Water body carbon emission long-term trend prediction method based on Bayesian additive regression tree

Abstract

The invention relates to a water body carbon emission long-term trend prediction method based on a Bayesian additive regression tree, and belongs to the field of water body carbon emission prediction. The method comprises the steps that water-gas interface carbon flux data and related hydrological and water environment data are collected, the data are preprocessed and subjected to dimension unification, and predictive variables are determined through correlation analysis and a structural equation model. Thirdly, constructing a prediction model by using a BART algorithm model, and verifying the performance of the model through a test set; in addition, the method further comprises the steps of utilizing a BART algorithm model to invert historical carbon emission conditions, and coupling with a global climate mode to predict carbon emission trends under different climate change scenes in the future. The method provides an effective way for evaluation and long-term trend prediction of the carbon emission level of the water body, and is helpful for deeply understanding and alleviating the influence of global climate change.

Description

Technical Field

[0001] The present invention belongs to the field of water body carbon emission prediction, and relates to a water body carbon emission long-term trend prediction method based on Bayesian additive regression tree. Background Art

[0002] Research on greenhouse gas emissions from water bodies not only covers changes in gas sources and sinks, spatiotemporal characteristics and their driving factors, but also includes the biogeochemical cycles of various biogenic elements. These studies not only focus on the emission characteristics of greenhouse gases from water bodies, but also aim to develop models to estimate and predict the long-term trends of these emissions. Modeling plays an important complementary role in estimating greenhouse gas concentrations or fluxes, especially in compensating for the limitations of field monitoring in time and space. Through modeling, quantitative relationships can be extended from limited observations to larger regional or global scales, thereby helping to formulate effective plans to respond to climate change.

[0003] However, most of the existing models are based on empirical relationships or assume linear relationships between independent variables. These models have limitations in dealing with the complex interactions of greenhouse gas fluxes such as CH4 and CO2 in freshwater ecosystems. Greenhouse gas fluxes in freshwater ecosystems are affected by many factors, and there are obvious interactions between these factors. Although current research has shifted from single observational studies to studies combining observations and models, existing models are still insufficient to deeply understand the carbon emission characteristics of freshwater ecosystems, making it difficult to accurately estimate and predict greenhouse gas fluxes in water bodies.

[0004] Data-driven machine learning models provide a new solution. Unlike traditional empirical models, machine learning models do not need to rely on complex parameters and are good at handling nonlinear relationships. These models can effectively explore data characteristics, clarify the interactions between variables, and simulate and predict trends in the future. However, building a successful machine learning model requires a large amount of data for training, and it is often difficult to obtain water carbon emission data, resulting in few machine learning models for water carbon emissions and high uncertainty in predictions. Therefore, there is an urgent need to develop a more effective data-driven machine learning model to more easily simulate and predict water carbon emissions. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a method for predicting the long-term trend of carbon emissions from water bodies based on a Bayesian additive regression tree, so as to provide technical support for better understanding the characteristics of carbon emissions from water bodies.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The long-term trend prediction method of water body carbon emissions based on Bayesian additive regression tree includes the following steps:

[0008] Data collection and preprocessing: Collect carbon flux data at the water-air interface and related hydrological and water environment data to form an original data set. Specific indicators include: CO2 flux (f CO2 )、CH4 flux (f CH4 ), inflow (Qin), water level (WL), water temperature (WT), chlorophyll a (Chl-a), total nitrogen (TN), total phosphorus (TP), and dissolved organic carbon (DOC). The time scale of each indicator needs to be unified. In order to ensure that the model effectively extracts the seasonal variation characteristics of water carbon emissions, it is recommended to carry out model training and verification with data on a monthly scale or finer time scale. At the same time, in order to ensure that there is enough data to fully train the water carbon emission prediction model, for the prediction of water carbon emissions, as much data as possible from different spatial points can be collected. The original data set is preprocessed, including data cleaning, removal of erroneous data and empty data, and unification of dimensions using methods such as log transformation.

[0009] Determination of predictor variables: Correlation analysis and structural equation model were used to analyze the relationship between water carbon flux and various hydrological and water environmental factors to determine the predictor variables. Variables that were significantly correlated with carbon flux at a significance level of 0.05 were selected as predictor variables. Carbon emissions from large water bodies have high spatial heterogeneity. When conducting preliminary analysis, they can be analyzed separately by section (such as main streams and tributaries) to ensure the effectiveness of the analysis.

[0010] Model construction and testing: The data set is divided for model training and testing, the prediction model is constructed using the BART algorithm, and the model performance is tested on the test set. The model training set and test set are divided for the preprocessed data set. The BART algorithm model is used to construct a water body carbon flux prediction model with various hydrological and water environmental factors as prediction variables through the training set, and the model performance is tested on the test set. In the process of model construction, cross-validation is used to optimize the model parameters. When dividing the training set and the test set, the ratio of the two should not be too different to prevent the model from overfitting.

[0011] The basic expression of the BART algorithm model is as follows:

[0012]

[0013] Where N is the CO2 flux f CO2 、CH4 flux f CH4, inflow Qin, water level WL, water temperature WT, chlorophyll aChl-a, total nitrogen TN, total phosphorus TP and dissolved organic carbon DOC and other parameters; Y is the response variable, X is the driving variable, ε is the noise, k is the number of decision trees, k = 1, 2, ..., K, Τ is the structure of the tree, M is the parameter of the leaf node, Τ M Represents a whole tree.

[0014] These parameters are given and need to be optimized through cross-validation. A mature R package (bartMachine) has been developed for the BART algorithm. When using this package to build a BART model, parameters are set for the number and structure of decision trees, and then optimized through cross-validation.

[0015] Cross-validation is used to optimize model parameters during model construction to ensure that the ratio of training sets to test sets is appropriate to prevent model overfitting.

[0016] Historical carbon emission inversion: The constructed BART model is used to invert the historical carbon emissions of the water body, and the carbon flux level of the entire water body is inverted through the basic conditions of hydrological and water environmental elements. Carbon emissions are calculated based on the water surface area, and the water surface area is estimated as accurately as possible. The inverted historical carbon emission level of the water body is compared with the actual monitored carbon emission level to verify the effectiveness of the model. Among them, due to the discreteness of actual sampling in time and space, the actual carbon emission level needs to be obtained by extrapolating the actual point-like instantaneous monitoring of carbon flux in time and space. In terms of time, the instantaneous carbon flux monitoring value on the sampling day of each month can be extrapolated to the flux level of the month. In terms of space, the area of ​​the backwater area that the sampling point can represent is determined, and a weight is assigned to it to calculate the overall carbon emissions of the entire backwater area (within the boundary range) according to the carbon flux level and the size of the water surface area. If the data used is monthly frequency data, refer to the following formula to extrapolate the point-like instantaneous monitoring carbon flux to obtain the corresponding annual carbon emissions:

[0017]

[0018] Among them, i is the sampling point number, ranging from 1 to n; m is the month, ranging from 1 to 12; F di is the daily carbon flux value obtained at sampling point i; β ik is the water surface area represented by sampling point i in the kth month, β ik ≤1;D k is the day of the kth month.

[0019] Future trend prediction: Couple the BART carbon flux prediction model with the downscaling results of the global climate model to predict the long-term trend of carbon emissions in water bodies under different future climate change scenarios. Compare the performance of multiple climate models downscaled to the studied area to provide the BART model with boundary conditions under future climate change scenarios. These steps together constitute the core technical process of the long-term trend prediction method of water body carbon emissions based on Bayesian additive regression trees.

[0020] The beneficial effects of the present invention are:

[0021] (1) The model based on Bayesian additive regression tree (BART) can simulate and predict water carbon emissions more accurately, especially in dealing with complex nonlinear relationships. This model constructs the relationship characteristics between hydrological and water environmental factor data and water carbon flux, directly uses hydrological and water environmental data to predict carbon flux, without the need for additional carbon flux monitoring, thereby improving the accuracy of the prediction.

[0022] (2) The present invention reduces the reliance on direct carbon flux monitoring data by utilizing hydrological and water environment data, thereby reducing monitoring costs and difficulty. This method extrapolates point-like instantaneous carbon flux data in time and space, and combines it with the water surface area to more accurately estimate water carbon emissions, further reducing reliance on direct monitoring data.

[0023] (3) The model can adapt to the characteristics of different water bodies and improve its applicability and prediction accuracy in different water bodies by adjusting the prediction variables and model parameters. This adaptability enables the model to be applied to various water bodies, including those with different characteristics and conditions.

[0024] (4) Combined with the global climate model, the present invention can predict the long-term trend of carbon emissions from water bodies under different climate change scenarios in the future, providing a scientific basis for responding to climate change. This long-term trend prediction capability is crucial for understanding and predicting the long-term changing characteristics of carbon emissions from water bodies.

[0025] (5) The water carbon emission information provided can help environmental managers formulate effective climate change mitigation strategies and measures to protect aquatic ecosystems. This decision support capability is crucial for formulating effective environmental management strategies.

[0026] (6) It provides new tools and methods for the study of greenhouse gas emissions from water bodies, promoting scientific progress in this field. This scientific support capability is crucial to promoting research in the field of greenhouse gas emissions from water bodies.

[0027] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0029] Figure 1 It is a flow chart of the present invention;

[0030] Figure 2 The results of the correlation analysis between water carbon flux and various hydrological and water environmental factors in the specific embodiment of the present invention;

[0031] Figure 3 The structural equation model analysis results of water carbon flux and various hydrological and water environmental elements in the specific embodiment of the present invention;

[0032] Figure 4 The BART algorithm model training and testing performance of CH4 flux and CO2 flux at the water-air interface in a specific embodiment of the present invention;

[0033] Figure 5 The inversion result of historical carbon emission of water bodies in a specific embodiment of the present invention;

[0034] Figure 6 1 and 2 show the performance of eight different climate modes in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0035] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0036] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0037] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0038] Figure 1 This is a flow chart of the present invention. This embodiment predicts the long-term trend of carbon emissions from the water-air interface of a water body through a method for predicting the long-term trend of carbon emissions from water bodies based on a Bayesian additive regression tree. The method specifically includes the following steps:

[0039] 1. Compile and organize the monthly scale water-air interface carbon flux data and related hydrological and water environment data obtained from field monitoring of a river for 5 years (2010-2011, 2015-2017, 2021-2023) to form an original data set. Specific indicators include: CO2 flux (f CO2 )、CH4 flux (f CH4 ), flow (Qin), water level (WL), water temperature (WT), chlorophyll a (Chl-a), total nitrogen (TN), total phosphorus (TP), dissolved organic carbon (DOC), and the example of raw data set preparation is shown in Table 1.

[0040] Table 1

[0041]

[0042] 2. Preprocess the original data set, check the data validity, remove erroneous data and empty data, and unify the dimensions. When unifying the dimensions, divide the CO2 flux by 100, and perform log transformation on the remaining variables.

[0043] 3. Use correlation analysis and structural equation model to analyze the relationship between CO2 flux, CH4 flux and various hydrological and water environmental factors of a certain water body to determine the predictor variables. Figure 2 ) and structural model analysis ( Figure 3 ) showed that Qin, WL, WT, Chl-a, TN, TP, and DOC all significantly affected the f of the Three Gorges mainstream and tributaries to varying degrees. CO2 or CH4 Therefore, the predictor variables for water body CO2 flux and CH4 flux are selected as all the above variables.

[0044] 4. The water body related indicator data set was divided into a training set (n=317) and a test set (n=207) at a ratio of 6:4, and the carbon flux prediction model was constructed using the BART algorithm model. Cross-validation was used to optimize the model parameters during the model construction process. The training and testing performance of the BART algorithm model for CH4 flux and CO2 flux are shown in Figure 2. Figure 4 shown.

[0045] 5. Use the BART algorithm model of water CH4 flux and CO2 flux to invert historical carbon emissions ( Figure 5 ). During the inversion, the entire reservoir area is considered as a whole, and the hydrological and water environmental elements are determined as the average values ​​of each point in the month. The BART algorithm model of water CH4 flux and CO2 flux can be used to invert the carbon flux level of the entire river, and then the carbon emissions are calculated based on the water surface area.

[0046] 6. The BART prediction model of water CH4 flux and CO2 flux is coupled with the downscaling results of the global climate model to predict the long-term trend of water carbon emissions under different future climate change scenarios. Specifically, the daily temperature and precipitation data from 1940 to 2014 of a total of 249 meteorological stations in the Yangtze River Basin obtained from the National Information Center of the China Meteorological Administration (http: / / data.cma.cn / ) are used as benchmark data, and the 8 climate models in CMIP6 are used to simulate the temperature and precipitation data of the same period. The quantitative index S based on the Taylor diagram is used to evaluate the performance of the 8 different climate models in the upper reaches of the Yangtze River. Table 2 shows the long-term trend prediction results of water emissions in a specific embodiment of the present invention.

[0047] Table 2

[0048]

[0049] The evaluation results show that the simulation performance of each climate model for temperature is not much different. Therefore, the ACCESS-CM2 model with the best precipitation simulation was finally selected to predict the long-term changes in temperature and precipitation in the upper reaches of the Yangtze River under four shared socio-economic pathways (SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5) climate change scenarios until 2100. The precipitation conditions obtained by the ACCESS-CM2 model were brought into the Distributed Time Variant Gain Model (DTVGM) to simulate the flow changes in the upper reaches of the Yangtze River until 2100 under the climate change scenario. The future changes in temperature and flow under the climate change scenario are used as the boundary conditions for our use of the BART algorithm model to predict the long-term trend of carbon emissions in a river in the Yangtze River Basin. On the basis of the above, we obtained the changes in CO2 and CH4 fluxes at the water-air interface of a water body until 2100 through the BART carbon flux prediction model. Based on the predicted carbon flux and water surface area, we finally calculated the long-term trend results of carbon emissions in the water body, as shown below: Figure 6 shown.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for predicting the long-term trend of carbon emissions from water bodies based on Bayesian additive regression trees, characterized by: The method comprises the following steps: Collect hydrological, water environment and carbon flux data; Preprocess the data and unify the dimensions; Analyze data relationships to identify predictor variables; Divide the dataset for model training and testing; Use the BART algorithm model to build a prediction model and conduct performance testing; Inverse historical carbon emissions to validate the model; Coupling the BART algorithm model with global climate models to predict future trends; Calculate carbon emissions.

2. The method for predicting long-term trend of carbon emissions from water bodies based on Bayesian additive regression tree according to claim 1 is characterized in that: The carbon flux data include CO2 flux, CH4 flux, chlorophyll a and dissolved organic carbon.

3. The method for predicting long-term trend of carbon emissions from water bodies based on Bayesian additive regression tree according to claim 1 is characterized in that: The hydrological and water environment data include inflow, water level, water temperature, total nitrogen and total phosphorus.

4. The method for predicting long-term trend of carbon emissions from water bodies based on Bayesian additive regression tree according to claim 1 is characterized in that: The preprocessing includes data cleaning and dimension unification.

5. The method for predicting long-term trend of carbon emissions from water bodies based on Bayesian additive regression tree according to claim 1 is characterized in that: The analysis of data relationships includes correlation analysis and establishing a structural equation model.

6. The method for predicting long-term trend of carbon emissions from water bodies based on Bayesian additive regression tree according to claim 1, characterized in that: The data set is divided using a cross-validation technique.

7. The method for predicting long-term trend of carbon emissions from water bodies based on Bayesian additive regression tree according to claim 1, characterized in that: The parameters of the BART algorithm model were optimized through cross-validation.

8. The method for predicting long-term trend of carbon emissions from water bodies based on Bayesian additive regression tree according to claim 1, characterized in that: The inversion of historical carbon emissions includes estimating carbon flux levels throughout the water column.

9. The method for predicting long-term trend of carbon emissions from water bodies based on Bayesian additive regression tree according to claim 1, characterized in that: The coupling of the BART algorithm model with a global climate model includes a comparison of the downscaling performance of different climate models.

10. The method for predicting long-term trend of carbon emissions from water bodies based on Bayesian additive regression tree according to claim 9, characterized in that: The predicting of future trends includes providing boundary conditions under future climate change scenarios.

Images