Method for determining influence of straw returning mode on carbon emission of tobacco and rice rotation soil

Abstract

The invention provides a method for determining the influence of a straw returning mode on tobacco-rice rotation soil carbon emission, and belongs to the technical field of soil carbon emission. The method comprises the steps that test plots are divided, and different straw returning treatment groups are arranged; installing a static gas tank measuring system and regularly collecting a gas sample; measuring the emission of CO2 and N2O and the concentration of CH4 by using a gas chromatographic method, and calculating the carbon emission flux of the soil; analyzing a straw decomposition process by applying a soil microorganism kinetic equation; analyzing the dynamic change of a soil carbon library based on a carbon recurrent neural network model; constructing a carbon flow network topological structure by using a minimum spanning tree algorithm and identifying key nodes; according to the method, comprehensive influences of environmental factors, crop growth parameters and microbial community structures on carbon emission are integrated through a multi-head cross attention mechanism, accurate quantification and prediction of straw returning carbon emission are achieved, and a scientific basis is provided for optimizing a farmland carbon management strategy and promoting conversion from a farmland carbon source to a carbon sink.

Description

Technical Field

[0001] The present invention belongs to the technical field of soil carbon emissions, and in particular relates to a method for determining the impact of a straw return method on soil carbon emissions in a tobacco-rice rotation. Background Art

[0002] As an important farmland ecological management measure, straw incorporation is widely used in agricultural production systems such as tobacco-rice rotation. Traditional straw incorporation techniques include direct incorporation, crushed incorporation, and straw mulching. These methods increase soil organic matter content, improve soil structure, and promote soil fertility. However, existing straw incorporation techniques often lack a systematic carbon emission assessment mechanism during their application.

[0003] Currently, farmland carbon emission monitoring primarily relies on methods such as static chambers and eddy covariance methods. These methods suffer from limitations such as insufficient temporal and spatial resolution, high monitoring costs, and complex data interpretation. In tobacco-rice rotation systems, in particular, soil carbon cycling is complex and variable due to crop transitions and seasonal variations. Existing single-factor or linear models struggle to accurately reflect the comprehensive impact of straw incorporation on soil carbon emissions.

[0004] Therefore, how to construct a system that can comprehensively consider multidimensional factors such as microbial activity, environmental factors, crop growth parameters, and accurately quantify and predict soil carbon emissions in tobacco-rice rotation under different straw return treatment methods has become a key technical issue that needs to be urgently solved in the current field of farmland carbon management. Summary of the Invention

[0005] In view of this, the present invention provides a method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation, which can solve the technical problem in the existing technology that it is difficult to accurately quantify and predict the impact of different straw return treatment methods on soil carbon emissions in tobacco-rice rotation.

[0006] The present invention is achieved as follows: The present invention provides a method for determining the impact of straw return methods on soil carbon emissions in a tobacco-rice rotation, comprising: selecting a tobacco-rice rotation test field and dividing it into several plots; setting up a straw return experimental treatment; installing a static gas box measurement system; collecting gas samples at a set frequency; recording temperature data and humidity data in the box during sampling; using gas chromatography to determine the carbon dioxide concentration and methane concentration in the gas sample, and calculating the soil carbon emission flux and accumulation; applying a soil microbial kinetic equation to analyze the straw decomposition process; analyzing the dynamic changes of the soil carbon pool under different straw return conditions based on a carbon cycle neural network model, and integrating the comprehensive effects of environmental factors, crop growth parameters and microbial community structure on soil carbon emission flux through a multi-head cross-attention mechanism; using a minimum spanning tree algorithm to construct a carbon flow network topology structure, identify key carbon flow nodes, and optimize farmland carbon management strategies.

[0007] Among them, a tobacco-rice rotation experimental field was selected, and the experimental area was divided into several plots of 120 square meters in area. Random block arrangement was adopted, and each treatment was repeated 3 times.

[0008] Among them, a straw return experiment treatment was set up, including a non-straw return control group, a simple straw return group, and a straw return plus decomposition promoter and urea treatment group. The straw return amount was 4,000 kg per hectare, the decomposition promoter added amount was 30 kg per hectare, and the urea added amount was 75 kg per hectare.

[0009] Among them, the static air box measurement system includes an acrylic box, a stainless steel base, an automatic temperature and humidity recorder and a gas collection component, and static air boxes are set up on and under the ridges respectively.

[0010] Among them, gas samples are collected at a frequency of once every 7 days, and during the fertilization period, the sampling is intensified to once every 1 to 2 days and lasts for a week, and the sampling time is from 8:00 to 11:00 in the morning; the temperature and humidity data in the box are recorded during sampling, and 30 ml of gas samples are collected at the 0th minute, 10th minute, 20th minute and 30th minute after placement.

[0011] Among them, the soil carbon emission flux is calculated using gas density, gas box height, gas concentration change rate and temperature correction factor, and the unit is milligrams per square meter per hour.

[0012] Among them, the cumulative amount of soil carbon emissions is calculated by multiplying the average emission flux of adjacent sampling periods by the number of days between sampling, and the unit is kilograms per hectare.

[0013] Among them, the soil microbial kinetic equation is used to describe the quantitative relationship between microbial activity and carbon emissions during the decomposition of straw organic carbon, taking into account the factors of initial carbon content, temperature response, moisture response, carbon-nitrogen ratio response and microbial biomass carbon response.

[0014] Among them, the specific structure of the carbon cycle neural network model is a hybrid architecture combining a multi-layer bidirectional long short-term memory network and a graph convolutional network, which includes a temporal feature extraction layer, a spatial relationship encoding layer, a multi-head cross-attention layer and a carbon cycle parameter prediction layer.

[0015] Among them, the number of heads in the multi-head attention mechanism is determined by the number of soil carbon pools, and the attention weight is determined by the intensity of the impact of environmental factors on each carbon pool. The model input includes time series data of soil physical and chemical indicators, microbial community structure characteristics and meteorological data, and the output is the predicted value of the dynamic change of soil carbon pools and soil carbon emission flux.

[0016] The present invention achieves comprehensive monitoring and accurate prediction of the carbon emission process of straw return to the field by establishing a comprehensive evaluation system including a static gas box measurement system, a microbial dynamics model, a carbon cycle neural network model and a minimum spanning tree algorithm. This method breaks through the single measurement limitation of traditional carbon emission assessments, integrates microbial decomposition dynamics, environmental factor response mechanisms and crop growth parameters into a unified analysis framework, and effectively captures the complex interactions between different factors through a multi-head cross-attention mechanism. In particular, the spatiotemporal feature extraction capability of the carbon cycle neural network model and the network optimization function of the minimum spanning tree algorithm are utilized to successfully solve the measurement problem of the complex carbon flow path and variable influencing factors in the tobacco-rice rotation system.

[0017] In summary, the present invention successfully solves the technical problem of accurately quantifying and predicting the impact of different straw return treatment methods on soil carbon emissions in tobacco-rice rotation by constructing a multi-dimensional and multi-scale method for determining the carbon emission impact of straw return, providing a scientific basis and technical support for the transformation of farmland carbon sources into carbon sinks and sustainable agricultural development. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of the method of the present invention;

[0019] Figure 2 The average daily temperature and daily precipitation during the 2024 trial period;

[0020] Figure 3 This is the effect of different straw return methods on soil CO2 emission flux in tobacco-rice rotation;

[0021] Figure 4 This is a diagram showing the effects of different straw return methods on the cumulative CO2 emissions from tobacco-rice rotation soils;

[0022] Figure 5 This is the effect of different straw return methods on soil CH4 emission flux in tobacco-rice rotation;

[0023] Figure 6 This is the effect of different straw return methods on the cumulative CH4 emissions from tobacco-rice rotation soil;

[0024] Figure 7 This is the effect of different straw return methods on soil N2O emission flux in tobacco-rice rotation;

[0025] Figure 8 This is the effect of different straw return methods on the cumulative N2O emissions from tobacco-rice rotation soil;

[0026] Figure 9 are environmental factors that affect soil carbon emissions in tobacco fields;

[0027] Figure 10are environmental factors that affect carbon emissions from paddy soil;

[0028] Figure 11 This is a pillar diagram of net ecosystem carbon harvest in tobacco-rice rotation farmland;

[0029] Figure 12 Bar chart showing factor analysis for tobacco-rice rotation farmland. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0031] like Figure 1 FIG. 1 is a flow chart of a method for determining the effect of straw return on soil carbon emissions in tobacco-rice rotation provided by the present invention. The method comprises the following steps:

[0032] S01. Select the tobacco-rice rotation experimental field and divide the experimental area into several areas of 120m 2 The plots were arranged in random blocks, with 3 replicates for each treatment.

[0033] S02. Experimental treatments for returning straw to the field were set up, including a control group without returning straw to the field, a group with returning straw to the field alone, and a group with returning straw to the field plus a rot-promoting agent and urea. The amount of straw returned to the field was 4000 kg·hm -2 The amount of the rot-promoting agent added is 30 kg·hm -2 , urea addition amount is 75kg·hm -2 ;

[0034] S03. Install a static gas box measurement system, including an acrylic box, a stainless steel base, an automatic temperature and humidity recorder, and a gas collection component. Set up static gas boxes above and below the ridges.

[0035] S04. Collect gas samples every 7 days. During the fertilization period, increase the sampling frequency to once every 1 to 2 days for one week. The sampling time is from 8:00 to 11:00 in the morning.

[0036] S05. During sampling, record the temperature and humidity data in the box, and collect 30 mL of gas samples at 0 minutes, 10 minutes, 20 minutes, and 30 minutes after placement;

[0037] S06. Use gas chromatography to measure the CO2 concentration and CH4 concentration in the gas sample, and calculate the soil carbon emission flux and soil carbon emission accumulation based on the measurement data;

[0038] S07. Apply soil microbial kinetic equations to analyze the straw decomposition process, calculate microbial carbon turnover rate, substrate utilization efficiency, carbon-nitrogen ratio response factor, and temperature-humidity response factor, and construct a prediction model for carbon emissions from straw decomposition;

[0039] S08. Analyze the dynamic changes of soil carbon pools under different straw return conditions based on a pre-trained carbon cycle neural network model, and integrate the comprehensive effects of environmental factors, crop growth parameters, and microbial community structure on soil carbon emission fluxes through a multi-head cross-attention mechanism;

[0040] S09. Use the minimum spanning tree algorithm to construct the carbon flow network topology structure, identify key carbon flow nodes, optimize farmland carbon management strategies, realize the transformation of farmland carbon source to carbon sink, and provide a scientific basis for sustainable agricultural development.

[0041] Among them, the static gas box is a gas collection device consisting of a sealed box and a stainless steel base. The box material is acrylic, and the outer layer of the box is wrapped with insulation material to prevent excessive temperature changes in the box. The side of the box is connected to a hose and a three-way valve, and an automatic temperature and humidity recorder is installed on the top of the box.

[0042] The calculation formula for soil carbon emission flux is: Where F is the greenhouse gas emission rate, in mg·m -2 ·h -1 , ρ is the density of the gas under standard conditions, in kg·m -3 , h is the height of the static air box, in m, is the rate of change of greenhouse gas concentration in the static air box per unit time, in mg·m -2 ·h -1 , T is the average temperature in the box during gas sampling, in °C.

[0043] The calculation formula for soil carbon emission accumulation is CE=∑[(F i +F i+1 ) / 2×10 -3 ×d×24], where CE is the cumulative gas emissions in kg·hm -2 , F i and F i+1 is the gas emission flux between two consecutive adjacent sampling periods, in mg·m -2 ·h -1 , d is the number of days between two consecutive sampling times.

[0044] The soil microbial dynamics equation is: The soil microbial kinetic equation is used to describe the quantitative relationship between microbial activity and carbon emissions during the decomposition of straw organic carbon. The input includes the initial carbon content C0, the temperature response function f(T), the moisture response function f(W), the carbon-nitrogen ratio response function f(CN), and the microbial biomass carbon response function f(MBC). The output is the carbon decomposition rate per unit time. The initial carbon content C0 is calculated by the amount of straw returned to the field and the percentage of straw carbon content; the temperature response function is Calculate, where Q 10 is the temperature sensitivity coefficient, T ref is the reference temperature, the temperature parameter of the temperature response function is derived from the temperature data recorded by the temperature and humidity automatic recorder; the moisture response function f(W) is calculated from the humidity data recorded by the temperature and humidity automatic recorder; the carbon-nitrogen ratio response function f(CN) is calculated from the carbon-nitrogen ratio of the straw; and the microbial biomass carbon response function f(MBC) is obtained by measuring the carbon content of soil microbial biomass.

[0045] The net ecosystem carbon budget is calculated as NECB = ∑C input -∑C output =(C straw +C fertilizer +NPP)-(R h +CH4+C harvest ), where C starw 、C fertilizer and C harvest They refer to the carbon content of returned straw, the carbon content of applied fertilizer and the carbon content of harvested crops, NPP is the net primary productivity, R h It is soil heterotrophic respiration. The carbon content of the returned straw is calculated by the amount of straw returned to the field and the percentage of the straw carbon content. The carbon content of the applied fertilizer is calculated by the amount of urea added and the percentage of the urea carbon content. The carbon content of the harvest is calculated by the dry weight of the crop harvest and its carbon content percentage.

[0046] Among them, the calculation formula of net primary productivity is NPP=NPP grain +NPP straw +NPP root +NPP rhizodeposit , where NPP grain 、NPP straw 、NPP root and NPP rhizodeposit represents the carbon content of tobacco leaves or rice grains, the carbon content of stems, the carbon content of roots, and the carbon content of rhizosphere exudates, respectively, in kg C·hm -2The carbon content of tobacco leaves or rice grains is obtained by calculating the dry weight of harvested tobacco leaves or rice grains and the percentage of their carbon content; the carbon content of the stems is obtained by calculating the dry weight of the crop stems and the percentage of their carbon content; the carbon content of the roots is obtained by collecting crop root samples, measuring the dry weight and the percentage of their carbon content; the carbon content of the rhizosphere secretions is obtained by calculating the root carbon content multiplied by the rhizosphere secretion coefficient.

[0047] Among them, the decay-promoting agent refers to an additive containing microbial agents, with an effective viable bacterial count of 5×10 7 ·g -1 or higher, to accelerate the decomposition process of straw in the soil.

[0048] The minimum spanning tree algorithm, a graph-theoretic algorithm used to find a spanning tree connecting all nodes with the minimum sum of edge weights, is used in the impact determination method to construct a carbon cycle network for farmland ecosystems. Farmland carbon pools and carbon flows are considered network nodes and edges. The minimum spanning tree is used to identify key carbon cycle pathways and optimize farmland management strategies to improve carbon sequestration efficiency. The farmland carbon pools include active organic carbon pools, stable organic carbon pools, microbial carbon pools, and rhizosphere carbon pools; carbon flows include microbial decomposition, plant photosynthesis, plant respiration, and soil heterotrophic respiration. In prior art, active organic carbon pools refer to organic carbon components in the soil that are easily decomposed and utilized by microorganisms, primarily including soluble organic carbon, microbial metabolites, and fresh plant residues. This carbon component has a rapid turnover rate, typically completing decomposition within days to months, and has a direct impact on soil respiration and carbon emissions. In tobacco-rice rotation systems, during the initial straw return period, a large amount of easily decomposable organic matter enters the active carbon pool, leading to a short-term increase in soil carbon emissions. The active organic carbon pool, sensitive to environmental factors such as temperature and humidity, is the most active component of the soil carbon cycle and a key indicator of soil quality and fertility. The stable organic carbon pool refers to organic carbon components in the soil that are difficult to decompose by microorganisms and have a long turnover cycle. These components primarily include humus, organic carbon bound to minerals, and organic carbon protected by aggregates. This carbon pool has a turnover period of years to millennia and plays a key role in soil carbon sequestration. In tobacco-rice rotation systems, long-term straw incorporation can promote the accumulation of the stable organic carbon pool, especially when supplemented with rot-promoting agents and urea, which accelerate the conversion of organic matter to humus and improve carbon sequestration efficiency. The stable organic carbon pool responds more slowly to climate change, but its changes have long-term impacts on the carbon balance of agricultural ecosystems. The microbial carbon pool refers to the total amount of carbon contained in living microorganisms (including bacteria, fungi, and actinomycetes) in the soil and is one of the most active carbon pools in the soil. Although microbial carbon pools only account for 2-5% of total soil organic carbon, they play a central role in organic matter decomposition, nutrient transformation, and carbon cycling. During straw incorporation, the addition of decomposition-promoting agents can significantly increase the size and activity of the microbial carbon pool, improving substrate utilization efficiency. The turnover rate of the microbial carbon pool is influenced by factors such as the carbon-nitrogen ratio, temperature, and humidity. Its dynamic changes directly reflect the regulatory role of soil biological processes on the carbon cycle and are an important parameter for predicting soil carbon emissions. The rhizosphere carbon pool refers to the sum of carbon components within a specific region surrounding plant roots, including root exudates, rhizosphere microorganisms and their metabolites, and shed root cells. The rhizosphere is a key pathway for soil carbon input, with approximately 20-40% of the carbon fixed by plants through photosynthesis released into the soil in the form of root exudates. In tobacco-rice rotation systems, the root characteristics of rice and tobacco differ significantly, leading to seasonal variations in the composition and turnover rate of the rhizosphere carbon pool.The rhizosphere carbon pool is closely linked to the microbial carbon pool, forming a "root-microbe" interactive system. This system plays a vital role in nutrient cycling and soil aggregate formation, while also influencing soil carbon sequestration capacity and greenhouse gas emissions. Managing the rhizosphere carbon pool is crucial for optimizing farmland carbon cycling and enhancing its carbon sink function.

[0049] Among them, the specific structure of the carbon cycle neural network model is a hybrid architecture combining a multi-layer bidirectional long short-term memory network and a graph convolutional network, which includes a time series feature extraction layer, a spatial relationship encoding layer, a multi-head cross attention layer and a carbon cycle parameter prediction layer. The number of heads of the multi-head attention mechanism is determined by the number of soil carbon pools, and the attention weight is determined by the intensity of the impact of environmental factors on each carbon pool. The model input includes soil physical and chemical index time series data, microbial community structure characteristics and meteorological data, and the output is the predicted value of the dynamic change of soil carbon pool and soil carbon emission flux; the steps of establishing the training data set in the process of training the carbon cycle neural network model specifically include collecting soil carbon pools in different climates across the country. Long-term monitoring data of soil carbon emissions from typical tobacco-rice rotation in the region, standardized soil physical and chemical properties and soil carbon emission flux data, construction of a spatiotemporal sequence sample set, division of the training set and validation set in an 80:20 ratio, and data enhancement processing of the training set to improve the generalization ability of the model; the steps of training the carbon cycle neural network model specifically include initializing network structure parameters, using the Adam optimizer for parameter optimization, training the model with the mean square error loss function as the optimization target, setting an early stopping mechanism to avoid overfitting, using a learning rate decay strategy to improve the convergence stability of the model, using cross-validation to evaluate model performance, and saving the model parameters with the best test performance.

[0050] Among them, environmental factors include soil temperature, soil moisture, soil pH value, soil organic matter content, soil total nitrogen content, soil total phosphorus content, soil total potassium content, soil alkaline nitrogen content, soil available phosphorus content, soil available potassium content, daily average temperature, and daily precipitation; microbial community structure characteristics include soil bacterial abundance, soil fungal abundance, soil bacterial diversity index, soil fungal diversity index, soil bacterial phylum composition percentage, and soil fungal phylum composition percentage; crop growth parameters include crop plant height, crop stem diameter, crop leaf area index, crop biomass, and crop yield.

[0051] The specific implementation of the above steps is described in detail below.

[0052] Step S01 selects a tobacco-rice rotation test field and divides it into regions to provide a standardized experimental site for the subsequent straw return experiment. First, determine the appropriate tobacco-rice rotation area, which requires uniform soil texture, flat terrain, and good drainage and irrigation conditions. Then divide the test area into several 120m 2Each plot was rectangular, 12 meters long and 10 meters wide, with a 50-cm-wide isolation zone around each plot to prevent cross-influence between treatments. Treatments were randomly assigned to each plot using a randomized block arrangement method, with three replicates per treatment to eliminate random error and ensure the reliability and statistical significance of the results. This standardized experimental plot division and scientific arrangement method laid the spatial foundation for studying the effects of straw return on soil carbon emissions in tobacco-rice rotations.

[0053] Step S02 sets up the experimental treatment of straw return to the field and establishes a comparative experimental system under different straw treatment conditions. First, three treatment groups are set up: the control group without straw return to the field (CK), the group with simple straw return to the field (S), and the group with straw return to the field plus a urea-accelerating agent (SDU). The amount of straw returned to the field is uniformly 4000 kg·hm -2 , determined with reference to the average straw yield of local farmland; the amount of rot-promoting agent added is 30kg·hm -2 , select the effective bacteria count of 5×10 7 ·g -1 The compound microbial preparation; the amount of urea added is 75kg·hm -2 , providing additional nitrogen for straw decomposition. Before returning the straw to the field, the straw was chopped into 5-10 cm lengths to ensure uniform decomposition. In the straw return plus decomposition-promoting agent and urea treatment group, the straw was first evenly spread on the soil surface, then sprayed with the prepared decomposition-promoting agent solution and spread with urea, and finally tilled to a 10-15 cm soil layer. This step, by setting up a control group and different treatment groups, provides an experimental basis for studying the effects of different straw return methods on soil carbon emissions.

[0054] Step S03: Install a static gas chamber measurement system to construct a soil greenhouse gas emission monitoring device. The static gas chamber system consists of four components: an acrylic enclosure, a stainless steel base, a temperature and humidity recorder, and a gas collection component. The enclosure is made of 5mm-thick acrylic sheet, measuring 50cm × 50cm × 80cm, with a transparency greater than 85%. It is wrapped in a 1cm-thick layer of insulation to prevent drastic temperature fluctuations within the enclosure. A DHT22 temperature and humidity recorder with an accuracy of ±0.5°C and ±2% RH is installed on the top of the enclosure. A gas inlet is provided on the side of the enclosure, connected to a polytetrafluoroethylene hose and a three-way valve for gas collection. The stainless steel base is 15cm high and inserted into the soil to a depth of 10cm. A rubber seal is placed between the enclosure and the system to ensure airtightness. During the tobacco season only, static gas chambers are installed above and below the ridges, with six chambers installed per treatment plot. During the rice season, three chambers are installed per plot, for a total of 36 chambers. This step provides technical support for accurately measuring soil carbon emission fluxes by constructing a standardized gas collection system.

[0055] Step S04 determines the gas sampling frequency and establishes a sampling scheme that adapts to the dynamic characteristics of farmland carbon emissions. According to the characteristics of the tobacco-rice rotation growing season, a sampling strategy combining conventional and encrypted sampling is adopted. The conventional sampling frequency is once every 7 days, covering the entire growing season; during the fertilization period, an encrypted sampling strategy is adopted, with a frequency of once every 1 to 2 days and lasting for a week to capture the emission peak after fertilization. The sampling time is fixed at 8 am to 11 am, when the temperature is moderate and can represent the average emission level throughout the day. Agricultural activities such as irrigation and fertilization are stopped 24 hours before sampling to avoid human interference. Before each sampling, check the integrity of the gas box system, including sealing, working status of temperature and humidity recorders, etc. This step ensures that the collected data can truly reflect the temporal dynamic change characteristics of soil carbon emissions through scientific and reasonable sampling frequency design, especially capturing emission peaks after key agricultural activities such as fertilization.

[0056] Step S05 implements the gas sampling process to obtain gas concentration data for carbon emission calculation. Before sampling, prepare a 50 ml airtight syringe, a vacuum gas sampling bottle, and a gas label. During sampling, first record the temperature and humidity data in the box, accurate to 0.1°C and 0.1% RH, to provide parameters for subsequent temperature correction. Immediately after the gas box is sealed, collect 30 mL of gas sample at the 0th minute, and then collect gas samples of the same volume at the 10th minute, 20th minute, and 30th minute, for a total of 4 time points. The collected gas is immediately injected into a pre-vacuumed sample bottle to ensure that the sampling pipeline is contaminant-free. Each sample is labeled with the processing number, repeat number, sampling time point, and date information. After the sample collection is completed, it should be sent to the laboratory for analysis within 12 hours. If the time exceeds the time, it should be stored in a 4°C environment with a shelf life of no more than 7 days. This step ensures that the obtained gas samples accurately represent the soil carbon emission characteristics through a standardized gas sampling process, providing reliable basic data for the subsequent calculation of carbon emission fluxes.

[0057] Step S06: Gas sample analysis and carbon emission calculation to determine soil carbon emission flux and cumulative amount. Agilent 7890B gas chromatograph was used to measure CO2 and CH4 concentrations in the gas sample. The gas chromatography conditions were: FID detector, temperature 250°C; column temperature 60°C; carrier gas was high-purity nitrogen, flow rate 30ml / min -1 The injection volume was 1 mL. The linear regression slope was calculated based on the changes in gas concentration at the four time points, and the regression coefficient R was determined. 2 Is it greater than 0.90? If it is lower than this threshold, the sample data will be discarded. The carbon emission flux is calculated using the formula F = ρ × h × dc / dt × 273 / (273 + T), where ρ is the gas density under standard conditions, and CO2 is 1.96 kg·m -3 , CH4 is 0.71 kg·m -3; h is the static air box height; dc / dt is the concentration change slope; T is the average temperature at the time of sampling. The cumulative carbon emissions are calculated by the formula CE=∑[(F i +F i+1 ) / 2×10 -3 The carbon fluxes from two consecutive samplings were calculated using the equation [×d×24]. The average of the emission fluxes from two consecutive samplings was multiplied by the time interval to obtain the periodic emission value. The total emission value was then accumulated across all periodic emission values. This step, using standardized analytical methods and calculation formulas, converted the gas concentration data into carbon emission flux and accumulation indicators, quantifying the effects of different straw return treatments on soil carbon emissions.

[0058] Step S07 applies the soil microbial kinetic equation to analyze the straw decomposition process and establishes a correlation model between biological mechanisms and carbon emissions. First, soil microbial parameters are measured, including microbial carbon turnover rate and substrate utilization efficiency, and microbial biomass carbon is measured using the chloroform fumigation method. Application formula A microbial dynamics model was constructed to describe the relationship between straw organic carbon decomposition and carbon emission. The initial carbon content C0 was determined by the amount of straw returned to the field (4000 kg·hm -2 ) multiplied by the percentage of straw carbon content (about 40%). Temperature response function In, Q 10 The value is set to 2.0~2.5, T ref The soil moisture response function, f(W), reaches its optimal value of 1 when the soil moisture content is 40% to 60%, and is less than 1 below or above this range. The carbon-nitrogen ratio response function, f(CN), reaches its optimal value when the straw carbon-nitrogen ratio is 25 to 30. Both high and low carbon-nitrogen ratios reduce the decomposition rate. Model parameters were fitted using nonlinear regression, and the parameters were adjusted until the root mean square error between the model predictions and the measured carbon emissions was less than 10%. This step, through the construction of a microbial kinetic model, reveals the quantitative relationship between microbial activity and carbon emissions during straw decomposition, providing a mechanistic basis for carbon emission prediction.

[0059] Step S08 analyzes the dynamic changes of soil carbon pools based on the carbon cycle neural network model to achieve carbon emission prediction. This step first imports a pre-trained carbon cycle neural network model, which adopts a hybrid architecture combining a multi-layer bidirectional long short-term memory network with a graph convolutional network, and has the ability to process time series data and spatial correlation data. Input parameters include environmental factors, crop growth parameters, and microbial community structure data; the output is the predicted dynamic changes of soil carbon pools and carbon emission fluxes. The model performance is optimized by a multi-head cross-attention mechanism, and the number of attention heads is set to 4, corresponding to the active organic carbon pool, stable organic carbon pool, microbial carbon pool, and rhizosphere carbon pool. The model prediction accuracy is determined by the coefficient of determination R 2 Evaluation, requiring R 2A value greater than 0.85 is required for prediction. Using input parameters under different straw incorporation conditions, the dynamic changes in soil carbon pools over 100 days, including carbon pool size, conversion rate, and carbon emission flux, were predicted. The predicted results were validated against measured data, with a relative error within ±15%. This step integrates multi-source data through a deep learning model to accurately predict the dynamic changes in soil carbon pools under different straw incorporation conditions, providing decision support for farmland carbon management.

[0060] Step S09 uses a minimum spanning tree algorithm to construct a carbon flow network topology and optimize farmland carbon management strategies. This step considers the farmland carbon cycle system as a network structure, with nodes representing carbon pools, edges representing carbon transfer processes, and edge weights representing carbon flux. A minimum spanning tree is constructed using the Prim algorithm, starting with the soil organic carbon pool and sequentially connecting the microbial carbon pool, crop carbon pool, and atmospheric carbon pool. Edge weights are defined as the energy consumption or time cost required for carbon conversion. By identifying key carbon flow nodes, the rate-limiting steps and critical control points of the carbon cycle in the farmland system are determined. Key node identification criteria include nodes with a degree centrality greater than 0.5 or a betweenness centrality greater than 0.3. Optimization strategies are designed for key nodes. For example, returning straw to the field and treating it with a rot-promoting agent can improve the conversion efficiency of microbial carbon pools to stable organic carbon, increasing carbon sequestration by 10% to 15%. Network analysis calculates the carbon source and sink status of the farmland system. If the net ecosystem carbon budget is greater than zero, the system is a carbon sink; if it is less than zero, the system is a carbon source. This step uses network topology analysis methods to reveal the internal structure and key control points of the farmland carbon cycle system, providing a scientific basis and optimization strategy for the transformation of farmland carbon sources into carbon sinks.

[0061] The carbon cycle neural network model utilizes a hybrid architecture combining a multi-layer bidirectional long short-term memory network with a graph convolutional network. It comprises four key layers: a temporal feature extraction layer, a spatial relationship encoding layer, a multi-head cross-attention layer, and a carbon cycle parameter prediction layer. The temporal feature extraction layer consists of three layers of bidirectional LSTM units, each with 128 neurons, to capture the time series characteristics of dynamic changes in soil carbon pools. The spatial relationship encoding layer utilizes a two-layer graph convolutional network, each with 64 convolution kernels, to process the spatial correlations between different carbon pools. The multi-head cross-attention layer has four attention heads, each focusing on the interactions between different carbon pools and environmental factors. Attention weights are determined through backpropagation optimization. The carbon cycle parameter prediction layer consists of a two-layer fully connected network, with the number of neurons in the output layer matching the prediction target dimension. The model uses a residual connection structure to mitigate the vanishing gradient problem, and a batch normalization layer to improve training stability. The training dataset was constructed by collecting long-term soil carbon emission data from tobacco-rice rotations at 23 monitoring stations across five climatic regions across China, covering a period of more than 10 years. Soil physical and chemical indicators and carbon emission flux data were standardized, outliers removed, and missing values filled in. A spatiotemporal sequence sample set was constructed, each containing 30 consecutive days of time series data and corresponding spatial features. The training and validation sets were divided into a training set and a validation set with an 80:20 ratio. Data augmentation was performed on the training set, including the addition of Gaussian noise, sliding time windows, and feature combination transformations, expanding the training sample size to three times the original data size to improve model generalization. The model was trained using the Adam optimizer with an initial learning rate of 0.001, which decayed by 10% every 50 epochs. The loss function was the mean squared error (MSE). The training batch size was 64, the maximum training epoch was 500, and an early stopping mechanism was implemented, stopping training if no performance improvement on the validation set was observed after 20 consecutive epochs. Model performance was evaluated using 5-fold cross-validation, and the model parameters with the best validation set performance were selected for carbon emission prediction.

[0062] The mathematical model or calculation process involved in the present invention is described in detail below.

[0063] In step S06, the soil carbon emission flux calculation formula is specifically expressed as follows:

[0064]

[0065] Where F is the greenhouse gas emission rate, in mg·m -2 ·h -1 ; ρ is the density of the gas under standard conditions, CO2 is 1.96 kg·m -3 , CH4 is 0.71 kg·m -3 ;h is the height of the static air box, in meters; is the rate of change of greenhouse gas concentration in the static air box per unit time, in mg·kg -1 ·h -1; T is the average temperature inside the chamber during gas sampling, in °C.

[0066] The parameter acquisition method is:

[0067] The ρ value is the gas density constant under standard conditions. The density values of CO2 and CH4 are 1.96 kg·m -3 and 0.71 kg·m -3 ; h is obtained by directly measuring the static air box height, accurate to 0.01m; It is obtained by linear regression calculation of gas concentration versus time. The specific steps are: collect gas samples at 0, 10, 20, and 30 minutes, measure the gas concentration at each time point, and perform linear regression with time as the independent variable and concentration as the dependent variable. The slope of the regression equation is value, the regression coefficient R is required 2 Greater than 0.90; T is calculated by taking the average value of the temperature data recorded by the temperature and humidity automatic recorder during the sampling process, accurate to 0.1℃.

[0068] This formula converts the gas concentration change rate measured under standard conditions (273K, 101.325kPa) into the emission flux under actual sampling temperature conditions based on the principle of the ideal gas state equation. The term is used for temperature correction, accounting for the effect of temperature on gas volume. Because rising temperatures cause gas to expand, the same mass of gas occupies a larger volume, which affects the concentration calculation. Compared to traditional calculation methods, this formula incorporates a temperature correction factor, making the measurement results more accurately reflect actual emissions.

[0069] In step S06, the calculation formula for the cumulative amount of soil carbon emissions is specifically expressed as follows:

[0070]

[0071] Where CE is the cumulative gas emissions, in kg·hm -2 ; F i and F i+1 is the gas emission flux between two consecutive adjacent sampling periods, in mg·m -2 ·h -1 ;d i is the number of days between two consecutive sampling times; n is the total number of sampling times; 10 -3 is the unit conversion factor, converting mg to g; 24 is the number of hours, converting hours to days.

[0072] The parameter acquisition method is:

[0073] F i and F i+1 Calculated by the above soil carbon emission flux formula; di The difference in days between two consecutive sampling dates is calculated. For conventional sampling, d i For 7 days, for encrypted sampling i For 1 or 2 days.

[0074] This formula is based on the principle of trapezoidal integration, treating the emissions between two consecutive sampling periods as the area of a trapezoid. The total emissions are then calculated by summing the emissions over all time intervals. This formula uses a trapezoidal method rather than a rectangular method to calculate the integral, allowing for a more accurate estimation of emission dynamics, particularly in situations where emissions fluctuate significantly. The average emission flux is calculated based on the term, which reduces the error caused by the sampling interval compared to directly integrating the single-point flux data.

[0075] In step S07, the soil microbial kinetic equation is specifically expressed as follows:

[0076]

[0077] Where, is the carbon decomposition rate per unit time, in kg·C hm -2 ·d -1 ; k is the basic decomposition rate constant, ranging from 0.01 to 0.05d -1 ; C0 is the initial carbon content, unit is kg·C hm -2 ; f(T) is the temperature response function, dimensionless; f(W) is the moisture response function, dimensionless; f(CN) is the carbon-nitrogen ratio response function, dimensionless; f(MBC) is the microbial biomass carbon response function, dimensionless.

[0078] Among them, the temperature response function f(T) is specifically expressed as:

[0079]

[0080] Where Q 10 is the temperature sensitivity coefficient, ranging from 2.0 to 2.5, dimensionless; T is the measured soil temperature, in °C; T ref is the reference temperature, which is 20℃.

[0081] The moisture response function f(W) is specifically expressed as:

[0082]

[0083] Where, W is the soil moisture content, in %; W opt The lower limit of the optimum water content is 40%. max The upper limit of the optimum moisture content is 60%.

[0084] The carbon-nitrogen ratio response function f(CN) is specifically expressed as:

[0085]

[0086] Where, CN is the carbon-nitrogen ratio of straw, dimensionless; opt The lower limit of the optimum carbon-nitrogen ratio is 25; max is the upper limit of the optimum carbon-nitrogen ratio, which is 30; α is the attenuation coefficient, which is 0.1.

[0087] The microbial biomass carbon response function f(MBC) is specifically expressed as:

[0088] f(MBC)=1-exp(-β×MBC);

[0089] Where MBC is the soil microbial biomass carbon content, in mg kg -1 ; β is the microbial activity coefficient, which is 0.005 kg·mg -1 .

[0090] The parameter acquisition method is:

[0091] C0 is calculated by the amount of straw returned to the field and the percentage of straw carbon content. The calculation formula is C0 = M straw ×C content ×10 -2 , where M straw is the amount of straw returned to the field, fixed at 4000 kg·hm -2 , C content is the percentage of carbon content in straw, which is about 40%; T is recorded by an automatic temperature and humidity recorder; W is obtained by measuring the soil moisture content, which is determined by the drying method. Fresh soil samples (dried at 105°C for 48 hours) are weighed and the weight loss percentage is calculated; CN is calculated by measuring the carbon content and nitrogen content of straw. The carbon content is determined by potassium dichromate oxidation method, and the nitrogen content is determined by Kjeldahl method; MBC is determined by chloroform fumigation method and the difference in extractable organic carbon content before and after fumigation is calculated; k, Q 10 The , α, and β parameters are obtained by fitting the measured data using the nonlinear regression method.

[0092] The kinetic equation combines the Arrhenius equation and the Michaelis-Menten kinetic principle, and comprehensively considers the effects of temperature, moisture, substrate quality, microbial activity and other factors on straw decomposition. The temperature response function f(T) is based on Q 10Theory describes the multiple increase in decomposition rate for every 10°C increase in temperature. The exponential form reflects the nonlinear growth relationship of microbial activity with increasing temperature. The moisture response function f(W) adopts a piecewise function form to describe the inhibition of microbial activity under water shortage and over-humidity conditions. The carbon-nitrogen ratio response function f(CN) takes into account the limitation of carbon-nitrogen imbalance on decomposition. When the carbon-nitrogen ratio is too high, nitrogen becomes a limiting factor and inhibits the decomposition process. The exponential decay is used to represent the strong inhibitory effect of high carbon-nitrogen ratio. The microbial biomass carbon response function f(MBC) uses an exponential saturation curve to describe the relationship between microbial abundance and decomposition rate, reflecting the influence of microbial community size on the decomposition process.

[0093] In step S07, the net ecosystem carbon budget calculation formula is specifically expressed as follows:

[0094] NECB=∑C input -∑C output =(C straw +C fertilizer +NPP)-(R h +CH4+C harvest );

[0095] Where NECB is the net ecosystem carbon budget, expressed in kg·C hm -2 ; C straw is the carbon content of returned straw, in kg·C hm -2 ; C fertilizer is the carbon content of the applied fertilizer, in kg·C hm -2 ; NPP is net primary productivity, unit is kg·C hm -2 ; R h is soil heterotrophic respiration, unit is kg·C hm -2 ; CH4 is methane carbon emissions, unit is kg·Chm -2 ; C harvest is the carbon content of the harvest, in kg·C hm -2 .

[0096] The parameter acquisition method is:

[0097] C straw It is calculated by the amount of straw returned to the field and the percentage of straw carbon content. The calculation formula is C straw =M straw ×C content ×10 -2 , where M straw is the amount of straw returned to the field, fixed at 4000 kg·hm -2 , C content is the percentage of carbon content in straw, which is about 40%; C fertilizerIt is calculated by the amount of urea added and the percentage of urea carbon content. The calculation formula is C fertilizer =M urea ×C urea ×10 -2 , where M urea is the amount of urea added, fixed at 75 kg·hm -2 , C urea is the percentage of carbon content in urea, which is about 20%; R h The total soil respiration was calculated by subtracting the autotrophic respiration from the total soil respiration. The total soil respiration was calculated by measuring CO2 emissions using the static gas box method, and the autotrophic respiration was estimated by the root exclusion method. CH4 was calculated by measuring methane emissions using the static gas box method. C harvest The dry weight of the harvested material and its carbon content percentage were calculated. The harvested tobacco or rice part was sampled and dried (80°C, 48 hours) to determine the dry weight. The carbon content was determined by an elemental analyzer.

[0098] The calculation formula of net primary productivity (NPP) is as follows:

[0099] NPP=NPP grain +NPP straw +NPP root +NPP rhizodeposit ;

[0100] Where, NPP grain is the carbon content of tobacco leaves or rice grains, in kg·C hm -2 ;NPP straw is the carbon content of the stem, in kg·C hm -2 ;NPP root is the root carbon content, unit is kg·C hm -2 ;NPP rhizodeposit is the carbon content of rhizosphere exudates, in kg·C hm -2 .

[0101] The parameter acquisition method is:

[0102] NPP grain It is obtained by harvesting tobacco leaves or rice grains, measuring dry weight and carbon content, and the calculation formula is NPP grain =Y grain ×C grain ×10 -2 , where Y grain is the tobacco leaf or rice grain yield, in kg·hm -2 , C grain The percentage of carbon content is about 45% in tobacco leaves and about 42% in rice grains; NPP strawIt is obtained by harvesting the stems, measuring the dry weight and carbon content, and the calculation formula is NPP straw =Y straw ×C straw ×10 -2 , where Y straw is the stalk yield, in kg·hm -2 , C straw The percentage of carbon content is about 40%; NPP root It is obtained by collecting root samples, measuring dry weight and carbon content. The calculation formula is NPP root =Y root ×C root ×10 -2 , where Y root is the root production, in kg·hm -2 , C root The carbon content percentage is about 38%; NPP rhizodeposit It is calculated by multiplying the root carbon content by the rhizosphere exudate coefficient. The calculation formula is NPP rhizodeposit =NPP root ×K rhizo , where K rhizo is the rhizosphere secretion coefficient, which is 0.5 for tobacco and 0.7 for rice.

[0103] The Net Ecosystem Carbon Budget (NECB) formula is based on the principle of carbon balance and comprehensively considers the carbon input and carbon output processes of the farmland system. The carbon input items in the formula include returning straw to the field, applying fertilizers and crop photosynthetic carbon sequestration, and the carbon output items include soil heterotrophic respiration, methane emissions and crop harvest removal. If NECB is greater than zero, it means that the system is a carbon sink, and if NECB is less than zero, it means that the system is a carbon source. The formula uses addition and subtraction to calculate the carbon budget, which intuitively reflects the contribution of each carbon flow process to the carbon balance of the system, helps to identify key carbon flow pathways and optimize carbon management measures. The net primary productivity (NPP) calculation adopts the group addition method to calculate the carbon sequestration of different plant organs separately, and especially considers the rhizosphere secretions, which is easily overlooked as a carbon input pathway, to make the carbon budget calculation more comprehensive and accurate.

[0104] In step S08, the multi-head cross attention weight calculation formula is specifically expressed as follows:

[0105]

[0106] Where A i is the output of the i-th attention head, representing the i-th carbon pool feature; Q i , K i 、V i They are query matrix, key matrix and value matrix respectively, with dimensions of n×d, where n is the number of samples and d is the feature dimension; dk is the dimension of the key vector; softmax is a normalization function that ensures that the sum of the weights is 1.

[0107] The calculation formula of attention head fusion is specifically expressed as follows:

[0108] Z=Concat(A1,A2,...,A h )W O ;

[0109] Where Z is the output of the multi-head attention layer; A1, A2, ..., A h is the output of each attention head; h is the number of attention heads, which is set to 4, corresponding to the active organic carbon pool, stable organic carbon pool, microbial carbon pool and rhizosphere carbon pool; W O It is a projection matrix used to fuse multi-head features; Concat is a connection operation that concatenates multiple matrices by column.

[0110] The parameter acquisition method is:

[0111] The attention weights are obtained through back-propagation optimization. The specific steps are as follows: initialize the weight matrix to random small values; calculate the mean square error loss between the model prediction value and the true value; update the weight parameters using the gradient descent method, and the learning rate is initially set to 0.001; repeat the training iterative process until the loss function converges and the performance of the validation set no longer improves.

[0112] The multi-head cross attention mechanism is based on the Transformer architecture principle and calculates the correlation between different input features through self-attention. Calculate the feature similarity matrix and divide by The dot product is scaled to prevent vanishing gradients. The softmax function converts similarities into probability distributions, achieving feature weighting. The multi-head design enables the model to focus on the interactions between different carbon pools and environmental factors, improving feature representation. Compared with traditional feedforward networks, this mechanism can more effectively capture the complex nonlinear relationships between soil carbon pools and environmental factors, enabling accurate prediction of carbon emission fluxes.

[0113] Specifically, the present invention is based on the integration of system dynamics of the soil carbon cycle and machine learning techniques. First, a controllable experimental foundation was established by setting up control experiments and different straw return treatment groups. The static air chamber measurement system provides direct monitoring data on soil carbon emissions with high temporal and spatial resolution, ensuring the accuracy and representativeness of the underlying data.

[0114] At the data processing level, this paper introduces the soil microbial dynamics equation to analyze the carbon emission mechanism during straw decomposition, incorporating key influencing factors such as temperature, humidity, carbon-nitrogen ratio, and microbial biomass into a unified mathematical framework. This equation is based on the theory of substrate-microbe interaction and can accurately describe the decomposition dynamics of straw organic carbon under the action of microorganisms. It is a theoretical bridge connecting environmental conditions and carbon emissions. The influence weight of each factor on carbon decomposition rate was quantified.

[0115] The core innovation of this paper lies in the construction of a carbon cycle neural network model that combines a bidirectional long short-term memory network with a graph convolutional network. This hybrid architecture targets the spatiotemporal characteristics of soil carbon emissions. The temporal feature extraction layer captures the seasonal and cyclical variations in carbon emissions, the spatial relationship encoding layer addresses the interactions between different carbon pools, and the multi-head cross-attention layer implements the complex nonlinear mapping between environmental factors, microbial characteristics, and carbon emissions.

[0116] The application of the minimum spanning tree algorithm further optimizes the carbon flow network topology. By treating the farmland carbon cycle as a network system, it identifies key carbon flow nodes and provides a basis for accurate decision-making in farmland management. Based on graph theory principles, this algorithm minimizes energy loss during carbon flow conversion, enabling the system to automatically discover the optimal carbon management path.

[0117] In summary, the present invention realizes the full-chain carbon emission impact analysis from microscopic biological mechanisms to macroscopic farmland management through multi-level and multi-angle data collection and model construction.

[0118] A specific embodiment 1 of the present invention is provided below. The specific implementation of each step in this embodiment 1 is described in detail as follows.

[0119] The specific implementation of steps S01-S05 in this embodiment is the same as above and will not be described in detail here.

[0120] The specific implementation of step S06 in this embodiment is gas sample analysis and carbon emission calculation to determine the soil carbon emission flux and cumulative amount. The CO2 concentration and CH4 concentration in the gas sample were measured using an Agilent 7890B gas chromatograph. The gas chromatography conditions were: FID detector, temperature 250°C; column temperature 60°C; carrier gas was high-purity nitrogen, flow rate 30ml·min -1 The injection volume was 1 mL. The linear regression slope was calculated based on the changes in gas concentration at the four time points, and the regression coefficient R was determined. 2 Is it greater than 0.90? If it is lower than this threshold, the sample data will be discarded. The soil carbon emission flux calculation formula is: Where F is the greenhouse gas emission rate, in mg kg-1 ·h -1 ; ρ is the density of the gas under standard conditions, CO2 is 1.96 kg·m -3 , CH4 is 0.71 kg·m -3 ;h is the height of the static air box, in meters; is the rate of change of greenhouse gas concentration in the static air box per unit time, in mg·kg -1 ·h -1 , calculated by linear regression of gas concentration versus time; T is the average temperature in the chamber during gas sampling, in °C. The cumulative amount of soil carbon emissions is calculated using the formula: Calculate, where CE is the cumulative gas emissions, in kg·hm -2 ; F i and F i+1 is the gas emission flux between two consecutive adjacent sampling periods, in mg·m -2 ·h -1 ;d i is the number of days between two consecutive sampling times; n is the total number of sampling times; 10 -3 is the unit conversion factor; 24 is the number of hours. This method, based on the ideal gas state equation, converts the rate of change of gas concentration measured under standard conditions into emission flux under actual sampling temperature conditions. Cumulative emissions are calculated using the trapezoidal integration method. Compared to traditional calculation methods, a temperature correction factor is added to ensure that the measurement results more accurately reflect actual emissions.

[0121] The specific implementation of step S07 is to apply the soil microbial kinetic equation to analyze the straw decomposition process and establish a correlation model between biological mechanisms and carbon emissions. First, soil microbial parameters are measured, including microbial carbon turnover rate and substrate utilization efficiency. The microbial biomass carbon is measured using the chloroform fumigation method. Apply the microbial kinetic equation: Where, is the carbon decomposition rate per unit time, in kg·C hm -2 ·d -1 ; k is the basic decomposition rate constant, ranging from 0.01 to 0.05d -1 ; C0 is the initial carbon content, which is calculated by the amount of straw returned to the field and the percentage of straw carbon content; f(T) is the temperature response function, and the calculation formula is where Q 10 is the temperature sensitivity coefficient, ranging from 2.0 to 2.5, T ref is the reference temperature, which is 20°C; f(W) is the moisture response function, and the calculation formula is: Where W opt The lower limit of the optimum water content is 40%, W maxis the upper limit of the optimum moisture content, which is 60%; f(CN) is the carbon-nitrogen ratio response function, and the calculation formula is: Among them, CN opt The lower limit of the optimum carbon-nitrogen ratio is 25, and CN max is the upper limit of the optimal carbon-nitrogen ratio, which is 30; α is the attenuation coefficient, which is 0.1; f(MBC) is the microbial biomass carbon response function, and the calculation formula is: f(MBC) = 1-exp(-β×MBC), where β is the microbial activity coefficient, which is 0.005 kg·mg -1 The kinetic equation combines the Arrhenius equation and the Michaelis-Menten kinetic principle, comprehensively considering the effects of temperature, moisture, substrate quality, microbial activity and other factors on straw decomposition. It is fitted by nonlinear regression method and the parameters are adjusted until the root mean square error between the model prediction value and the measured carbon emission value is less than 10%. At the same time, combined with the net ecosystem carbon budget calculation formula: NECB = ∑C input -∑C output =(C straw +C fertilizer +NPP)-(R h +CH4+C harvest ) and the net primary productivity calculation formula: NPP = NPP grain +NPP straw +NPP root +NPP rhizodeposit , evaluate the carbon source and sink status of the farmland system under different straw return conditions, and provide a theoretical basis for farmland carbon management.

[0122] The specific implementation method of step S08 is to analyze the dynamic changes of soil carbon pools based on the carbon cycle neural network model to realize carbon emission prediction. This step first imports a pre-trained carbon cycle neural network model, which adopts a hybrid architecture combining a multi-layer bidirectional long short-term memory network with a graph convolutional network, and has the ability to process time series data and spatial correlation data. The input parameters include environmental factors, crop growth parameters, and microbial community structure data; the output is the predicted dynamic changes of soil carbon pools and carbon emission fluxes. The model performance is optimized by a multi-head cross-attention mechanism, and the number of attention heads is set to 4, corresponding to the active organic carbon pool, stable organic carbon pool, microbial carbon pool, and rhizosphere carbon pool. The multi-head cross-attention weight calculation formula is: Where A i is the output of the i-th attention head, representing the i-th carbon pool feature; Q i , K i 、V i are query matrix, key matrix and value matrix respectively; d kis the dimension of the key vector; softmax is the normalization function. The attention head fusion calculation formula is: Z = Concat(A1, A2, ..., A h )W O , where Z is the output of the multi-head attention layer; A1, A2, ..., A h is the output of each attention head; h is the number of attention heads; W O is the projection matrix. Compared with the traditional feedforward network, this mechanism can more effectively capture the complex nonlinear relationship between soil carbon pool and environmental factors. The attention weight is optimized by the back propagation algorithm, and the model prediction accuracy is determined by the coefficient of determination R. 2 Evaluation, requiring R 2 A value greater than 0.85 is required for prediction. Using input parameters under different straw incorporation conditions, the dynamic changes in soil carbon pools over 100 days, including carbon pool size, conversion rate, and carbon emission flux, were predicted. The predicted results were validated against measured data, with a relative error within ±15%. This step integrates multi-source data through a deep learning model to accurately predict the dynamic changes in soil carbon pools under different straw incorporation conditions, providing decision support for farmland carbon management.

[0123] The specific implementation of step S09 is to use the minimum spanning tree algorithm to construct the carbon flow network topology structure and optimize the farmland carbon management strategy. This step regards the farmland carbon cycle system as a network structure, with nodes as carbon pools, edges as carbon flow processes, and edge weights as carbon flow. The Prim algorithm is used to construct a minimum spanning tree, starting from the soil organic carbon pool, connecting the microbial carbon pool, crop carbon pool and atmospheric carbon pool in sequence, and the edge weight is defined as the energy consumption or time cost required for carbon conversion. The algorithm steps are: initialize the spanning tree set to an empty set, select the soil organic carbon pool node as the starting point to join the spanning tree set; iteratively perform the following operations until all nodes are added to the spanning tree set: find the edge with the smallest weight among all edges connected to the current spanning tree set, and add the node connected by the edge to the spanning tree set; the final edge set forms a minimum spanning tree, representing the carbon cycle path with the least energy consumption. By identifying the key nodes of carbon flow, the rate-limiting steps and key control points of the carbon cycle in the farmland system are determined. The key node judgment standard is a node with a degree centrality greater than 0.5 or a betweenness centrality greater than 0.3. The degree centrality calculation formula is: Where DC(i) is the degree centrality of node i; A ij is an element of the adjacency matrix, which is 1 if node i is connected to node j, otherwise it is 0; n is the total number of nodes. The betweenness centrality calculation formula is: Where BC(i) is the betweenness centrality of node i; σ st is the number of shortest paths from node s to node t; σ st(i) is the number of shortest paths from s to t that pass through node i. Optimization strategies are designed for key nodes. For example, returning straw to the field with a decomposition-promoting agent can improve the conversion efficiency of microbial carbon pools to stable organic carbon, increasing carbon sequestration by 10% to 15%. Network analysis is used to calculate the carbon source and sink status of the farmland system. If the net ecosystem carbon budget is greater than zero, it is a carbon sink; if it is less than zero, it is a carbon source. This step uses network topology analysis to reveal the internal structure and key control points of the farmland carbon cycle system, providing a scientific basis and optimization strategies for transforming farmland carbon sources into carbon sinks.

[0124] The carbon cycle neural network model of this embodiment adopts a hybrid architecture combining a multi-layer bidirectional long short-term memory network and a graph convolutional network. It includes four key layers: temporal feature extraction layer, spatial relationship encoding layer, multi-head cross attention layer, and carbon cycle parameter prediction layer. The temporal feature extraction layer consists of three layers of bidirectional LSTM units, each with 128 neurons, which are used to capture the time series characteristics of the dynamic changes of the soil carbon pool. The calculation formula of the LSTM unit is: f t =σ(W f ·[h t-1 , x t ]+b f ), i t =σ(W i ·[h t-1 , x t ]+b i ), o t =σ(W o ·[h t-1 , x t ]+b o ), h t =o t ×tanh(C t ), where f t 、i t 、o t They are forget gate, input gate and output gate respectively; C t is the cell state; h t is the hidden state; W and b are weight and bias parameters; σ is the sigmoid activation function; tanh is the hyperbolic tangent activation function. The spatial relationship encoding layer uses a two-layer graph convolutional network with 64 convolution kernels in each layer to process the spatial correlation between different carbon pools. The graph convolution calculation formula is: Among them H (l) is the node feature matrix of the lth layer; To add the adjacency matrix of self-connection; is the degree matrix; W (l)is the weight matrix of the first layer. The multi-head cross attention layer sets 4 attention heads, focusing on the interaction between different carbon pools and environmental factors. The carbon cycle parameter prediction layer consists of a 2-layer fully connected network. The calculation formula of the fully connected layer is: y = σ (Wx + b), where x is the input feature; W and b are weight and bias parameters; σ is the activation function, and the ReLU function is usually selected: ReLU (x) = max (0, x). The model uses a residual connection structure to alleviate the gradient disappearance problem. The residual connection formula is: H (l+1) =H (l) +F(H (l) ), where F(H (l) ) is the residual mapping function. The batch normalization layer improves training stability and is calculated as: where μ B and are the mean and variance of the batch data, respectively; γ and β are learnable scaling and offset parameters; ∈ is a smoothing term to prevent division by zero errors. The model is trained using the Adam optimizer, and the loss function is the mean squared error (MSE), calculated as: where y i is the true value; is the predicted value; n is the number of samples. Through the above model structure and training method, the carbon cycling neural network can effectively capture the complex patterns of dynamic changes in soil carbon pools, providing a powerful tool for accurately predicting the impact of straw return on soil carbon emissions.

[0125] In order to better understand and implement the present invention, Example 2 of a specific application scenario of the present invention is provided below: This example takes Changting County, Longyan City, Fujian Province as the research area, and selects local typical red soil tobacco-rice rotation farmland as the test site. The region has a subtropical monsoon climate, with an average annual temperature of 18.3°C, an annual precipitation of 1700mm, red soil type, pH value of 4.96, organic matter content of 30.53g / kg, total nitrogen content of 1.08g / kg, total phosphorus content of 0.68g / kg, and total potassium content of 20.75g / kg. It is implemented according to the steps described in "A method for determining the impact of a straw return method on tobacco-rice rotation soil carbon emissions".

[0126] First, select the test area and divide it into nine plots with an area of 120m 2 The plots were 12 meters long and 10 meters wide, with 50-cm-wide isolation strips around them. A randomized block arrangement was used, with three treatments, each replicated three times: a control group (CK, no straw return), a straw return group (SC), and a straw return plus a decomposition-promoting agent and urea group (SD).

[0127] Secondly, the straw return experiment was carried out. The amount of straw returned to the field in both the SC and SD treatment groups was 4000 kg·hm-2 The amount of accelerator added to the SD treatment group was 30 kg·hm -2 (The effective bacterial count reaches 5×10 7 ·g -1 The amount of urea added was 75 kg·hm -2 Before returning the straw to the field, the straw was chopped into 5-10 cm lengths. In the SD treatment, the straw was evenly spread on the soil surface, then the prepared decay-promoting agent solution was sprayed and urea was spread, and finally the soil was plowed to a 10-15 cm soil layer.

[0128] Thirty-six static gas chamber measurement systems were installed in the experiment, with four chambers per treatment plot (two above and two below the ridge). The static gas chamber system consisted of an acrylic chamber (50 cm × 50 cm × 50 cm), a stainless steel base (15 cm high, inserted 10 cm deep into the soil), a DHT22 automatic temperature and humidity recorder (accuracy ±0.5°C and ±2% RH), and gas collection components. The chambers were wrapped in 1 cm thick insulation material and had gas inlets on the sides connected to a polytetrafluoroethylene hose and a three-way valve.

[0129] Gas sampling followed a strategy that combined regular and intensified sampling. Regular sampling occurred every seven days throughout the growing season; during the fertilization period, intensified sampling was performed every one to two days for a week. Sampling was performed between 8:00 AM and 11:00 AM. Temperature and humidity data were recorded during each sampling session. Immediately after the chamber was sealed, a 30 mL gas sample was collected at the 0th minute, followed by samples of the same volume at the 10th, 20th, and 30th minutes.

[0130] The collected gas samples were measured using an Agilent 7890B gas chromatograph. The gas chromatography conditions were: FID detector, temperature 250°C; column temperature 60°C; carrier gas was high-purity nitrogen, flow rate 30 ml min -1 The injection volume was 1 mL. The linear regression slope was calculated based on the changes in gas concentration at the four time points, and the regression coefficient R was determined. 2 Is it greater than 0.90? Soil carbon emission flux and accumulation are calculated according to the formula and calculate.

[0131] Soil and crop samples were collected seasonally. Surface soil samples (0-20 cm) were used to determine soil physicochemical properties and microbiological parameters. Aboveground and belowground biomass samples were collected at tobacco and rice harvest for yield and carbon and nitrogen content analysis.

[0132] The results of the tobacco growing season experiment show that the soil CO2 emission flux under different treatments is shown in Table 1:

[0133] Table 1 Soil CO2 emission flux during tobacco growing season (mg·m -2 ·h -1 )

[0134] Sampling date CK SC SD February 21 (after applying base fertilizer) 433.85±20.72 629.86±22.09 590.65±32.14 March 6 268.75±19.81 454.47±11.23 424.81±6.48 March 21 235.24±20.71 518.08±52.32 495.69±58.76 April 7 (budding period) 707.18±44.44 1208.68±36.55 1165.97±63.47 April 24 278.47±25.79 462.04±35.54 426.54±12.30 May 8 (maturity) 253.63±35.19 312.32±50.81 298.26±21.89 May 22 147.87±22.51 191.64±50.38 186.61±40.20 June 5 130.03±23.80 222.99±11.27 208.18±19.71

[0135] The changes in soil CO2 emission flux during the rice growing season are shown in Table 2:

[0136] Table 2 Soil CO2 emission flux during rice growing season (mg·m -2 ·h -1 )

[0137]

[0138] Based on the above flux data, the calculated cumulative CO2 emissions are shown in Table 3:

[0139] Table 3 Cumulative CO2 emissions from soil in tobacco-rice rotation system (kg·hm -2 )

[0140]

[0141]

[0142] Note: Different letters in the same row indicate significant differences (P<0.05).

[0143] The soil CH4 emission flux during the tobacco growing season is shown in Table 4:

[0144] Table 4 Soil CH4 emission flux during tobacco growing season (mg·m -2 ·h -1 )

[0145] Sampling date CK SC SD February 21 (after applying base fertilizer) 0.64±0.11 0.58±0.08 0.58±0.09 March 6 -0.52±0.50 -0.49±0.28 -0.51±0.05 March 21 -0.13±0.05 -0.12±0.03 -0.12±0.04 April 7 (budding period) 0.49±0.09 0.51±0.09 0.51±0.11 April 24 0.07±0.02 0.09±0.02 0.09±0.02 May 8 0.19±0.03 0.19±0.03 0.19±0.04 May 22 0.34±0.03 0.35±0.03 0.33±0.02 June 5 0.06±0.01 0.06±0.06 0.05±0.04

[0146] The soil CH4 emission flux data during the rice growing season are shown in Table 5:

[0147] Table 5 Soil CH4 emission flux during rice growing season (mg·m -2 ·h -1 )

[0148]

[0149] The cumulative CH4 emissions from the tobacco-rice rotation system are shown in Table 6:

[0150] Table 6 Cumulative CH4 emissions from soil in tobacco-rice rotation system (kg·hm -2 )

[0151] growing season CK SC SD Tobacco Season 0.03±0.01b 0.18±0.04a -0.44±0.05c Rice season 93.75±3.95b 86.13±4.13c 105.76±3.88a annual 93.78±3.95b 86.30±4.09b 105.31±3.84a

[0152] Note: Different letters in the same row indicate significant differences (P<0.05).

[0153] According to the net ecosystem carbon budget calculation formula, the carbon budget data under different treatments are shown in Table 7:

[0154] Table 7 Net ecosystem carbon budget of tobacco-rice rotation system (kg·C hm -2 )

[0155]

[0156]

[0157] Note: Different letters in the same column indicate significant differences (P<0.05); C straw 、C fertilizer and C harvest They refer to the carbon content of returned straw, applied fertilizer and aboveground biomass, NPP is net primary productivity, R h is soil heterotrophic respiration, and NECB is net ecosystem carbon budget.

[0158] The effects of returning straw to fields on soil microbial parameters are shown in Table 8:

[0159] Table 8 Effects of different treatments on soil microbial parameters

[0160]

[0161] Note: Different letters in the same column indicate significant differences (P<0.05).

[0162] The prediction results of the dynamic changes of soil carbon pool by combining straw return treatment with the carbon cycle neural network model are shown in Table 9:

[0163] Table 9 Prediction of the effect of returning straw to the field on soil carbon pool after 100 days (g kg -1 )

[0164]

[0165] Note: Different letters in the same column indicate significant differences (P<0.05).

[0166] The carbon flow network topology structure analysis constructed by the minimum spanning tree algorithm identified the key control nodes in the straw return system and their centrality as shown in Table 10:

[0167] Table 10 Centrality of key nodes in carbon flow network under different treatments

[0168]

[0169]

[0170] Note: Different letters in the same column indicate significant differences (P<0.05).

[0171] Analysis of the results showed that both straw incorporation treatments significantly increased soil CO₂ emissions in tobacco-rice rotations, particularly during the tobacco growing season, with increases reaching 66.39% for the SC treatment and 58.37% for the SD treatment. However, the effects were smaller during the rice growing season, with increases of 3.21% for the SC treatment and 1.08% for the SD treatment. This suggests that straw incorporation has a greater impact on soil carbon emissions in dryland crops than in paddy fields. Compared to the SC treatment, the SD treatment, while promoting straw decomposition, reduced soil CO₂ emissions. This may be due to the fact that urea addition reduced the soil microbial C / N ratio, thereby alleviating the microbial "nitrogen starvation" effect.

[0172] From the perspective of net ecosystem carbon budget (NECB), both straw return treatments changed the tobacco-rice rotation system from a carbon balance state (NECB of CK treatment was 148.50 kg·C hm -2 ) into an obvious carbon sink (NECB of SC treatment was 1087.71 kg·Chm -2 , NECB of SD treatment is 1490.07 kg·C hm -2 The NECB of the SD treatment was 36.99% higher than that of the SC treatment, indicating that the straw return method with the addition of microbial decomposition promoters and urea is more conducive to improving the carbon sequestration capacity of the farmland system.

[0173] Traditional studies on the impact of straw return mainly focus on single indicators such as soil greenhouse gas emissions or organic matter content, ignoring the comprehensive assessment of the overall carbon balance of the system and the dynamic changes of the carbon pool. The present invention adopts a method that combines a static gas box with a carbon cycle neural network model. It not only measures the soil carbon emission flux, but also constructs a farmland carbon flow network topology structure, and evaluates the impact of straw return on the carbon source and sink function of farmland at the system level. Compared with traditional methods, the progress of the present invention is reflected in the following aspects: (1) the accuracy of carbon emission prediction is improved by integrating environmental factors, crop growth parameters and microbial community structure through a multi-head cross-attention mechanism; (2) the carbon flow network topology structure is constructed using the minimum spanning tree algorithm, and the key control points of the carbon cycle are identified, providing a theoretical basis for precise policy implementation; (3) based on the net ecosystem carbon budget accounting, the contribution of different straw return methods to the carbon sequestration capacity of the farmland system is quantitatively evaluated. The research results show that straw return treatment with the addition of microbial decomposition promoters and urea reduces carbon emissions while improving the system carbon sequestration capacity, providing a feasible technical solution for agricultural carbon emission reduction and improving the carbon sequestration function of farmland.

[0174] The following provides a specific embodiment 3 of the present invention to explore the effects of different straw return methods on soil carbon emissions in tobacco-rice rotation. Figure 2 As shown, researchers conducted a systematic field experiment at the tobacco experimental base in Changting County, Longyan City, Fujian Province from October 2023 to October 2024. The experimental base is located in a subtropical marine monsoon climate zone with an average annual temperature of 18.3°C, an average annual precipitation of about 1700 mm, and an average annual frost-free period of 260 days. The soil used in the experiment was red paddy soil. The main physical and chemical properties before the experiment began were: pH 4.97, organic matter content 30.53 g kg -1 , total nitrogen content 1.08g·kg -1 , total phosphorus content 0.68g·kg -1 , total potassium content 20.75g·kg -1 , alkaline nitrogen content 106.56 mg kg -1 , available phosphorus content 62.43 mg·kg -1 , fast-acting potassium content 186.76 mg·kg -1 . Weather data records such as Figure 2 As shown, the daily average temperature and daily precipitation during the 2024 experiment generally showed seasonal changes, with the highest temperature in July and August, with an average temperature exceeding 28°C, and precipitation mainly concentrated in April-June.

[0175] Three treatments were set up in the experiment: ① Control group (CK): no straw return was carried out; ② Conventional straw return group (SC): rice straw was crushed and evenly returned to the field at a rate of 4000 kg·hm -2 ; ③ Improved straw return group (SD): rice straw 4000 kg·hm -2 + 30kg·hm2 of rot-promoting agent -2 + urea 75kg·hm -2 Each treatment was repeated three times, with randomized block arrangement and a plot area of 120 m 2 The number of effective bacteria of the rot-promoting agent is ≥5×10 7 ·g -1 The main ingredients are complex strains such as Bacillus subtilis and Bacillus licheniformis.

[0176] After the autumn rice harvest, each treatment group was implemented according to the experimental plan. For both the SC and SD groups, rice straw was chopped into approximately 5 cm lengths and evenly spread across the field. For the SD group, the straw surface was evenly coated with a decomposition-accelerating agent and urea. After soaking the straw with water for 3-5 days, the straw was treated with a rotary tiller, maintaining a 7-10 cm water layer and retting for 35 days. Drainage ditches were dug 15-20 days before ploughing to allow the water to drain naturally. The field was then plowed and ridged for tobacco planting.

[0177] Tobacco varieties, Yunyan 87, were selected and transplanted on January 15, 2024, at a planting density of 1.2 m × 0.5 m. Rice varieties, Yongyou 1540, were selected and transplanted on July 6, 2024, at a planting density of 0.2 m × 0.2 m. Both tobacco and rice were managed according to local high-yield cultivation standards. Aside from the experimental treatments, all other fertilization rates and management measures were identical.

[0178] Soil greenhouse gas emissions were monitored using a static gas chamber coupled with gas chromatography. Gas samples were collected every seven days, with sampling increased to one or two days for a week during critical periods such as fertilization. Analysis of the monitoring data revealed soil greenhouse gas emission characteristics and the net ecosystem carbon budget (NECB) under different treatments, as shown in Table 11.

[0179] Table 11 Effects of different straw return methods on cumulative greenhouse gas emissions from tobacco-rice rotation soils

[0180]

[0181] Note: Different lowercase letters indicate significant differences among treatments (P<0.05).

[0182] As can be seen from Table 11, Figure 3 、 Figure 4 As shown in Figure 2, during the tobacco growing season, the SC and SD treatments significantly increased the cumulative CO2 emissions compared to the control treatment, with increases of 66.39% and 58.37%, respectively. However, the SD treatment reduced CO2 emissions compared to the SC treatment. Figure 5 、 Figure 6 As shown in the figure, for CH4 emissions, the SD treatment showed a negative value (-0.38 kg·hm -2 ), indicating that this treatment was a weak CH4 sink during the tobacco season, and the SC treatment was significantly higher than the CK treatment. Figure 7 、 Figure 8 As shown in the figure, N2O emissions during the tobacco season were SD>SC>CK, and the SD treatment increased by 34.88% compared with the CK treatment.

[0183] During the rice-growing season, the differences in cumulative CO₂ emissions among the three treatments were relatively small, with the SC and SD treatments increasing by 3.16% and 1.64%, respectively, compared to the CK treatment. CH₄ emissions were significantly higher in the SD treatment than in the CK treatment (by 12.81%), while those in the SC treatment were slightly lower. N₂O emissions during the rice season also showed a trend of SD > SC > CK.

[0184] like Figure 11 、 12 As shown in Table 12, in order to gain a deeper understanding of the effects of different treatments on the carbon balance of farmland, the net ecosystem carbon budget (NECB) was calculated.

[0185] Table 12 Effects of different straw return methods on the net ecosystem carbon budget of tobacco-rice rotation

[0186]

[0187] Note: C straw 、C fertilizer and C harvest =C refers to the carbon content of returned straw, applied fertilizer, and harvested crops, respectively. NPP is net primary productivity, and Rh is soil heterotrophic respiration. Different lowercase letters indicate significant differences among treatments (P < 0.05).

[0188] As shown in Table 11, compared with the CK treatment, the straw return treatments (SC and SD) significantly improved the net ecosystem carbon budget (NECB). Among them, the NECB value of the SD treatment reached 1490.07 kg C·hm -2 , which was 37.00% higher than that of SC treatment and 903.01% higher than that of CK treatment. The high NECB of SD treatment was mainly attributed to the direct input of straw carbon (C straw ) and significantly increased net primary productivity (NPP). Although the SD treatment also increased carbon output (including C harvest , Rh and CH4), but the increase in carbon input was greater, thus improving the carbon sequestration capacity of the farmland ecosystem as a whole.

[0189] like Figure 9 、 Figure 10 As shown, the Mantel test is a non-parametric statistical method for evaluating the correlation between two sets of distance matrices. The present invention uses the Mantel test to analyze the correlation between soil carbon emissions and soil factors, where Pearson's r refers to the degree of correlation between soil factors.

[0190] Mantel's r value: This value indicates the degree of correlation between soil carbon emissions and soil factors. Typically, it is a value between -1 and 1: r = 1 indicates a completely positive correlation; r = -1 indicates a completely negative correlation; and r = 0 indicates no linear correlation. Generally, the closer r is to 1 or -1, the stronger the correlation between the two matrices.

[0191] Mantel's p-value indicates the significance of the test results. Generally speaking, a p-value less than 0.05 indicates statistical significance, indicating a significant correlation between the two matrices. To further explore the factors influencing soil carbon emissions, this study analyzed changes in soil physical and chemical properties and enzyme activities under different treatments, as shown in Table 13.

[0192] Table 13 Effects of different straw return methods on soil physical and chemical properties and enzyme activities in tobacco-rice rotation

[0193]

[0194] Note: SOC is soil organic carbon, TN is total nitrogen, AN is alkaline-hydrolyzed nitrogen, AK is available potassium, S-β-GC is β-glucosidase activity, and S-PPO is polyphenol oxidase activity. Different lowercase letters indicate significant differences among treatments (P < 0.05).

[0195] As shown in Table 12, compared with the CK treatment, the SD treatment significantly increased soil pH, organic carbon content (SOC), total nitrogen content (TN), alkaline nitrogen content (AN), available potassium content (AK), and β-glucosidase (S-β-GC) and polyphenol oxidase (S-PPO) activities, but decreased the carbon-to-nitrogen ratio (C / N). These changes in soil physicochemical properties and enzyme activities directly affect soil carbon emissions. Correlation analysis results showed that soil AK content and S-PPO enzyme activity were key factors affecting soil CO2 emissions, while soil pH, SOC, TN, C / N, and AN content were key factors affecting N2O emissions.

[0196] The results of this study show that compared with traditional straw shredding and returning (SC) and no straw returning (CK), the modified straw returning (SD) method with the addition of a decomposition promoter and urea has significant advantages in carbon emission reduction and carbon sequestration efficiency. Although the SD treatment increased CH4 and N2O emissions in rice fields, after comprehensively considering the carbon balance, the SD treatment still showed the highest net ecosystem carbon budget (NECB), significantly improving the carbon sequestration function of the farmland ecosystem.

[0197] Traditional farmland straw management mainly adopts the methods of burning or directly returning it to the field. Although straw burning is simple and easy to implement, it will cause air pollution and direct carbon emissions; and although directly returning it to the field can increase soil organic matter, due to the high C / N ratio of straw, it decomposes slowly, which will cause temporary nutrient fixation in the short term and may even lead to high carbon emissions due to enhanced carbon mineralization. The key technical innovation of the present invention is that by adding decomposition promoters and urea to the straw, on the one hand, it accelerates the decomposition process of the straw, and on the other hand, it reduces the C / N ratio of the straw and alleviates the nitrogen competition during the microbial decomposition process, thereby achieving the multiple goals of reducing soil CO2 emissions, improving crop productivity, and enhancing the soil carbon sink function.

[0198] This invention not only solves the technical problems of low utilization rate and slow decomposition of traditional straw return to the field, but also significantly improves the carbon sequestration capacity of the farmland ecosystem by optimizing the carbon-nitrogen cycle process of the tobacco-rice rotation system, providing a feasible technical path for agricultural carbon emission reduction and soil carbon sequestration, and has important application value for promoting green and low-carbon development of agriculture.

[0199] It should be noted that the variables involved in the present invention are explained in detail as shown in Table 14 below.

[0200] Table 14 Variable explanation table

[0201]

[0202]

[0203] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation, characterized in that: include: A tobacco-rice rotation experimental field was selected and divided into several plots; a straw return experiment treatment was set up; Install a static gas chamber measurement system; collect gas samples at a set frequency; record temperature and humidity data within the chamber during sampling; use gas chromatography to determine carbon dioxide and methane concentrations in the gas samples, and calculate soil carbon emission flux and accumulation; apply soil microbial kinetic equations to analyze the straw decomposition process; Based on the carbon cycle neural network model, the dynamic changes of soil carbon pool under different straw return conditions are analyzed, and the comprehensive influence of environmental factors, crop growth parameters and microbial community structure on soil carbon emission flux is integrated through a multi-head cross-attention mechanism; the minimum spanning tree algorithm is used to construct the carbon flow network topology structure, identify the key nodes of carbon flow, and optimize the farmland carbon management strategy.

2. The method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation according to claim 1, characterized in that: A tobacco-rice rotation experimental field was selected and the experimental area was divided into several plots of 120 square meters each. Random block arrangement was adopted and three replicates were set for each treatment.

3. The method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation according to claim 2, characterized in that: A straw return experiment was set up, including a non-straw return control group, a simple straw return group, and a straw return plus decomposition promoter and urea treatment group. The straw return amount was 4,000 kg per hectare, the decomposition promoter amount was 30 kg per hectare, and the urea amount added was 75 kg per hectare.

4. The method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation according to claim 3, characterized in that: The static air box measurement system includes an acrylic box, a stainless steel base, an automatic temperature and humidity recorder, and a gas collection component. Static air boxes are set up on the ridges and below the ridges.

5. The method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation according to claim 4, characterized in that: Gas samples are collected at a frequency of once every 7 days. During the fertilization period, sampling is intensified to once every 1 to 2 days and continues for a week. The sampling time is from 8:00 to 11:00 in the morning. The temperature and humidity data in the box are recorded during sampling, and 30 ml of gas samples are collected at 0 minutes, 10 minutes, 20 minutes and 30 minutes after placement.

6. The method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation according to claim 5, characterized in that: Soil carbon emission fluxes were calculated using gas density, gas chamber height, gas concentration change rate, and temperature correction factor, and were expressed in milligrams per square meter per hour.

7. The method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation according to claim 6, characterized in that: The cumulative amount of soil carbon emissions was calculated by multiplying the average emission flux of adjacent sampling periods by the number of days between sampling, and the unit is kilograms per hectare.

8. The method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation according to claim 7, characterized in that: The soil microbial kinetic equation was used to describe the quantitative relationship between microbial activity and carbon emissions during straw organic carbon decomposition, taking into account factors such as initial carbon content, temperature response, moisture response, carbon-nitrogen ratio response, and microbial biomass carbon response.

9. The method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation according to claim 8, characterized in that: The specific structure of the carbon cycle neural network model is a hybrid architecture that combines a multi-layer bidirectional long short-term memory network with a graph convolutional network, which includes a temporal feature extraction layer, a spatial relationship encoding layer, a multi-head cross-attention layer, and a carbon cycle parameter prediction layer.

10. The method for determining the impact of straw return methods on soil carbon emissions in tobacco-rice rotation according to claim 9, characterized in that: The number of heads in the multi-head attention mechanism is determined by the number of soil carbon pools, and the attention weight is determined by the intensity of the impact of environmental factors on each carbon pool. The model input includes time series data of soil physical and chemical indicators, microbial community structure characteristics and meteorological data, and the output is the predicted value of the dynamic change of soil carbon pools and soil carbon emission flux.

Images