote sensing data, ground-based in-situ data, and experimental analysis data;

[0090] M: Model parameter set, containing three core parameters: first, the true value of organic carbon content (denoted as C). trueThe first variable is the objective and true value of the organic carbon storage in the sediments of the target lake / reservoir, which is the core variable that the model needs to estimate; the second is the systematic bias of the data source (denoted as B). i (i = 1, 2, 3) correspond to the air-based, ground-based, and experimental data sources, respectively, characterizing the degree of systematic deviation of each data source from the true value due to equipment errors, environmental interference, etc.; the third is the random error of the data source (denoted as ε). i,j ), representing the random fluctuation of the j-th data point in the i-th data source, which follows a normal distribution. This is the measure of uncertainty for that data point;

[0091] P(M|D): Posterior probability distribution, which represents the probability distribution of the model parameter set M after obtaining the observation dataset D. It is the core result of the model output and reflects the confidence range of the parameters.

[0092] P(D|M): Likelihood function, representing the probability of observing dataset D given a model parameter set M, defined jointly by the observation error distribution models of each data source; in this embodiment, the likelihood function is set according to the characteristics of each data source: the experimental analysis data has the highest accuracy, and its error distribution model is set as a narrow variance normal distribution; the ground-based in-situ data is the next most accurate, and is set as a medium variance normal distribution; the airborne remote sensing data is more affected by environmental interference, and is set as a wide variance normal distribution;

[0093] P(M): Prior probability distribution, representing the probability distribution assumptions about the model parameter set M based on historical data and domain knowledge before obtaining the observed dataset D; for example, the true value of organic carbon content C. true The prior distribution is set as a normal distribution based on the measured values ​​of the target lake / reservoir over the past 5 years. Where μ hist This is the historical average. Historical variance; data source systematic bias B i The prior distribution is set based on the equipment calibration report, such as the prior distribution of the system bias of a satellite sensor. The variance of equipment calibration error;

[0094] P(D): Evidence factor, a normalization constant. When calculating the posterior probability of a specific parameter, its value does not affect the relative relationship of the parameter distribution, but only ensures that the integral value of the posterior probability distribution is 1. In this embodiment, its value is calculated by numerical integration.

[0095] The role and realization of latent variables: Data source reliability latent variable (denoted as W) iThe weights of different data sources in the model are dynamically adjusted as follows: For experimental analysis data, which has high reliability, the value range of W3 is set to [0.8, 1.0]; for ground-based in-situ data, the value range of W2 is [0.5, 0.8]; and for airborne remote sensing data, the value range of W1 is [0.2, 0.5]. During the iterative calculation process, the weights of W3 and W1 are dynamically adjusted according to the degree of fit between the observed data and the prior assumptions. i The specific value is determined by factors such as the degree of fit between the airborne remote sensing data and the experimental data during a certain period. For example, if the fit between the airborne remote sensing data and the experimental data is high during a certain period, W1 is increased, and vice versa.

[0096] Spatiotemporal correlation latent variables (denoted as R) s,t This is used to transmit uncertain information in the spatiotemporal dimension: for spatially adjacent data points (such as two in-situ sensing nodes less than 1km apart), R is set. s,t The spatial correlation coefficient is 0.8-0.9, indicating that its uncertainty is strongly correlated; a decrease in the uncertainty of one data point will lead to a decrease in the uncertainty of adjacent points. For data points that are continuous in time (such as in-situ data with an interval of 1 hour), let R... s,t The temporal correlation coefficient is 0.7-0.8, indicating that its uncertainty has a moderate correlation. Through this latent variable, the model can transmit the uncertainty information of high-confidence data (such as experimental data) to the surrounding spatiotemporal data points, thereby improving the overall quantification accuracy.

[0097] Uncertainty metric calculation and output: The posterior probability distribution P(M|D) is sampled and calculated using the Markov Chain Monte Carlo (MCMC) algorithm, with the number of iterations set to 10,000 to ensure the convergence of the sampling results; during the iteration process, the Gelman-Rubin statistic of the parameters is monitored in real time, and when the statistic is less than 1.1, the sampling is considered to have converged and the iteration is stopped.

[0098] After the iteration converges, the random error variance of each data point in the model parameter set is extracted. This variance is a measure of the uncertainty of the corresponding data point: The smaller the value, the higher the reliability of the data point; conversely, the larger the value, the lower the reliability. The standardized observations of each data point are then compared with their corresponding uncertainty measures. The data is correlated to generate a dataset with an accompanying uncertainty metric, which serves as input to the multi-source data fusion processing module.

[0099] Implementation of the multi-source data fusion processing module: The multi-source data fusion processing module adopts a spatiotemporal adaptive fusion algorithm based on uncertainty weighting to perform optimal fitting on the dataset with attached uncertainty measurement, generating a spatiotemporally continuous organic carbon storage fusion data product, and simultaneously calculating the uncertainty of the fusion result. The specific implementation process is as follows:

[0100] The core formula and parameter definition of the fusion algorithm: In this embodiment, the core of the spatiotemporal adaptive fusion algorithm is to assign adaptive weights to each data point. The weight calculation takes into account the uncertainty of the data, the spatiotemporal distance, and the reliability of the data source. Finally, multi-source data fusion is achieved through weighted averaging. The core mathematical formula is as follows:

[0101] Calculation of the combined estimate of organic carbon reserves

[0102]

[0103] The definitions and physical meanings of each parameter are as follows:

[0104] The fused estimate of organic carbon storage at spatial location s and time point t is the core indicator of the fused product;

[0105] n: Total number of data source types. In this example, n = 3 (airborne, ground-based, experimental).

[0106] m: The total number of data points within the set spatiotemporal neighborhood window. The window size is set according to the data density. The spatial window range is 1km×1km, and the temporal window range is 24 hours, ensuring that the window contains enough data points to support fusion calculation.

[0107] Z i,j (s,t): The standardized observation of the j-th data point in the i-th data source at location s and time t. If the original location or time of the data point does not coincide with (s,t), the observation of the point at (s,t) is obtained by spatiotemporal interpolation.

[0108] λ i,j (s,t): The adaptive weight of the j-th data point in the i-th data source at (s,t) is the core of the fusion algorithm and directly determines the contribution of each data point to the fusion result.

[0109] Adaptive weight calculation

[0110]

[0111] The definitions and physical meanings of each parameter are as follows:

[0112] ω i The overall reliability coefficient of the i-th type of data source is set based on prior knowledge. In this embodiment, ω1 = 0.3 (empty basis), ω2 = 0.5 (foundation), and ω3 = 0.8 (experiment), which is consistent with the value logic of the latent variables of data source reliability.

[0113] K s (‖ss j||: Spatial kernel function, characterizing the influence of the spatial distance between the data point and the fusion target point (s,t) on the weights. This embodiment uses the Gaussian kernel function. Where d = ||ss j || represents spatial distance, L s For spatial scale parameters, L is set according to the heterogeneity of the lake and reservoir sedimentary environment in this embodiment. s =500m; the smaller the spatial distance, the more K s The larger the (d) value, the higher the weight of the data point;

[0114] K t (‖tt j ||): The time kernel function characterizes the impact of the time distance between the data point and the fusion target point (s,t) on the weights. It also uses a Gaussian kernel function. Where τ=‖tt j ‖ represents the time distance, L t For the time scale parameter, L in this embodiment t =12h; the smaller the time interval, the more K t The larger the (τ) value, the higher the weight of the data point;

[0115] The uncertainty measure of the j-th data point in the i-th data source at (s,t) is directly provided by the uncertainty quantification module. Its value is inversely proportional to the adaptive weight, that is, the smaller the uncertainty, the higher the weight.

[0116] Fusion calculation process: First, set the spatiotemporal resolution of the fused product: the spatial resolution is consistent with the remote sensing data grid, and the temporal resolution is set to 1 day to ensure that the product can reflect spatial continuity while capturing dynamic changes at the daily scale; Second, traverse all fusion target points (s,t), and perform the following steps for each target point:

[0117] Determine the spatiotemporal neighborhood window: With (s,t) as the center, delineate a 1km×1km spatial window and a 24-hour time window, and extract all data points within the window (covering three types of data sources: airborne, ground-based, and experimental).

[0118] Calculate the adaptive weight λ for each data point i,j (s,t): According to the weighting formula above, substituting ω i Spatial distance d, temporal distance τ, and uncertainty measure Calculate the weight of each data point;

[0119] Calculate the fusion estimate According to the fusion estimation formula, the sum of "weight × observation value" of all data points in the window is obtained by dividing by the total weight to obtain the fusion estimate of organic carbon storage at (s,t).

[0120] Uncertainty in the fusion result: Based on the law of error propagation, the formula for calculating the uncertainty (variance) of the fusion result is derived:

[0121]

[0122] in

[0123]

[0124] That is, the proportion of adaptive weights to total weights, characterizing the sensitivity of the data point's contribution to the fusion result; the fusion estimate is calculated using this formula. Corresponding uncertainty To quantify the uncertainty of the fusion results.

[0125] After completing the fusion calculations for all target points, a spatiotemporally continuous fused organic carbon storage data product of lake and reservoir sediments is generated, which includes the fused organic carbon storage estimate for each spatiotemporal point. and its uncertainty Stored in the system database for subsequent output and verification.

[0126] Implementation of the Verification and Feedback Module: The verification and feedback module evaluates the accuracy of the fused data product using independent measured data, and optimizes the model parameters of the uncertainty quantification module based on the evaluation results, forming a closed loop for improving system accuracy. The specific implementation process is as follows:

[0127] Acquisition of independent validation data: Independent sampling and analysis are conducted every six months during the monitoring period. Sampling points are selected from areas not previously involved in data collection (including air sample, ground sample, and experimental samples) to ensure the independence of the validation data. The sampling and analysis process is consistent with the experimental data analysis unit, obtaining independent measured values ​​of total organic carbon content in sediments, which serve as the validation baseline (denoted as Z). ref The spatiotemporal locations of the verification data correspond one-to-one with the spatiotemporal locations of the fused data products to ensure the effectiveness of the comparison.

[0128] Calculation of accuracy assessment index: The estimated value of organic carbon storage from the fused data products. Compared with the verification benchmark Z ref Comparisons were made at the same spatiotemporal locations, and three core accuracy evaluation indicators were calculated:

[0129] Root Mean Square Error (RMSE): Characterizes the overall deviation between the fused estimate and the true value, and is expressed by the formula:

[0130]

[0131] Where N is the total number of validation data points, Z k Z is the fusion estimate for the k-th validation point.ref,k This represents the measured value at the k-th verification point; the smaller the RMSE value, the higher the accuracy.

[0132] Mean Absolute Error (MAE): Characterizes the average bias of the fused estimates, and is expressed by the formula:

[0133]

[0134] The smaller the MAE value, the smaller the average deviation of the fusion result;

[0135] Coefficient of determination (R) 2 ): Characterizes the degree of linear correlation between the fusion estimate and the measured value, and the formula is:

[0136]

[0137] in, R is the mean of the measured values. 2 The closer the value is to 1, the stronger the correlation between the fusion result and the measured value, and the better the model fit.

[0138] In this embodiment, the calculated RMSE is 0.12 g / kg, MAE is 0.09 g / kg, and R... 2 The accuracy of 0.89 indicates that the accuracy of the fused data product meets the needs of monitoring organic carbon storage in lake and reservoir sediments, but there is still room for optimization.

[0139] Model parameter feedback optimization: Based on the accuracy evaluation results, an adjustment signal is generated and fed back to the hierarchical probability model of the uncertainty quantification module to iteratively optimize the model parameters.

[0140] If the fusion result of a certain type of data source deviates significantly from the measured value (e.g., the fusion deviation ratio of the space-based remote sensing data exceeds 40%), then the prior probability distribution parameters of that data source should be adjusted. For example, increase the prior variance of the systematic bias B1 of the space-based data source to reduce its prior confidence; at the same time, adjust the error distribution model of the likelihood function to increase the likelihood variance of the space-based data so that the model reduces the weight of that data source in subsequent calculations.

[0141] If the uncertainty of the fusion result does not match the actual deviation (e.g.) If the spatial correlation coefficient is small but the actual deviation is large, then adjust the parameters of the potential variables of spatiotemporal correlation: for example, reduce the spatial correlation coefficient to reduce the uncertainty transmission from high-confidence data to low-confidence regions, so that the uncertainty of the fusion result is more in line with the actual deviation.

[0142] After parameter adjustment, the optimized model parameters are substituted into the uncertainty quantification module, and the uncertainty quantification and multi-source data fusion process is re-executed to calculate the new fused data product. The accuracy is then evaluated again using independent validation data until the accuracy index reaches a preset threshold (e.g., RMSE ≤ 0.10 g / kg, R...). 2 (≥0.92), complete the iterative optimization of model parameters.

[0143] In summary, this embodiment demonstrates the technical process of multi-source data acquisition, preprocessing standardization, uncertainty quantification, spatiotemporal adaptive fusion, and verification feedback optimization by implementing a dynamic monitoring and accounting system for organic carbon storage in typical freshwater lakes and reservoirs.

[0144] Example 2

[0145] like Figure 2 As shown in this embodiment, the specific working steps of a dynamic monitoring and accounting system for organic carbon storage in lake and reservoir sediments during actual operation are described in detail. This system achieves dynamic monitoring and accurate accounting of organic carbon storage in lake and reservoir sediments through multi-source data acquisition, preprocessing, uncertainty quantification, data fusion, and verification feedback. The specific steps are as follows:

[0146] S100: Synchronous acquisition of multi-source data:

[0147] The following three types of data are acquired simultaneously through airborne remote sensing platforms (satellites, drones), ground-based in-situ sensor networks deployed in lakes and reservoirs, and regular field sampling and laboratory analysis:

[0148] Airborne remote sensing inversion data: including parameters related to organic carbon in water bodies (such as DOC, POC, CDOM, etc.);

[0149] In-situ ground-based sensing data includes hydrological and water quality parameters such as water temperature, pH, dissolved oxygen, and spectral data.

[0150] Experimental analysis data includes the physicochemical properties of sediment samples, such as dry density, water content, and total organic carbon content.

[0151] S200: Data Preprocessing and Standardization

[0152] The raw data collected was processed as follows:

[0153] Spatiotemporal reference unification: All data are converted to a unified spatial coordinate system (such as CGCS2000) and time reference (UTC), and point data is converted into area data through interpolation methods;

[0154] Quality control: Identify and remove outlier data based on statistical methods (such as the 3σ criterion) and domain knowledge thresholds;

[0155] Standardization processing: Normalize the valid data to eliminate differences in units and generate a standardized dataset.

[0156] S300: Quantification of Uncertainty

[0157] Using a hierarchical probabilistic model based on Bayes' theorem, the uncertainty of each data point in a standardized dataset is quantified, and a dataset with an accompanying uncertainty metric is output. This model considers factors such as data source reliability and spatiotemporal correlation to dynamically evaluate the credibility of each data point.

[0158] S400: Multi-source data fusion processing

[0159] An uncertainty-weighted spatiotemporal adaptive fusion algorithm is employed to fuse multi-source data with accompanying uncertainties, generating a spatiotemporally continuous organic carbon storage fusion data product. This product includes an estimated organic carbon storage value and its uncertainty index for each spatiotemporal location.

[0160] S500: Data Product Output and Storage

[0161] The merged data products are output to the system database and stored in standard geographic information data formats (such as NetCDF and GeoTIFF) to support subsequent queries, visualization, and analysis.

[0162] S600: Verification and Feedback Optimization:

[0163] Introduce independent measured data that was not involved in the fusion process to evaluate the accuracy of the fused data product (calculate RMSE, MAE, R...). 2 The system uses indicators such as [list of indicators] and feeds the evaluation results back to the uncertainty quantification module to iteratively optimize the model parameters and improve the overall accuracy of the system.

[0164] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A dynamic monitoring and accounting system for organic carbon storage in lake and reservoir sediments, characterized in that, include: The multi-source data acquisition module is used to acquire optical data retrieved from airborne remote sensing, hydrological and water quality parameter data from ground-based in-situ sensors, and physicochemical property data from sediment experiments. The data preprocessing and standardization module is used to perform spatiotemporal benchmark unification, quality control and standardization processing on the heterogeneous data acquired by the multi-source data acquisition module, and generate a spatiotemporally aligned standardized dataset. The uncertainty quantification module, connected to the data preprocessing and standardization module, is used to quantify the uncertainty of various types of data in the standardized dataset based on a probabilistic graphical model, and output the uncertainty measure of each data point. The multi-source data fusion processing module is connected to the uncertainty quantification module. It is used to receive the standardized dataset and its corresponding uncertainty measure, and to use a spatiotemporal adaptive fusion algorithm based on uncertainty weighting to perform optimal fitting on the multi-source data, thereby generating a spatiotemporally continuous fusion data product of lake and reservoir sediment organic carbon storage. The system output module is connected to the multi-source data fusion processing module and is used to output and store the fused data products.

2. The system for dynamic monitoring and accounting of organic carbon storage in lake and reservoir sediments according to claim 1, characterized in that, The uncertainty quantification module uses a probabilistic graphical model, which is a hierarchical probabilistic model based on Bayes' theorem. This model quantifies the uncertainty of multi-source heterogeneous data in a structured way by introducing latent variables of data source reliability and spatiotemporal correlation. The mathematical expression of this model is the formula for calculating the posterior probability distribution: Wherein, P(M|D) represents the posterior probability distribution of the model parameter set M under the condition of obtaining the observed dataset D, the model parameter set M includes the true value of organic carbon content, the systematic bias of the data source and random error, P(D|M) represents the likelihood function, which characterizes the probability of observing dataset D under the model parameter set M. This likelihood function is jointly defined by the observation error distribution model of each data source, P(M) represents the prior probability distribution of the model parameter set M, which is determined based on historical data and domain knowledge and is used to introduce constraints on the reasonable range of parameter values, and P(D) represents the evidence factor, which is regarded as a normalization constant when calculating the posterior probability of a specific parameter; The hierarchical probability model dynamically adjusts the weights of different data sources in the prior definition and likelihood function through the data source reliability latent variable, and transmits uncertainty information in the spatiotemporal dimension through the spatiotemporal correlation latent variable.

3. The system for dynamic monitoring and accounting of organic carbon storage in lake and reservoir sediments according to claim 1, characterized in that, The spatiotemporal adaptive fusion algorithm based on uncertainty weighting used in the multi-source data fusion processing module operates as follows: Assign an adaptive weight to each standardized data point that is inversely proportional to its uncertainty measure, and perform a weighted average within a spatiotemporal window, while also considering the spatiotemporal correlation between data points. The mathematical expression of this algorithm is as follows: in, Z represents the fused estimate of organic carbon storage at spatial location s and time point t, where n represents the total number of data source types, m represents the total number of data points within the defined spatiotemporal neighborhood window, and Z... i,j (s,t) represents the standardized observation of the j-th data point in the i-th data source at location s and time t, where λ i,j (s,t) represents the adaptive weight of the j-th data point in the i-th data source when participating in the fusion calculation, ω i K represents the overall credibility coefficient assigned to the i-th type of data source based on prior knowledge. s (‖ss j ‖) represents the spatial kernel function, used to characterize the spatial distance between the target point and the fusion target point as ‖ss j The weight of a data point decreases with distance according to the law of K. t (‖tt j ‖) represents the time kernel function, used to characterize the time distance between the target point and the fusion point as ‖tt. j The decay pattern of the weight of the data points in the equation with increasing time interval. This represents the uncertainty metric value of the j-th data point in the i-th type of data source provided by the uncertainty quantification module at location s and time t.

4. The system for dynamic monitoring and accounting of organic carbon storage in lake and reservoir sediments according to claim 1, characterized in that, The multi-source data acquisition module specifically includes: The airborne remote sensing data unit is used to retrieve data on dissolved organic carbon concentration, particulate organic carbon concentration, colored soluble organic matter concentration, water surface temperature, and chlorophyll a concentration of a large area of ​​lake and reservoir surface water by using multispectral and hyperspectral sensors mounted on any platform such as satellites or drones. The ground-based in-situ sensing data unit is used to periodically and in real-time collect water level, water temperature, transparency, pH value, conductivity, dissolved oxygen concentration and underwater spectral data of monitoring points through a network of intelligent sensing nodes deployed in the lake and reservoir water. The experimental analysis data unit is used to collect sediment column or surface samples from lakes and reservoirs, and then conduct laboratory physicochemical analysis to obtain data on the dry density, water content, grain size distribution, total organic carbon content, and stable isotope ratios of the sediments.

5. The system for dynamic monitoring and accounting of organic carbon storage in lake and reservoir sediments according to claim 1, characterized in that, The execution flow of the data preprocessing and standardization module includes: The spatiotemporal reference unification submodule is used to convert all data acquired by the multi-source data acquisition module to a unified spatial coordinate system and time reference, and to use the Kriging spatial interpolation algorithm and time series interpolation algorithm to convert the point in-situ data and experimental data into area data that matches the remote sensing data grid, fill in the missing data periods, and generate a preliminary spatiotemporally aligned dataset. The quality control and outlier removal submodule is connected to the spatiotemporal benchmark unification submodule. It is used to perform anomaly detection on each data channel in the preliminary dataset based on the statistical 3σ criterion and knowledge thresholds in the field of lake and reservoir water environment, and remove outlier data points that deviate significantly from the normal range. The standardization processing submodule, connected to the quality control and outlier removal submodule, is used to normalize the dataset after quality control, eliminate the differences caused by different physical dimensions, and convert the numerical range of all datasets to the same interval, generating the final standardized dataset input to the uncertainty quantification module.

6. The system for dynamic monitoring and accounting of organic carbon storage in lake and reservoir sediments according to claim 1, characterized in that, The uncertainty quantification module works in conjunction with the multi-source data fusion processing module to ensure the complete transfer of uncertainty information from quantification to fusion. The uncertainty quantification module outputs the uncertainty measure. It is directly used as an important input parameter for the fusion algorithm in the multi-source data fusion processing module; After completing the fusion calculation, the multi-source data fusion processing module also performs uncertainty propagation calculations on the fusion results. Its output includes not only the fusion estimate of organic carbon storage. Simultaneously output the uncertainty of this estimate. The formula for calculating the uncertainty of the fusion result is derived based on the error propagation law: in, Indicates the fusion estimate The variance at position s and time t is used to measure its uncertainty; Indicates the fusion estimate Relative to the j-th observation Z in the i-th data source i,j The partial derivative of (s,t), This represents the uncertainty measure of the j-th observation in the i-th data source.

7. The system for dynamic monitoring and accounting of organic carbon storage in lake and reservoir sediments according to claim 1, characterized in that, It also includes a verification and feedback module that connects to the system output module; The verification and feedback module includes: An independent verification data import unit is used to import measured organic carbon storage data of lake and reservoir sediments that have not participated in the fusion calculation process of the multi-source data fusion processing module, as a verification benchmark. The accuracy evaluation unit, connected to the independent verification data import unit, is used to compare the fused data product generated by the system output module with the verification benchmark at the same spatiotemporal location, and calculate multiple accuracy evaluation indicators such as root mean square error, mean absolute error, and coefficient of determination. The model parameter feedback optimization unit is connected to the accuracy evaluation unit and is used to generate an adjustment signal based on the accuracy evaluation result, and feed the adjustment signal back to the probabilistic graphical model in the uncertainty quantification module. The prior distribution parameters or likelihood function parameters in the probabilistic graphical model are iteratively optimized.

8. A method for dynamic monitoring and calculation of organic carbon storage in lake and reservoir sediments, applicable to the dynamic monitoring and calculation system for organic carbon storage in lake and reservoir sediments as described in any one of claims 1-7, characterized in that, The specific steps of this method are as follows: S100. The multi-source data acquisition module synchronously acquires airborne remote sensing inversion optical data, ground-based in-situ sensor hydrological and water quality parameter data, and sediment experimental analysis physicochemical property data. S200. The data preprocessing and standardization module performs a spatiotemporal reference unification operation on the acquired airborne remote sensing inversion optical data, the ground-based in-situ sensor hydrological and water quality parameter data, and the sediment experimental analysis physicochemical property data, converting all data to a unified spatial coordinate system and time reference. Then, a quality control operation is performed to remove abnormal data points. Finally, standardization processing is performed to eliminate dimensional differences and generate a standardized dataset. S300. The standardized dataset is received through the uncertainty quantification module, and the hierarchical probability model based on Bayes' theorem is used to analyze various types of data in the standardized dataset, quantify the uncertainty of each data point, and output a dataset with an uncertainty metric. S400. The multi-source data fusion processing module receives the dataset with uncertainty measurement and uses a spatiotemporal adaptive fusion algorithm based on uncertainty weighting to perform fusion calculation on the multi-source data, generating a spatiotemporal continuous data product that simultaneously contains the fusion estimate of organic carbon storage and its uncertainty index. S500, The spatiotemporal continuous data product is output and stored through the system output module; S600. The verification and feedback module uses independent measured data to evaluate the accuracy of the spatiotemporal continuous data product, and generates a feedback signal based on the evaluation results to optimize the model parameters in the uncertainty quantification module.

Images