Basin carbon reserve migration assessment method and system based on tree growth model
By employing a watershed carbon storage assessment method based on tree growth models, combined with multi-source data and a self-attention mechanism, the problems of insufficient assessment accuracy and poor environmental adaptability in existing technologies are solved. This enables accurate assessment and prediction of watershed carbon storage transport, supporting watershed ecological protection and carbon emission reduction decisions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YELLOW RIVER INST OF HYDRAULIC RES YELLOW RIVER CONSERVANCY COMMISSION
- Filing Date
- 2026-02-10
- Publication Date
- 2026-05-26
AI Technical Summary
Existing watershed carbon storage assessment methods neglect the dynamic relationship between tree growth and carbon storage transport, have insufficient accuracy in multi-source data fusion, cannot adapt to different environmental conditions, and lack carbon storage transport prediction models for water resource allocation scenarios, resulting in insufficient assessment accuracy and difficulty in providing precise decision support.
A watershed carbon storage assessment method based on a tree growth model is adopted. By preprocessing multi-source data and dynamically allocating weights through a self-attention mechanism, an assessment model integrating tree growth, carbon storage accounting, and runoff-driven carbon transport is constructed. Combined with satellite remote sensing data, carbon storage assessment and prediction in water resource scheduling scenarios are realized.
It has improved the scientific rigor and accuracy of carbon storage and transport assessment, enhanced the model's adaptability to complex watershed environments, and enabled precise predictions in water resource allocation scenarios, providing technical support for watershed ecological protection and carbon emission reduction decision-making.
Smart Images
Figure CN122087340A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of carbon cycle monitoring technology, and more specifically to a method and system for assessing watershed carbon storage and transport based on a tree growth model. Background Technology
[0002] Tree growth and death have a significant impact on watershed carbon cycling, which in turn directly influences global climate change. Domestic and international watershed carbon transport assessment methods primarily employ sample plot inventory and volumetric methods. These methods often require on-site sampling and measurement, incurring substantial costs, and are difficult to implement long-term simulations of watershed carbon transport. From a methodological perspective, the InVEST model is an important method for carbon transport assessment, while land use simulation models such as CA-Markov, FLUS, MCE, GeoSOS, and PLUS are widely used multi-scenario prediction models. These models typically achieve accuracy in assessing watershed carbon transport through extensive and repetitive parameter adjustments, often requiring researchers with deep professional knowledge and model-building experience. They cannot provide accurate and rapid assessments of watershed carbon transport, and these models also lack the ability to integrate multi-source data.
[0003] With the deepening implementation of the dual-carbon strategy, the dynamic changes and transport processes of carbon storage in watersheds, as an important component of terrestrial ecosystems, have become a research hotspot in the field of ecological environment. Watershed carbon storage is not only affected by biogeochemical processes such as vegetation growth and soil respiration, but also closely related to runoff changes caused by water resource allocation. Trees, as the main carbon sink carriers in watershed ecosystems, directly determine the efficiency of carbon fixation and accumulation through their growth process. Tree rings, as natural recorders of tree growth, can objectively reflect the coupling relationship between tree growth and environmental factors on a long-term scale.
[0004] However, existing methods for assessing carbon storage in watersheds have several limitations: First, most methods ignore the dynamic relationship between tree growth and carbon transport, estimating carbon storage solely through static biomass accounting, making it difficult to quantify the impact of changes in tree growth rate on carbon transport; second, multi-source data fusion employs a fixed-weight strategy, which cannot adapt to the dynamic changes in data importance under different environmental conditions, resulting in insufficient data fusion accuracy; third, carbon transport models often focus on single environmental factors, failing to fully consider the carbon migration and diffusion mechanisms driven by runoff, and their complexity is too high, making it difficult to achieve rapid assessment by combining satellite remote sensing data; fourth, carbon transport prediction under unified water resource management scenarios lacks targeted model support, failing to provide accurate basis for watershed ecological protection and carbon emission reduction decisions.
[0005] Therefore, how to propose a watershed carbon storage and transport assessment method and system based on tree growth models to overcome the shortcomings of existing technologies is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0006] In view of this, the present invention provides a method and system for assessing watershed carbon transport based on a tree growth model. This method couples the tree growth process, accurately integrates multi-source data, and quantifies the assessment of runoff-driven carbon transport, enabling efficient and accurate assessment and prediction of watershed carbon transport in water resource allocation scenarios. This addresses the problems of insufficient assessment accuracy, difficulty in quantifying coupling relationships, and inability to adapt to water resource allocation scenarios in existing technologies.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: A watershed carbon storage and transport assessment method based on a tree growth model includes: Collect multi-source data, preprocess the multi-source data, and obtain comprehensive characteristic data of factors affecting watershed carbon storage and transport; Construct a watershed carbon storage and transport model that integrates tree growth model; The comprehensive feature data is divided into training set, validation set and test set, and input into the watershed carbon storage and transport model for training. The model parameters are optimized by comparing the learning loss function and the elastic weight solidification method to obtain the optimal evaluation model. By acquiring real-time satellite remote sensing data and water resource scheduling scheme parameters, preprocessing and extracting their features, and then inputting them into the optimal evaluation model, the evaluation and prediction results of the spatial distribution, transport flux and changing trend of watershed carbon storage under unified water resource scheduling are obtained.
[0008] Optionally, the multi-source data includes satellite remote sensing data, tree growth-related data, runoff data, and biomass data. Among them, the satellite remote sensing data covers vegetation index, surface temperature, and topographic information; the tree growth-related data includes tree ring width, tree height, diameter at breast height, and growth rate; the runoff data includes runoff volume and flow velocity; and the biomass data includes measured and estimated values of biomass of trees, shrubs, and herbs.
[0009] Optionally, the multi-source data is preprocessed, including: noise reduction, outlier removal, and normalization, to obtain standardized data; the standardized data is spatiotemporally aligned to eliminate time and spatial deviations between different data sources; the spatiotemporally aligned standardized data is input into a self-attention mechanism to dynamically allocate weight coefficients for each type of data; and the different types of data are weighted and concatenated according to the weight coefficients.
[0010] Optionally, the watershed carbon storage transport model includes an input layer, a tree growth simulation module, a carbon storage accounting module, a runoff-driven carbon transport module, a dynamic correction module, and an output layer. The input layer is used to receive comprehensive feature data; The tree growth simulation module constructs a growth prediction model based on tree ring data and environmental factors to simulate the process of tree biomass accumulation. The carbon storage calculation module calculates the total carbon storage of the watershed based on tree biomass and other vegetation biomass data, combined with the carbon conversion coefficient. The runoff-driven carbon transport module constructs a carbon migration and diffusion model based on runoff data to quantify the impact of runoff on the spatial transport of carbon storage. The dynamic correction module uses a learnable error compensation function to correct model prediction biases in real time; the output layer is used to output carbon storage migration assessment and prediction results.
[0011] Optionally, the tree growth simulation module constructs a growth prediction model based on tree-ring data and environmental factors to simulate the tree biomass accumulation process, including: A tree growth prediction model is constructed using formulas: ; in, for Tree biomass at all times This is the initial biomass. for Real-time vegetation index for Standardized value of surface temperature at any given time. for Moisture stress coefficient at any time , , For model parameters, Let be the width of the tree rings in year i. This represents the average width of tree rings. The number of years observed.
[0012] Optionally, the carbon storage calculation module calculates the total carbon storage of the watershed based on tree biomass and other vegetation biomass data, combined with the carbon conversion coefficient, including: Calculate the total carbon storage of the watershed using the following formula: ; in, The total carbon storage of the basin. The number of watershed partitions, For the first Total tree biomass in the zone The carbon conversion coefficient of trees. For the first Total biomass of other vegetation in the zone For the first Carbon conversion coefficient of other vegetation in the zone For the first Zonal soil carbon storage.
[0013] Optionally, the runoff-driven carbon transport module constructs a carbon migration and diffusion model based on runoff data to quantify the impact of runoff on the spatial transport of carbon storage, including: A carbon migration and diffusion model is constructed using formulas: ; in, Coordinates at time t carbon concentration at that location For time step, The diffusion coefficient is... This is the runoff velocity vector. for Time coordinates The amount of carbon concentration change caused by tree growth.
[0014] Optionally, the dynamic correction module employs a learnable error compensation function to correct model prediction bias in real time, including: The error compensation value is calculated using the formula: ; in, for Time coordinates Error compensation value at the location, To predict carbon concentration for the model, To measure the carbon concentration, , These are learnable parameters; The corrected carbon concentration is: .
[0015] Optionally, the step of optimizing model parameters by comparing the learning loss function and the elastic weight solidification method to obtain the optimal evaluation model includes: The comparative learning loss function is calculated using the following formula: ; in, To compare the learning loss function, Cosine similarity for positive sample pairs. For the first Cosine similarity of negative sample pairs For temperature parameters, This represents the number of negative sample pairs. The elastic weight solidification loss function is calculated using the following formula: ; in, For the elastic weight solidification loss function, This is a set of indexes for key parameters of the model. For the corresponding parameters in the Fisher information matrix The value, The value of the current training parameter i, For parameters The optimal value in historical training; The total loss function is: ; in, For mean square error loss, , , The loss weights are used to minimize the total loss function through backpropagation, thereby optimizing the model parameters.
[0016] Optionally, a watershed carbon storage and transport assessment system based on a tree growth model includes: The data acquisition module is used to collect data from multiple sources. The data processing module is used to preprocess the multi-source data to obtain comprehensive characteristic data of the factors affecting watershed carbon storage and transport. The model building module is used to build a watershed carbon storage and transport model that integrates tree growth model; The model training module is used to divide the comprehensive feature data into training set, validation set and test set, input them into the watershed carbon storage and transport model for training, and optimize the model parameters by comparing the learning loss function and the elastic weight solidification method to obtain the optimal evaluation model. The assessment and prediction module is used to acquire real-time satellite remote sensing data and water resource scheduling scheme parameters. After preprocessing and feature extraction, the data is input into the optimal assessment model to obtain the assessment and prediction results of the spatial distribution, transport flux and changing trend of watershed carbon storage under unified water resource scheduling.
[0017] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method and system for assessing watershed carbon storage and transport based on a tree growth model, which has the following beneficial effects: This invention proposes a watershed carbon transport assessment method based on a tree growth model, comprising: collecting multi-source data; preprocessing the multi-source data to obtain comprehensive feature data of factors influencing watershed carbon transport; constructing a watershed carbon transport model fused with a tree growth model; dividing the comprehensive feature data into training, validation, and test sets, inputting them into the watershed carbon transport model for training, optimizing model parameters through a comparative learning loss function and an elastic weight solidification method to obtain an optimal assessment model; acquiring real-time satellite remote sensing data and water resource scheduling scheme parameters, preprocessing and extracting features from them, and inputting them into the optimal assessment model to obtain assessment and prediction results of the spatial distribution, transport flux, and changing trends of watershed carbon under unified water resource scheduling. This invention, by fusing a tree growth model and a carbon transport model, quantifies the coupling relationship between tree growth and carbon transport, solves the problem of traditional methods neglecting the dynamic influence of tree growth, and improves the scientific rigor and accuracy of carbon transport assessment. A self-attention mechanism is employed to dynamically assign weights to multi-source data, effectively adapting to changes in data importance under different environmental conditions, improving data fusion accuracy, and enhancing the model's adaptability to complex watershed environments. The constructed runoff-driven carbon transport module can quantify the impact of runoff changes caused by water resource allocation on carbon transport, achieving accurate prediction of carbon storage and transport under unified water resource allocation scenarios, providing targeted technical support for the implementation of the watershed's dual-carbon strategy. The model uses a dynamic correction module and an optimized loss function to reduce systematic errors and overfitting risks. Furthermore, its lightweight design enables rapid evaluation by combining satellite remote sensing data, meeting the real-time requirements of practical applications. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0019] Figure 1 This invention provides a schematic flowchart of a watershed carbon storage and transport assessment method based on a tree growth model. Detailed Implementation
[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0021] This invention discloses a method for assessing watershed carbon storage and transport based on a tree growth model, such as... Figure 1 As shown, it includes: Collect multi-source data, preprocess the multi-source data, and obtain comprehensive characteristic data of factors affecting watershed carbon storage and transport; Construct a watershed carbon storage and transport model that integrates tree growth model; The comprehensive feature data is divided into training set, validation set and test set, and input into the watershed carbon storage and transport model for training. The model parameters are optimized by comparing the learning loss function and the elastic weight solidification method to obtain the optimal evaluation model. By acquiring real-time satellite remote sensing data and water resource scheduling scheme parameters, preprocessing and extracting their features, and then inputting them into the optimal evaluation model, the evaluation and prediction results of the spatial distribution, transport flux and changing trend of watershed carbon storage under unified water resource scheduling are obtained.
[0022] Furthermore, the multi-source data includes satellite remote sensing data, tree growth-related data, runoff data, and biomass data. Among them, the satellite remote sensing data covers vegetation index, surface temperature, and topographic information; the tree growth-related data includes tree ring width, tree height, diameter at breast height, and growth rate; the runoff data includes runoff volume and flow velocity; and the biomass data includes measured and estimated values of tree, shrub, and herbaceous biomass.
[0023] Furthermore, the multi-source data is preprocessed, including denoising, outlier removal, and normalization, to obtain standardized data; the standardized data is spatiotemporally aligned to eliminate time and spatial deviations between different data sources; the spatiotemporally aligned standardized data is input into a self-attention mechanism to dynamically allocate weight coefficients for each type of data; and the different types of data are weighted and concatenated according to the weight coefficients.
[0024] In a specific implementation, the step of performing spatiotemporal alignment on standardized data, inputting the spatiotemporally aligned standardized data into a self-attention mechanism, and dynamically allocating weight coefficients for each type of data includes: Spatial location matching and time synchronization of multi-source data were achieved using a SLAM-based spatial calibration algorithm and a hardware-triggered timing synchronization method, with time deviation controlled within ±5ms and spatial registration error less than 1 pixel. Weighting coefficients for tree growth-related data were calculated using a formula. (1); in, These are the weighting coefficients for tree growth-related data. For activation function, , For learnable parameters, The average growth rate of trees, Standardized value for tree ring width; The weighting coefficients for runoff data are calculated using the following formula: (2); in, These are the weighting coefficients for runoff data. For activation function, Average runoff, This is the largest runoff in history. The average flow velocity, This is the highest flow rate in history. , These are learnable parameters; The weighting coefficients of satellite remote sensing data and biomass data were calculated using the similarity of the self-attention mechanism, and the sum of the weighting coefficients of each type of data was 1.
[0025] Furthermore, the watershed carbon storage and transport model includes an input layer, a tree growth simulation module, a carbon storage accounting module, a runoff-driven carbon transport module, a dynamic correction module, and an output layer. The input layer is used to receive comprehensive feature data; The tree growth simulation module constructs a growth prediction model based on tree ring data and environmental factors to simulate the process of tree biomass accumulation. The carbon storage calculation module calculates the total carbon storage of the watershed based on tree biomass and other vegetation biomass data, combined with the carbon conversion coefficient. The runoff-driven carbon transport module constructs a carbon migration and diffusion model based on runoff data to quantify the impact of runoff on the spatial transport of carbon storage. The dynamic correction module uses a learnable error compensation function to correct model prediction biases in real time; the output layer is used to output carbon storage migration assessment and prediction results.
[0026] Furthermore, the tree growth simulation module constructs a growth prediction model based on tree ring data and environmental factors to simulate the tree biomass accumulation process, including: A tree growth prediction model is constructed using formulas: (3); in, for Tree biomass at all times This is the initial biomass. for Real-time vegetation index for Standardized value of surface temperature at any given time. for Moisture stress coefficient at any time , , For model parameters, Let be the width of the tree rings in year i. This represents the average width of tree rings. The number of years observed.
[0027] Furthermore, the carbon storage calculation module calculates the total carbon storage of the watershed based on tree biomass and other vegetation biomass data, combined with the carbon conversion coefficient, including: Calculate the total carbon storage of the watershed using the following formula: (4); in, The total carbon storage of the basin. The number of watershed partitions, For the first Total tree biomass in the zone The carbon conversion coefficient of trees. For the first Total biomass of other vegetation in the zone For the first Carbon conversion coefficient of other vegetation in the zone For the first Zonal soil carbon storage.
[0028] Furthermore, the runoff-driven carbon transport module constructs a carbon migration and diffusion model based on runoff data to quantify the impact of runoff on the spatial transport of carbon storage, including: A carbon migration and diffusion model is constructed using formulas: (5); in, Coordinates at time t carbon concentration at that location For time step, The diffusion coefficient is... This is the runoff velocity vector. for Time coordinates The amount of carbon concentration change caused by tree growth.
[0029] Furthermore, the dynamic correction module employs a learnable error compensation function to correct model prediction biases in real time, including: The error compensation value is calculated using the formula: (6); in, for Time coordinates Error compensation value at the location, To predict carbon concentration for the model, To measure the carbon concentration, , These are learnable parameters; The corrected carbon concentration is: .
[0030] Furthermore, the optimization of model parameters by comparing the learning loss function and the elastic weight solidification method to obtain the optimal evaluation model includes: The comparative learning loss function is calculated using the following formula: (7); in, To compare the learning loss function, Cosine similarity for positive sample pairs. For the first Cosine similarity of negative sample pairs For temperature parameters, This represents the number of negative sample pairs. The elastic weight solidification loss function is calculated using the following formula: (8); in, For the elastic weight solidification loss function, This is a set of indexes for key parameters of the model. For the corresponding parameters in the Fisher information matrix The value, The value of the current training parameter i, For parameters The optimal value in historical training; The total loss function is: ; in, For mean square error loss, , , The loss weights are used to minimize the total loss function through backpropagation, thereby optimizing the model parameters.
[0031] In a specific implementation, a watershed carbon storage and transport assessment system based on a tree growth model includes: The data acquisition module is used to collect data from multiple sources. The data processing module is used to preprocess the multi-source data to obtain comprehensive characteristic data of the factors affecting watershed carbon storage and transport. The model building module is used to build a watershed carbon storage and transport model that integrates tree growth model; The model training module is used to divide the comprehensive feature data into training set, validation set and test set, input them into the watershed carbon storage and transport model for training, and optimize the model parameters by comparing the learning loss function and the elastic weight solidification method to obtain the optimal evaluation model. The assessment and prediction module is used to acquire real-time satellite remote sensing data and water resource scheduling scheme parameters. After preprocessing and feature extraction, the data is input into the optimal assessment model to obtain the assessment and prediction results of the spatial distribution, transport flux and changing trend of watershed carbon storage under unified water resource scheduling.
[0032] In a specific implementation, a watershed carbon storage and transport assessment method based on a tree growth model is provided to address the problems of insufficient assessment accuracy, difficulty in quantifying coupling relationships, and inability to adapt to water resource allocation scenarios in existing technologies. To achieve the above objectives, the embodiments of the present invention provide the following technical solutions: A watershed carbon storage and transport assessment method based on a tree growth model includes the following steps: S1. Collect multi-source data, including satellite remote sensing data, tree growth-related data, runoff data, and biomass data; S2. The multi-source data is preprocessed and spatiotemporally aligned. Weights are dynamically allocated and weighted fusion is performed through a self-attention mechanism to obtain comprehensive characteristic data of factors affecting watershed carbon storage and transport. S3. Construct a watershed carbon storage and transport model that integrates tree growth model; S4. Input the comprehensive feature data into the model for training and optimization to obtain the optimal evaluation model; S5. Based on real-time satellite remote sensing data and optimal evaluation models, the assessment and prediction of watershed carbon storage and transport under unified water resource scheduling are realized.
[0033] In a specific implementation, S1 collects multi-source data including: The comprehensiveness and accuracy of multi-source data are the foundation for reliable evaluation results. The specific types of data collected in this step are as follows: Satellite remote sensing data: acquired through remote sensing platforms such as Gaofen satellite and MODIS, including Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Land Surface Temperature (LST), Digital Elevation Model (DEM), land use type data, etc. The data resolution is selected according to the assessment accuracy requirements, with a spatial resolution range of 10m-1km and a temporal resolution of 1 day-1 month.
[0034] Tree growth-related data: obtained through field measurements and laboratory analysis, including tree ring sample collection and width measurement, tree height, diameter at breast height and crown width measurement data of trees of different ages, combined with long-term observation data from tree growth monitoring stations, to obtain data such as tree growth rate and annual ring growth trend; at the same time, growth characteristic parameters of major tree species in the study area were collected.
[0035] Runoff data: These are measured data from hydrological monitoring stations within the basin, including daily runoff, monthly runoff, runoff velocity, and runoff depth. Data such as water conservancy project scheduling plans are also collected to analyze runoff variation patterns.
[0036] Biomass data: The biomass of trees, shrubs and herbs was obtained by quadrat survey, and the spatial distribution data of biomass at the watershed scale was obtained by combining the data with the remote sensing inversion model; Soil carbon storage data was obtained by soil sampling and analysis, including parameters such as organic carbon content and bulk density of soil at different depths.
[0037] In a specific implementation, data preprocessing, spatiotemporal alignment, and feature fusion in S2 include: (1) First, preprocess the collected multi-source data: Satellite remote sensing data preprocessing: Radiometric correction and atmospheric correction are used to eliminate sensor errors and atmospheric effects. Maximum value synthesis (MVC) is used to reduce noise interference from clouds, fog and other factors. The DEM data is filled and smoothed, and topographic factors such as slope, aspect and water system are extracted.
[0038] Preprocessing of tree growth-related data: Cross-dating and standardization of tree ring width data to eliminate the influence of tree age effect and growth trend; outlier removal of measured data such as tree height and diameter at breast height, and interpolation to supplement missing data.
[0039] Runoff and biomass data preprocessing: Consistency checks are performed on runoff data to remove outliers caused by observation errors; measured biomass data are standardized and converted to a unified unit of tons per hectare; and data fusion is performed in conjunction with remote sensing inversion data to improve the spatial continuity of the data.
[0040] (2) Then, spatiotemporal alignment is performed: Spatial alignment: Based on the spatial resolution of satellite remote sensing data, bilinear interpolation or nearest neighbor interpolation is used to unify tree growth data, runoff data, and biomass data into the same spatial grid; based on the sensor extrinsic calibration matrix, the spatial position of satellite remote sensing data and ground measured data is accurately matched, and the spatial registration error is controlled within 1 pixel.
[0041] Time alignment: Hardware-triggered time synchronization and dynamic time warping (DTW) algorithms are used to achieve time synchronization of data with different time resolutions; for satellite remote sensing data, time aggregation is performed according to evaluation requirements to obtain monthly or quarterly time series data; tree growth data and runoff data are matched according to timestamps, and the time deviation is controlled within ±5ms.
[0042] (3) Finally, feature fusion is performed through a self-attention mechanism: The self-attention mechanism can dynamically allocate weights based on the relevance and importance of data features. The specific process is as follows: Constructing data feature matrices: Convert various types of spatiotemporally aligned data into feature vectors to form satellite remote sensing feature matrices, tree growth feature matrices, runoff feature matrices, and biomass feature matrices.
[0043] Dynamic weight calculation: The weight coefficients of tree growth-related data are calculated by formula (1), and the weight coefficients of runoff data are calculated by formula (2). The weights of satellite remote sensing data and biomass data are obtained by similarity calculation through self-attention mechanism, that is, the feature vector is mapped to a high-dimensional space by linear transformation, the attention score between different feature vectors is calculated, and the weight coefficients are obtained after Softmax normalization.
[0044] Weighted fusion: Based on the weight coefficients of various types of data, the feature matrix is weighted and spliced to obtain comprehensive feature data with unified dimensions. The feature dimensions after fusion are set according to the actual data situation, preferably 256-1024 dimensions.
[0045] In a specific implementation, the watershed carbon storage and transport model that integrates the tree growth model in S3 includes: The watershed carbon storage and transport model constructed in this invention integrates tree growth simulation, carbon storage calculation, runoff-driven carbon transport, and dynamic correction functions. The specific design of each module is as follows: Input layer: Receives comprehensive feature data obtained from S2, including various environmental factors, tree growth parameters, runoff parameters, etc. after data preprocessing, to provide comprehensive input information for the model.
[0046] Tree growth simulation module: Based on tree ring data and environmental factors (vegetation index, surface temperature, water conditions, etc.), a growth prediction model is constructed, and the tree biomass accumulation process is simulated through formula (3). This module can reflect the changes in the growth rate of trees under different environmental conditions and quantify the contribution of tree growth to carbon fixation.
[0047] Carbon storage accounting module: Based on the tree biomass data output by the tree growth simulation module, combined with the biomass data of other vegetation (shrubs, herbs), the total carbon storage of the watershed is calculated by formula (4); among which the tree carbon conversion coefficient adopts the default value of different tree species and is calibrated according to the measured data, and the soil carbon storage is calculated by soil organic carbon content, bulk density and soil layer thickness.
[0048] Runoff-driven carbon transport module: A carbon migration and diffusion model is constructed based on runoff data. The influence of runoff on the spatial transport of carbon storage is quantified by formula (5). This model considers factors such as runoff velocity, diffusion coefficient, and carbon concentration gradient. It can simulate the migration path and diffusion range of carbon in water bodies and reflect the driving effect of runoff changes caused by water resource scheduling on carbon transport.
[0049] Dynamic correction module: In order to improve the prediction accuracy of the model, a learnable error compensation function is adopted, and the error compensation value is calculated by formula (6) to correct the prediction results of the model in real time. This module can dynamically adjust the model parameters according to the deviation between the measured data and the predicted data to reduce the system error.
[0050] Output layer: Outputs spatial distribution data of watershed carbon storage, carbon transport flux, carbon transport pathways and trend prediction results. The data format supports seamless integration with GIS platforms, facilitating visualization and decision analysis.
[0051] In a specific implementation, model training optimization in S4 includes: The comprehensive feature data obtained from S2 is divided into training, validation, and test sets in a 7:2:1 ratio and input into the watershed carbon storage and transport model for training and optimization. Initialize model parameters: Use the Xavier initialization method to initialize the weight parameters in the model, set the learning rate to 0.001-0.01, and the number of iterations to 100-500.
[0052] Loss function design: The total loss function is the weighted sum of the contrastive learning loss function, the elastic weight solidification loss function and the mean squared error loss function. The loss of each part is calculated by formulas (7) and (8). The contrastive learning loss function can enhance the model's ability to distinguish similar samples, the elastic weight solidification loss function can prevent the model from catastrophic forgetting during training, and the mean squared error loss function can reduce the deviation between the predicted value and the measured value.
[0053] Model optimization: The Adam optimizer is used to minimize the total loss function, and the model parameters are updated through the backpropagation algorithm. During training, the model performance is monitored in real time using the validation set, and an early stopping strategy is adopted to prevent overfitting. When the validation set loss no longer decreases after 10 consecutive iterations, training is stopped, the current optimal model parameters are saved, and the optimal evaluation model is obtained.
[0054] In a specific implementation, the evaluation and prediction in S5 include: Acquire real-time satellite remote sensing data (including the latest vegetation index, surface temperature, topographic data, etc.) and water resource allocation scheme parameters (such as allocation flow rate, allocation period, etc.), perform preprocessing and feature extraction in S2, and input them into the optimal evaluation model obtained in S4: Carbon storage and transport assessment: The model outputs the spatial distribution of carbon storage, carbon transport flux and carbon transport path in the current watershed. The spatial resolution of the assessment results is consistent with that of satellite remote sensing data, and the temporal resolution is set to 1 day to 1 month according to actual needs.
[0055] Carbon storage and transport prediction: Based on parameters of different water resource scheduling schemes, predict the trend of watershed carbon storage changes and carbon transport characteristics in different time periods (short-term: 1-3 months, medium-term: 6-12 months, long-term: 1-5 years), and provide a scientific basis for unified water resource scheduling and carbon emission reduction decisions in the watershed.
[0056] Results Output and Visualization: The output results support multiple data formats such as GeoJSON and TIFF, and can be integrated with GIS platforms to achieve real-time map rendering; at the same time, a carbon storage and migration assessment report is generated, including carbon storage statistics, identification of key carbon migration areas, and carbon effect analysis of water resource scheduling schemes.
[0057] In a specific embodiment, an electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of a watershed carbon storage transport assessment method based on a tree growth model.
[0058] In a specific embodiment, a non-transitory computer-readable medium stores a computer program thereon, which, when executed by a processor, implements the steps of a watershed carbon storage transport assessment method based on a tree growth model.
[0059] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.
[0060] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for assessing watershed carbon storage and transport based on a tree growth model, characterized in that, include: Collect multi-source data, preprocess the multi-source data, and obtain comprehensive characteristic data of factors affecting watershed carbon storage and transport; Construct a watershed carbon storage and transport model that integrates tree growth model; The comprehensive feature data is divided into training set, validation set and test set, and input into the watershed carbon storage and transport model for training. The model parameters are optimized by comparing the learning loss function and the elastic weight solidification method to obtain the optimal evaluation model. By acquiring real-time satellite remote sensing data and water resource scheduling scheme parameters, preprocessing and extracting their features, and then inputting them into the optimal evaluation model, the evaluation and prediction results of the spatial distribution, transport flux and changing trend of watershed carbon storage under unified water resource scheduling are obtained.
2. The watershed carbon storage and transport assessment method based on a tree growth model according to claim 1, characterized in that, The multi-source data includes satellite remote sensing data, tree growth-related data, runoff data, and biomass data. Among them, satellite remote sensing data covers vegetation index, surface temperature, and topographic information; tree growth-related data includes tree ring width, tree height, diameter at breast height, and growth rate; runoff data includes runoff volume and flow velocity; and biomass data includes measured and estimated values of biomass of trees, shrubs, and herbs.
3. The watershed carbon storage and transport assessment method based on a tree growth model according to claim 1, characterized in that, The multi-source data is preprocessed, including denoising, outlier removal, and normalization, to obtain standardized data. The standardized data is then spatiotemporally aligned to eliminate time and spatial deviations between different data sources. The spatiotemporally aligned standardized data is then input into a self-attention mechanism to dynamically allocate weight coefficients for each type of data. Based on these weight coefficients, different types of data are then weighted and concatenated.
4. The watershed carbon storage and transport assessment method based on a tree growth model according to claim 1, characterized in that, The watershed carbon storage and transport model includes an input layer, a tree growth simulation module, a carbon storage accounting module, a runoff-driven carbon transport module, a dynamic correction module, and an output layer. The input layer is used to receive comprehensive feature data; The tree growth simulation module constructs a growth prediction model based on tree ring data and environmental factors to simulate the process of tree biomass accumulation. The carbon storage calculation module calculates the total carbon storage of the watershed based on tree biomass and other vegetation biomass data, combined with the carbon conversion coefficient. The runoff-driven carbon transport module constructs a carbon migration and diffusion model based on runoff data to quantify the impact of runoff on the spatial transport of carbon storage. The dynamic correction module uses a learnable error compensation function to correct model prediction biases in real time; the output layer is used to output carbon storage migration assessment and prediction results.
5. The watershed carbon storage and transport assessment method based on a tree growth model according to claim 4, characterized in that, The tree growth simulation module constructs a growth prediction model based on tree ring data and environmental factors to simulate the tree biomass accumulation process, including: A tree growth prediction model is constructed using formulas: ； in, for Tree biomass at all times This is the initial biomass. for Real-time vegetation index for Standardized value of surface temperature at any given time. for Moisture stress coefficient at any time , , For model parameters, Let be the width of the tree rings in year i. The average width of tree rings. The number of years observed.
6. The watershed carbon storage and transport assessment method based on a tree growth model according to claim 4, characterized in that, The carbon storage calculation module calculates the total carbon storage of the watershed based on tree biomass and other vegetation biomass data, combined with the carbon conversion coefficient, including: Calculate the total carbon storage of the watershed using the following formula: ； in, The total carbon storage of the basin. The number of watershed partitions, For the first Total tree biomass in the zone The carbon conversion coefficient of trees. For the first Total biomass of other vegetation in the zone For the first Carbon conversion coefficient of other vegetation in the zone For the first Zonal soil carbon storage.
7. The watershed carbon storage and transport assessment method based on a tree growth model according to claim 4, characterized in that, The runoff-driven carbon transport module constructs a carbon migration and diffusion model based on runoff data, quantifying the impact of runoff on the spatial transport of carbon storage, including: A carbon migration and diffusion model is constructed using formulas: ； in, Coordinates at time t carbon concentration at that location For time step, Where is the diffusion coefficient. This is the runoff velocity vector. for Time coordinates The amount of carbon concentration change caused by tree growth.
8. The watershed carbon storage and transport assessment method based on a tree growth model according to claim 4, characterized in that, The dynamic correction module employs a learnable error compensation function to correct model prediction bias in real time, including: The error compensation value is calculated using the formula: ； in, for Time coordinates Error compensation value at the location, To predict carbon concentration for the model, To measure the carbon concentration, , These are learnable parameters; The corrected carbon concentration is: .
9. The watershed carbon storage and transport assessment method based on a tree growth model according to claim 1, characterized in that, The process of optimizing model parameters by comparing the learning loss function and the elastic weight solidification method to obtain the optimal evaluation model includes: The comparative learning loss function is calculated using the following formula: ； in, To compare the learning loss function, Cosine similarity for positive sample pairs. For the first Cosine similarity of negative sample pairs For temperature parameters, This represents the number of negative sample pairs. The elastic weight solidification loss function is calculated using the following formula: ； in, For the elastic weight solidification loss function, This is a set of indexes for key parameters of the model. For the corresponding parameters in the Fisher information matrix The value, The value of the current training parameter i, For parameters The optimal value in historical training; The total loss function is: ； in, For mean square error loss, , , The loss weights are used to minimize the total loss function through backpropagation, thereby optimizing the model parameters.
10. A watershed carbon storage and transport assessment system based on a tree growth model, characterized in that, include: The data acquisition module is used to collect data from multiple sources. The data processing module is used to preprocess the multi-source data to obtain comprehensive characteristic data of the factors affecting watershed carbon storage and transport. The model building module is used to build a watershed carbon storage and transport model that integrates tree growth model; The model training module is used to divide the comprehensive feature data into training set, validation set and test set, input them into the watershed carbon storage and transport model for training, and optimize the model parameters by comparing the learning loss function and the elastic weight solidification method to obtain the optimal evaluation model. The assessment and prediction module is used to acquire real-time satellite remote sensing data and water resource scheduling scheme parameters. After preprocessing and feature extraction, the data is input into the optimal assessment model to obtain the assessment and prediction results of the spatial distribution, transport flux and changing trend of watershed carbon storage under unified water resource scheduling.