Model for estimating carbon reserves of dead residues of soil microorganisms

Abstract

The invention discloses a model for estimating carbon reserves of soil microbial death residues, which considers that the carbon of the microbial death residues has a great contribution to the formation of total organic carbon in soil, and establishes a set of estimation algorithm representing the microbial death residues and develops four modules for realizing the algorithm, so that the carbon reserves of the soil microbial death residues are estimated. The reserve of microbial residual carbon in soil types can be estimated. In order to solve the problem that the carbon reserves of dead soil microorganism residues are difficult to accurately estimate on a large spatial scale, a neural network machine learning method is combined in an algorithm module, and a regional and even global soil microorganism residue carbon reserve map can be drawn through measurement data of some observation points. The core technology of the method is based on a theoretical algorithm for estimating microbial residues, computer software for realizing the algorithm is developed, the computer software comprises four modules, measurement data of field observation points are input, and spatial distribution of soil microbial residue carbon reserves and contribution to soil total organic carbon can be automatically, rapidly and accurately estimated.

Description

Technical Field

[0001] This invention belongs to the field of soil environmental protection technology, specifically involving an algorithm that can accurately and automatically estimate the carbon storage and spatial distribution of soil microbial residues, and can estimate a set of data products for soil microbial residue carbon storage. Background Technology

[0002] The continuous rise in atmospheric CO2 concentration in recent years has led to significant changes in global climate, a matter of great concern. Since soil organic carbon (SOC) is the largest carbon pool in terrestrial ecosystems, even small changes in soil carbon can be fed back into the atmosphere, causing significant changes in atmospheric CO2. Currently, the soil carbon cycle remains the least estimated link in terrestrial carbon cycle research, and estimates of the soil carbon pool also contain significant errors; current models have an uncertainty of 50% in estimating global SOC. Therefore, incorporating mechanisms confirmed by observational experiments into models remains crucial for more accurately simulating and predicting carbon cycles and climate change.

[0003] Microorganisms not only participate in the biogeochemical cycles of terrestrial ecosystems but are also a major driving force in the decomposition of soil organic matter (SOC). Most of the SOC decomposed by microorganisms originates from relatively unstable plant-derived carbon, with the remaining plant residues after microbial decomposition accounting for a significant proportion of the total SOC. Previous studies suggested that plant carbon is converted into CO2 and stable SOC; however, research on the impact of dead soil microbial residues on soil carbon is limited. Plant-derived carbon can be directly input into SOC through extracellular enzymatic hydrolysis by microorganisms or indirectly input into SOC through microbial circulation to the microbial biomass carbon (MBC) pool, and then form stable SOC through interactions with soil minerals. Due to the different molecular structures of microbial and plant-derived residue carbon, these significant differences determine the stability, aggregation, and reactivity of soil organic matter (SOM). Therefore, carbon input pathways do not necessarily affect SOM content but significantly determine its structure.

[0004] Living microorganisms in the soil constitute a relatively small biomass pool. Due to their relatively rapid metabolism, most of the organic carbon input into the soil can be continuously metabolized into various products and dead remains by microorganisms, leading to the accumulation of microbial-derived carbon. Compared to plant-derived carbon, microbial-derived carbon may more readily combine with soil minerals to form stable soil organic matter (SOC), and it makes a significant contribution to the formation of stable soil organic matter (SOM) through microbial metabolic cycles in each life cycle. Therefore, exploring the relationship between microbial-derived carbon and stable SOC will help improve the prediction of SOC.

[0005] Therefore, the problems existing in the above theories and technologies are as follows:

[0006] ① These methods do not distinguish between microbial dead body banks and plant residue banks.

[0007] ② Current technical methods do not explicitly consider the participation of microbial residue carbon pools.

[0008] ③ Estimating microbial and plant residues at the same decomposition rate will inevitably lead to some error.

[0009] ④ Currently, there is no technical method to calculate the amount of soil microbial residue in a region using data from a few observation points. Summary of the Invention

[0010] The purpose of this invention is to solve the above problems by providing a model that can accurately and quickly estimate the carbon storage of soil microbial death residues.

[0011] To achieve the aforementioned objective, the technical solution proposed in this invention comprises four steps and corresponding to four software processing modules, as detailed below:

[0012] 1) Module ①: Data Preprocessing Module

[0013] S1: 13 Data acquisition and preprocessing of C-isotope-labeled microbial residue decomposition experiments.

[0014] In module ①, the input data is 13 Observational data on the decomposition of C-labeled microbial remains ( Figure 1 (S1), the example data in this study includes observational data published by four studies ( Figure 3 These four observation data are merely experimental data for this invention and do not limit the scope of application of this invention. Other types of data can be used as long as the data input conditions are met.

[0015] For the collected 13 The C-labeled microbial residue decomposition experimental data were standardized by processing all data into a percentage of decomposition relative to the initial time to address data consistency issues under different conditions and observations. The processed data was then input into step 2 (S2).

[0016] 2) Module ②: Code implementation module for the decomposition equation of microbial residues.

[0017] S2: Establish the theoretical equations for the carbon decomposition of microbial residues through code. Based on the decomposition equation formulas, implement the equations using MATLAB.

[0018] The specific equations and techniques involved in this step are as follows:

[0019] First, the equational structure for the decomposition of soil microbial residues is established, specifically in... Figure 4 The detailed demonstration is presented below, and its mathematical expression is a system of differential equations:

[0020] (1)

[0021] (2)

[0022] (3)

[0023] (4)

[0024] (5)

[0025] Among them, C P Plant-derived organic carbon; C B Microbial biomass carbon; C NF : Microbial residue carbon fast library; C N-MAOM : Microbial carbon slow reservoir (mainly due to mineral adsorption); C DP Plant-derived dissolved organic carbon; C DN Microbial-derived dissolved organic carbon. C N This refers to total microbial residue carbon (MNC). Units: mg C g -1 soil. I is the annual average carbon input rate of plant litter to the soil (g C m). -2 s -1 The calculation is based on the input flux of litter layer minus the heterotrophic respiration of litter, and then the units are converted to mg / C / g using soil bulk density data. -1 soil hr -1 V max,P V max,N They are C P and C N Maximum decomposition rate (mg C mg) -1 MBC hr -1 K M,P K M,N They are C P and C N The half-saturation constant (mg C g) at which the decomposition rate reaches half of its maximum value. -1 soil). CUE P CUE N C respectively P and C N Carbon utilization efficiency (unitless). k BIt is the average mortality rate of the microbial community (hr) -1 f BNF It is C NF The percentage of total microbial mortality (unit not specified). R MAOM-F It is C N-MAOM With C NF The ratio of the decomposition rates (unitless).

[0026] In the equation, the input and output parts of each carbon library are composed of different units marked with plus and minus signs. A plus sign indicates the input of that carbon library, and a minus sign indicates the output. As a complete cyclic system, the output of the previous carbon library often becomes the input of the next carbon library.

[0027] Setting the rate of change of all carbon pools to zero, the above differential equations can be rewritten as a system of multivariate nonlinear equations. Solving this system yields the calculation formulas for each carbon pool under steady-state conditions (when each carbon pool no longer changes with time and soil processes reach a steady state):

[0028] (6)

[0029] (7)

[0030] (8)

[0031] (9)

[0032] (10)

[0033] (11)

[0034] Secondly, the method for estimating the equation parameters first uses the ODE45 solver provided by MATLAB to calculate the differential equation y = f(t,y) with initial conditions y0 from t0 to t n Numerical integral ( Figure 1 (S2). ODE45 is a numerical single-step solver for ordinary differential equations based on the explicit Runge–Kutta(4,5) equations, requiring only the previous time step y. tn-1 Solve for the result.

[0035] Finally, the input of this data in this process comes from the output data in step 1 (S1). The output data in S1 is input into module 2 corresponding to this step. By implementing the above set of equations, the change of each carbon pool over time and the corresponding parameter values ​​in the equations can be solved.

[0036] 3) Module ③: Equation parameter estimation module.

[0037] S3: Optimize the equation parameters in step S2. Use the lsqnonlin function provided by MATLAB to fit the observed data. This function mainly uses the nonlinear least squares method to fit the data. The fitting slope and Ri of the simulation results and observed data are calculated using the following formulas. 2 Root mean square error (RMSE) and mean absolute error (MAE) are used as indicators to quantify the accuracy of model estimation.

[0038] (12)

[0039] (13)

[0040] (14)

[0041] in, y is the simulated value, and y is the observed value. is the average value, and n is the number of observations.

[0042] Then, the parameters are constrained within a given range and solved to obtain the optimal equation parameter results under different experimental observation conditions, which are used for subsequent calculations. Figure 1 S3).

[0043] 4) Module ④: Spatial extrapolation module for equations.

[0044] S4: Predict the spatial distribution of microbial residue carbon using a neural network random forest method. The parameters of the soil microbial residue carbon decomposition equation obtained in step S3 are used as input to step S4. A neural network random forest method is used to train the relationship between the two, establishing a nonlinear relationship between the equation parameters and environmental variables. Through multiple iterations, the spatial distribution of microbial residue carbon is obtained. The output is a binary image file with the .tif extension, which can be directly opened and used with ENVI and ArcGIS software, and has the same map projection coordinate system as the input image data.

[0045] This invention develops four corresponding software modules through the aforementioned four steps. These four modules are sequentially structured, meaning that the output data of one module becomes the input data of the next. The final result from the combined output of the four modules realizes the method proposed in this invention for estimating the carbon storage of soil microbial death residues.

[0046] The method for estimating soil microbial residual carbon storage proposed in this invention can solve the following problems:

[0047] The necessity of incorporating microbial carbon pools into Earth system models can be assessed.

[0048] It can improve the accuracy of soil organic carbon estimation using the carbon pool equation representing microbial residues.

[0049] It can help us understand the decomposition rate of microbial residual carbon in different types of soil.

[0050] The contribution of microbial residue carbon to SOC can be estimated.

[0051] ⑤ It can provide a scientific basis for the country's major strategy of carbon neutrality.

[0052] Based on the above process, a method and process for estimating the carbon storage of soil microbial death residues were developed using the MATLAB computer programming language. The detailed calculation process is indicated in the specific implementation, and the method flow, the relationship between system modules, and related estimation results examples are given in the attached figures.

[0053] This invention includes an established decomposition equation for microbial residue carbon, based on the latest theories proven in cutting-edge research, which can accurately estimate the decomposition of soil microbial residues. This invention not only establishes the decomposition equation for soil microbial residues but also estimates the fate of microorganisms after death under simulated actual soil conditions. Based on the latest theoretical equation, this invention has developed computer software to implement the algorithm, comprising four modules. Inputting measurement data from field observation points, it can automatically, quickly, and accurately estimate the spatial distribution of soil microbial residue carbon storage and its contribution to total soil organic carbon. Through the aforementioned theoretical methods and technical means, operational efficiency and estimation accuracy can be improved.

[0054] In addition, this invention can lay an important scientific foundation for a better understanding, simulation and prediction of ecosystem carbon sinks, and provide a scientific basis for major national strategies for carbon neutrality. Attached Figure Description

[0055] Figure 1 A flowchart for a model to estimate the carbon storage of soil microbial death residues;

[0056] Figure 2 A deployment structure diagram for a model to estimate the carbon storage of soil microbial death residues;

[0057] Figure 3 Microbial residues in four different types of soil 13 Observational data on the change of C recovery rate over time;

[0058] Figure 4 The structure of microbial carbon residues participating in the soil carbon cycle;

[0059] Figure 5 Simulation results of the equations;

[0060] Figure 6 The proportion of carbon in microbial residues to total organic carbon. Detailed Implementation

[0061] The present invention will now be described in detail with reference to the accompanying drawings:

[0062] The invention uses 13 Experimental observation data on the decomposition of microbial remains labeled with C isotope are described in detail in section S1 of the invention. Searching for information on the Web of Science platform... 13 Observational data on the decomposition of C-labeled microbial remains were screened and compiled into four publicly available research data sets. Figure 3 These studies investigated soil microorganisms in soil rich in... 13 Culturing in substrates of C, killing them forms a rich... 13 Microbial residues of C. Rich in 13 Microbial remains of C were placed in soil for decomposition. Studies 1 and 2 involved culturing in the laboratory and using... 13 C-labeled E. coli residues were collected from soil samples derived from temperate farmland and forest ecosystems, respectively. Studies 3 and 4 involved in-situ decomposition experiments using... 13 The soil samples, including C-labeled fungi, Gram-positive bacteria (including actinomycetes), and Gram-negative bacteria, originated from temperate and tropical forest ecosystems.

[0063] Of the four studies, Studies 1 and 2 had relatively short observation periods, approximately 250 days, while Studies 3 and 4 had relatively long observation periods, approximately 900–1150 days (Figure 3). In comparison, this invention classifies Studies 1 and 2 as short-term studies and Studies 3 and 4 as long-term studies. These four observational data are merely experimental data from this invention and do not limit the scope of its application; other types of data can be used as long as the data input conditions are met.

[0064] In addition, this invention requires some auxiliary data, such as: SOC input rate I from the output data of Community LandModel (CLM5.0), which is published by the National Center for Atmospheric Research (NCAR) (https: / / www.earthsystemgrid.org). The monthly litter rate file is clm50_r270_1deg_GSWP3V1_iso_newpopd_hist.clm2.h0.LITFALL.185001-201412.nc, and the monthly litter heterotrophic respiration file is clm50_r270_1deg_GSWP3V1_iso_newpopd_hist.clm2.h0.LITTERC_HR.185001-201412.nc. A subset between 2000 and 2014 was selected and processed into annual average data.

[0065] The above is the preliminary preparation of the data. In this invention, it is specifically implemented through the following steps:

[0066] First, the equational structure for the decomposition of soil microbial residues is established, specifically in... Figure 4 The detailed demonstration is presented below, and its mathematical expression is a system of differential equations:

[0067] (1)

[0068] (2)

[0069] (3)

[0070] (4)

[0071] (5)

[0072] Among them, C P Plant-derived organic carbon; C B Microbial biomass carbon; C NF : Microbial residue carbon fast library; C N-MAOM : Microbial carbon slow reservoir (mainly due to mineral adsorption); C DP Plant-derived dissolved organic carbon; C DN Microbial-derived dissolved organic carbon. C N This refers to total microbial residue carbon (MNC). Units: mg C g -1 soil. I is the annual average carbon input rate of plant litter to the soil (g C m). -2 s -1The calculation is based on the input flux of litter layer minus the heterotrophic respiration of litter, and then the units are converted to mg / C / g using soil bulk density data. -1 soil hr -1 V max,P V max,N They are C P and C N Maximum decomposition rate (mg C mg) -1 MBC hr -1 K M,P K M,N They are C P and C N The half-saturation constant (mg C g) at which the decomposition rate reaches half of its maximum value. -1 soil). CUE P CUE N C respectively P and C N Carbon utilization efficiency (unitless). k B It is the average mortality rate of the microbial community (hr) -1 f BNF It is C NF The percentage of total microbial mortality (unit not specified). R MAOM-F It is C N-MAOM With C NF The ratio of the decomposition rates (unitless).

[0073] In the equation, the input and output parts of each carbon library are composed of different units marked with plus and minus signs. A plus sign indicates the input of that carbon library, and a minus sign indicates the output. As a complete cyclic system, the output of the previous carbon library often becomes the input of the next carbon library.

[0074] Setting the rate of change of all carbon pools to zero, the above differential equations can be rewritten as a system of multivariate nonlinear equations. Solving this system yields the calculation formulas for each carbon pool under steady-state conditions (when each carbon pool no longer changes with time and soil processes reach a steady state):

[0075] (6)

[0076] (7)

[0077] (8)

[0078] (9)

[0079] (10)

[0080] (11)

[0081] Secondly, the method for estimating the equation parameters first uses the ODE45 solver provided by MATLAB to calculate the differential equation y = f(t,y) with initial conditions y0 from t0 to t n Numerical integral ( Figure 1 (S2). ODE45 is a numerical single-step solver for ordinary differential equations based on the explicit Runge–Kutta(4,5) equations, requiring only the previous time step y. tn-1 The results are then obtained. Next, the observed data are fitted using the lsqnonlin function provided by MATLAB. This function primarily employs nonlinear least squares to fit the data, constraining the parameters within a given range and solving for the optimal combination of parameters. Figure 1 S3).

[0082] In the steps of establishing the decomposition equation of carbon in microbial residues ( Figure 1 (S2). We used the Michaelis-Menten equation to describe the nonlinear variation of substrate decomposition rate with MBC. Because we used culture experimental data under controlled conditions, we did not define a functional relationship between the decomposition rate and environmental factors. In the equation (the decomposition equation and its structure diagram are given in the specific implementation), after the soil microorganisms die, the microbial remains (i.e., the portion leaving the MBC library) are transferred proportionally to the MNC fast library (C). NF ) and MNC slow library (C N-MAOM In order to avoid introducing more parameters and overfitting, the number of other carbon libraries is reduced to a minimum.

[0083] In the step of estimating the parameters in the decomposition equation of carbon in microbial remains ( Figure 1 (S3). Based on previous research, some parameters already have prior fixed values, while most parameters are given a range. Within this range, the parameters are estimated by random sampling, assuming that they follow a uniform distribution. For some unique model parameters proposed in this model, parameters with similar significance to those in previous studies are selected as prior values ​​or prior ranges.

[0084] Finally, based on multiple environmental variables and optimized parameter values, a deep learning neural network model was designed. This model consists of four processing layers, each containing 256, 512, 512, and 256 neurons respectively. A non-linear activation function was used to enable the neural network to explore the non-linear relationship between the input and the final output, and the neuron weights were adjusted based on the loss function results. Finally, using the trained neural network model, the spatial distribution of the equation parameters was extrapolated based on the rasterized environmental variables to generate a parameter map of the regional distribution. This map was then substituted into the equation to obtain the spatial distribution of the microbial residues. Figure 1 in S4).

[0085] Example 1: A model for estimating the carbon storage of dead soil microorganisms

[0086] A model for estimating the carbon storage of dead soil microorganisms is employed. The following detailed description, in conjunction with accompanying drawings and embodiments, further illustrates the various details involved in the technical solution of this invention. This embodiment uses a personal computer (PC) for simulation implementation, with software based on a 64-bit Windows 10 operating system and the MATLAB secondary development platform. The embodiment used involves searching for information on the Web of Science platform. 13 Observational data on the decomposition of C-labeled microbial residues, based on the technical solution ( Figure 1 Write a MATLAB program to process the system. The system program structure is as follows: Figure 2 The spatial distribution of carbon storage in the dead soil microorganisms was obtained by debugging and running the program.

[0087] Develop a program to estimate the carbon storage of dead soil microorganisms: Based on ① the theoretical equations of microbial residue decomposition in the technical solution, implement the program; ② solve the system of differential equations and optimize the model parameters; ③ predict its spatial distribution through neural network machine learning, encompassing four modules, such as... Figure 2 .

[0088] Soil microbial residue decomposition data used in the examples are as follows: Figure 3 As shown, they are all standardized to the same unit level.

[0089] Differential equations 1-5 are used to simulate the decomposition of microbial remains, and equations 6-11 are used to calculate the value of each carbon pool under steady-state conditions.

[0090] The results of the differential equation were fitted using the lsqnonlin function provided by MATLAB to solve for the optimal equation parameters. The simulation results of the equation on the observed values ​​are as follows: Figure 5 As shown.

[0091] Predicting the spatial distribution of carbon in microbial residues using neural network machine learning ( Figure 6 In this invention, the continental United States region was selected as the prediction range for machine learning. This region is merely an example in this embodiment and does not limit the prediction range of this invention.

Claims

1. A model for estimating the carbon storage of soil microbial death residues, comprising: 1) Data Preprocessing Module For the collected 13 The experimental data on the decomposition of microbial residues labeled with C were standardized by converting all data into a percentage of decomposition relative to the initial time. 2) Code implementation module for microbial residue decomposition equations The theoretical equations for the carbon decomposition of microbial residues are established through code implementation. Based on the decomposition equation formula, the equations are implemented using code. The theoretical equations for the decomposition of soil microbial residues are expressed mathematically as a set of differential equations: （1） （2） （3） （4） （5） Among them, C P Plant-derived organic carbon; C B Microbial biomass carbon; C NF : Microbial residue carbon fast library; C N-MAOM : Slow carbon reservoir of microbial remains; C DP Plant-derived dissolved organic carbon; C DN Microbial-derived dissolved organic carbon; C N Total microbial residues carbon (MNC), unit: mg C g -1 soil; I is the annual average carbon input rate of plant litter to the soil, in g C m. -2 s -1 The calculation is based on the input flux of litter layer minus the heterotrophic respiration of litter, and then the units are converted to mg / C / g using soil bulk density data. -1 soil hr -1 V max,P V max,N They are C P and C N Maximum decomposition rate (mg C mg) -1 MBC hr -1 ;K M,P K M,N They are C P and C N The half-saturation constant mg ​​C g when the decomposition rate reaches half of its maximum value -1 soil. CUE P CUE N C respectively P and C N carbon utilization efficiency; k B The average mortality rate of the microbial community is hr -1 ;f BNF It is C NF The proportion of total microbial death; R MAOM-F It is C N-MAOM With C NF The ratio of their decomposition rates; The input and output parts of each carbon library in the equation are composed of different units with plus and minus signs. Each unit is preceded by a positive sign to represent the input part of that carbon library, and a negative sign to represent the output part of that carbon library. As a complete cycle system, the output of the previous carbon library is often the input of the next carbon library. Setting the rate of change of all carbon pools to zero, the above differential equations can be rewritten as a system of multivariate nonlinear equations. Solving this system yields the formulas for calculating each carbon pool under steady-state conditions, where each carbon pool no longer changes with time and the soil process reaches a steady state. （6） （7） （8） （9） （10） （11） The method for estimating the equation parameters first uses the ODE45 solver provided by MATLAB to calculate the differential equation y = f(t,y) with initial conditions y0 from t0 to t0. n Numerical integration; ODE45 is a numerical single-step solver for ordinary differential equations based on the explicit Runge–Kutta (4,5) equations, requiring only the previous time step y. tn-1 Solve for the result; In this process, the input of the data comes from the output data in S1. The output data in S1 is input into S2. By implementing the above set of equations, the change of each carbon pool over time and the corresponding parameter values ​​in the equations can be solved. 3) Equation parameter estimation module Module Optimization of Equation Parameters for Microbial Residue Decomposition Equations The lsqnonlin function provided by MATLAB is used to fit the observed data. This function mainly uses the nonlinear least squares method to fit the data. The fitting slope and R-squared of the simulation results and the observed data are calculated using the following formulas. 2 Root mean square error (RMSE) and mean absolute error (MAE) are used as indicators to quantify the accuracy of model estimation. （12） （13） （14） in, y is the simulated value, and y is the observed value. It is the average value, and n is the number of observations; Then, the parameters are constrained within a given range and solved to obtain the optimal equation parameter results under different experimental observation conditions, which can be used for subsequent calculations; 4) Spatial extrapolation module of equations Predicting the spatial distribution of carbon in microbial residues using a neural network random forest method. The parameters of the soil microbial residue carbon decomposition equation obtained from the equation parameter estimation module are used as input to the spatial extrapolation module of the equation. The relationship between the two is trained using a neural network random forest method to establish a nonlinear relationship between the equation parameters and environmental variables. Through multiple iterations of training, the spatial distribution image of microbial residue carbon is obtained. The output is directly as a binary image file with the file extension .tif, which can be directly opened and used with ENVI and ArcGIS software. It has the same map projection coordinate system as the input image data.

2. The model for estimating the carbon storage of soil microbial death residues according to claim 1, characterized in that: Implemented using MATLAB software.

Images