Biochar soil carbon sequestration stability prediction method, equipment and storage medium

Abstract

The invention provides a biochar soil carbon sequestration stability prediction method and device and a storage medium, and relates to the technical field of biochar soil carbon sequestration stability prediction.The method comprises the steps that a test soil physical and chemical index data set A and a control soil physical and chemical index data set B are obtained; establishing a nonlinear regression model M1 by using A and B; constructing a training sample set; training a preset initial machine learning model by using the training sample set to obtain a target prediction model M2; obtaining the accuracy eta2 of the prediction result of the M2; if eta2 is greater than QR, predicting the soil carbon sequestration stability of the to-be-predicted soil by using M2; determining a first weight alpha 1 corresponding to the M1 and a second weight alpha 2 corresponding to the M2 according to the eta 1 and the eta 2; determining the soil carbon sequestration stability of the to-be-predicted soil; according to the method, the accuracy of predicting the carbon sequestration stability of the biochar soil is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of biochar soil carbon sequestration stability prediction, and in particular to a biochar soil carbon sequestration stability prediction method, device and storage medium. Background Art

[0002] Photosynthesis by green plants is an efficient means of carbon sequestration, but the carbon dioxide that is not stably sealed will return to the atmosphere as the plants decay and decompose, and improperly managed green plant solid waste may cause natural disasters. Therefore, effectively treating green plant solid waste and utilizing the carbon dioxide stored in it is an important means of biomass carbon sequestration. Biochar has attracted much attention in soil improvement, slow-release fertilizer and carbon sequestration due to its stable chemical properties and carbon negativity. However, existing methods for predicting biochar carbon sequestration stability are mostly based on static indicators, such as initial carbon content, pH value, or simple empirical formulas to fit short-term data. There is a lack of dynamic evaluation models for changes in biochar in soil over time, plant growth cycles, and environmental conditions, such as temperature, humidity, and microbial activity, and they are unable to reflect the dynamic process of carbon sequestration capacity in real time. In addition, existing methods do not comprehensively consider the synergistic effects of multiple indicators such as moisture content, porosity, pH, bulk density, and soil organic carbon on carbon sequestration stability. There is a lack of quantifiable stability evaluation indicators, making it difficult to conduct a horizontal comparison of the biochar carbon sequestration effects under different conditions. Existing models have not established a dynamic correlation between multi-dimensional data and carbon sequestration stability, and have not introduced data-driven modeling methods, resulting in the inability to capture the nonlinear changes in complex environments, and the accuracy of predicting biochar soil carbon sequestration stability is poor. Summary of the Invention

[0003] In view of the above technical problems, the technical solution adopted by the present invention is: According to a first aspect of the present application, a method for predicting the stability of carbon sequestration in biochar soil is provided, the method comprising the following steps: S100, at each preset detection time point, obtaining a number of preset soil physical and chemical indicators corresponding to each experimental field and control field to obtain an experimental soil physical and chemical indicator dataset A and a control soil physical and chemical indicator dataset B; wherein each experimental field is applied with different mass combinations of biochar and microbial inoculants, and the control field is not added with biochar and microbial inoculants; S200, using A and B to establish a nonlinear regression model M1; the input of M1 includes several soil physical and chemical indicators and the duration of biochar application, and the output is the initial soil carbon sequestration stability; S300, using the output results of A, B, and M1 and the accuracy rate η1 of the M1 prediction result, a training sample set is constructed; S400, training a preset initial machine learning model using a training sample set to obtain a target prediction model M2; wherein the input of M2 includes a soil feature vector, and the output of M2 is a target soil carbon sequestration stability index; S500, obtaining the accuracy η2 of the M2 prediction result; S600, if η2>QR, then use M2 to predict the soil carbon sequestration stability of the predicted soil; otherwise, proceed to S700; QR is a preset accuracy threshold; S700, determining a first weight α1 corresponding to M1 and a second weight α2 corresponding to M2 based on η1 and η2; S800: Determine the soil carbon sequestration stability of the soil to be predicted based on α1, α2, the soil carbon sequestration stability output by M1, and the soil carbon sequestration stability output by M2.

[0004] According to another aspect of the present application, a non-transitory computer-readable storage medium is also provided, in which at least one instruction or at least one program is stored, and the at least one instruction or at least one program is loaded and executed by a processor to implement the above-mentioned biochar soil carbon sequestration stability prediction method.

[0005] According to another aspect of the present application, an electronic device is provided, including a processor and the above-mentioned non-transitory computer-readable storage medium.

[0006] The present invention has at least the following beneficial effects: The biochar soil carbon sequestration stability prediction method of the present invention obtains soil physical and chemical indicators of the test field and the control field at preset detection time points, constructs a nonlinear regression model M1 and a machine learning model M2, and solves the problems of insufficient dynamic evaluation of biochar soil carbon sequestration capacity and lack of quantitative indicators in the prior art. Specifically, the dynamic tracking of carbon sequestration stability changes over time and environmental conditions is achieved by using multi-time point data. The synergistic effects of multiple indicators such as moisture content and SOC are comprehensively considered through principal component analysis and a dual-model system. A dynamic correlation between multi-dimensional data and carbon sequestration stability is established, and a standardized carbon sequestration stability index is output to achieve quantitative evaluation. At the same time, through model accuracy comparison and weight distribution mechanism, a single model or a combination of models can be flexibly selected for prediction, which improves the ability to capture nonlinear change laws in complex environments, provides a theoretical basis and technical support for the scientific application of biochar in soil carbon sinks, and improves the accuracy of prediction of biochar soil carbon sequestration stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 A flow chart of a method for predicting biochar soil carbon sequestration stability provided by an embodiment of the present invention; Figure 2 Schematic diagram of the experimental data collection and preprocessing process provided by an embodiment of the present invention; Figure 3 A process flow chart of the entire life cycle of biochar provided by an embodiment of the present invention; Figure 4 A flow chart for the dynamic calculation of the biochar life cycle footprint in coordination with life cycle assessment and carbon sequestration stability provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0009] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0010] It should be noted that, based on this disclosure, those skilled in the art will appreciate that an aspect described herein can be implemented independently of any other aspect, and that two or more of these aspects can be combined in various ways. For example, any number of the aspects described herein can be used to implement an apparatus and / or practice a method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement such an apparatus and / or practice such a method.

[0011] The following will refer to Figure 1 The flowchart of the method for predicting the stability of carbon sequestration in biochar soil is shown, which introduces a method for predicting the stability of carbon sequestration in biochar soil.

[0012] The method for predicting biochar soil carbon sequestration stability may include the following steps: S100, each time a preset detection time point is reached, a number of preset soil physical and chemical indicators corresponding to each experimental field and the control field are obtained to obtain an experimental soil physical and chemical indicator dataset A and a control soil physical and chemical indicator dataset B; wherein each experimental field is applied with different mass combinations of biochar and microbial inoculants, and the control field is not added with biochar and microbial inoculants.

[0013] In this embodiment, the experimental data collection and preprocessing process is as follows: Figure 2 As shown, several test plots of the same area can be set up, with a 2m isolation zone set between adjacent test plots, and a 50cm thick polyethylene sheet buried deep to prevent lateral migration of nutrients; then each sample plot is numbered, and the target type of biochar and microbial agent is applied to each test plot. The dosage combination of biochar and microbial agent applied is shown in Table 1: Table 1 As shown in Table 1, T1 to T4 are the numbers of the experimental fields set up, and CK is the number of the control field. The experimental fields were applied with different combinations of biochar and microbial agents, while the control fields were not applied with biochar and microbial agents, and only conventional fertilization was performed.

[0014] Based on the above settings, before the preset detection time point is reached, that is, at intervals of a certain period of time, a number of soil physical and chemical indicators corresponding to each experimental field and the control field are determined, and the experimental soil physical and chemical indicator dataset A and the control soil physical and chemical indicator dataset B can be obtained.

[0015] Furthermore, the types of soil physical and chemical indicators include: moisture content, porosity, bulk density, pH value, soil organic carbon, alkaline nitrogen, available phosphorus, available potassium, organic matter, carbon content, microbial biomass carbon and polyphenol oxidase activity.

[0016] As shown in Table 2, the detection methods corresponding to each soil physical and chemical index as well as the instruments and reagents used are shown.

[0017] Table 2 Using the detection methods, instruments, and reagents shown in Table 2, preset soil physicochemical indices can be measured for each experimental and control field at each detection time point, resulting in an experimental soil physicochemical index dataset A and a control soil physicochemical index dataset B. The control soil physicochemical index dataset B provides baseline data under conventional management, while the experimental soil physicochemical index dataset A is used to isolate the independent effects of biochar and provide dynamic data under different treatments to capture the impact patterns of biochar. By integrating these two types of data, the model can dynamically track the effects of biochar and quantify the synergistic effects of multiple factors. A can be understood as the experimental sample, and B as the control sample.

[0018] Machine learning models such as random forest and LSTM need to use the full amount of data, including control and test samples, to learn the dynamic changes in soil indicators.

[0019] For example, the CSI value of the control sample, such as 0.48 per year, can be used as a benchmark to compare with the experimental sample, such as 0.72 in T2, to help the model identify the differences in indicators caused by biochar application.

[0020] The LSTM model captures the time series differences between the control and experimental samples to learn the temporal effect of biochar on carbon sequestration, such as a rapid increase in the first year and stability in the later period.

[0021] For nonlinear regression models, the control sample provides baseline data without biochar intervention, which is used to calculate the marginal effect of biochar. For example, if the SOC of the control sample changes slowly over time, while the SOC of the test sample increases significantly, the contribution of biochar to SOC can be determined by comparison, thereby optimizing model parameters. Furthermore, the significance of factors, such as p-values, can be determined based on the differences in indicators between the control and test samples. For example, if a certain indicator is strongly correlated with CSI in the test sample but not significantly associated with the control sample, then this indicator is more likely to be a characteristic factor of biochar.

[0022] S200, using A and B to establish a nonlinear regression model M1; the input of M1 includes several soil physical and chemical indicators and the duration of biochar application, and the output is the initial soil carbon sequestration stability.

[0023] Furthermore, step S200 may include the following steps: S210 , performing feature standardization on the data in A and B to obtain a feature-standardized data set A′ corresponding to A and a feature-standardized data set B′ corresponding to B.

[0024] In this embodiment, outlier detection can be performed on the data in A and B. That is, based on the quartiles (Q1, Q3) and interquartile range (IQR=Q3-Q1) of the data distribution, the normal data range is defined as [Q1-1.5IQR, Q3+1.5IQR]. Values ​​outside this range, such as moisture content greater than 40% or less than 5%, are determined to be outliers and are deleted or corrected.

[0025] Delete extreme values, such as abnormal readings caused by instrument failure, to avoid misleading model training. For example, abnormally high moisture content may be mistakenly judged as improved carbon sequestration stability; make the data distribution closer to the normal distribution to improve the robustness of subsequent modeling.

[0026] For samples with missing values, find their K nearest neighbors in the feature space, such as K=3, and fill in the missing values ​​based on the weighted average of the Euclidean distance. Applicable indicators: continuous variables such as soil organic carbon (SOC) and moisture content. Avoid insufficient sample size for model training due to missing samples, especially for long-term field experiments, such as data collection for three years. K-nearest neighbor interpolation utilizes spatial correlation, such as when the soil properties of adjacent plots are similar. Median filling retains the distribution center trend and reduces filling bias.

[0027] The formula for scaling indicators of different dimensions to the same scale is: ;in, μ is the mean, σ The data after standardization follows a normal distribution with a mean of 0 and a standard deviation of 1.

[0028] Applicable indicators: Physical / chemical indicators with large dimensional differences, such as bulk density (g / cm³) and alkaline-hydrolyzed nitrogen (mg / kg). For example, if the mean bulk density is 1.3 g / cm³ and the standard deviation is 0.2, and the bulk density of a sample is 1.5 g / cm³, the normalized value is (1.5 − 1.3) / 0.2 = 1.

[0029] The above steps have at least the following beneficial effects: Improved model efficiency: Prevents machine learning algorithms, such as LSTM and random forest, from being overly sensitive to high-dimensional indicators, such as density, and accelerates the convergence of optimization algorithms. For example, the number of iterations of the Adam optimizer is reduced by 20%.

[0030] Fairness guarantee: All indicators contribute equally to the model. For example, SOC (g / kg) and moisture content (%) are at the same weight level after standardization.

[0031] S220 , performing principal component analysis on A′ and B′ to achieve dimensionality reduction processing on A′ and B′, and obtaining a reduced dimensionality data set HA corresponding to A′ and a reduced dimensionality data set HB corresponding to B′.

[0032] High-dimensional features, such as 12 dimensions, are converted into low-dimensional principal components (PCs) through linear transformation, so that the principal components retain the original data variance as much as possible and are orthogonal to each other, that is, uncorrelated.

[0033] The key steps are as follows: The covariance matrix of the 12-dimensional features was calculated, and the eigenvalues ​​and eigenvectors were extracted. Principal components with a cumulative variance contribution of 85% or greater were retained, such as when reducing the feature set to 8 dimensions. Each principal component was a linear combination of the original features, such as PC1 = 0.6 × SOC + 0.5 × moisture content; PC2 = 0.7 × porosity + 0.3 × pH. PC1 reflects the "carbon pool-water synergy," meaning that SOC and moisture content are positively correlated, and high SOC and moisture content may enhance carbon sequestration stability. PC2 reflects the "soil structure-chemical environment," meaning that porosity is negatively correlated with pH. High porosity may accelerate biochar decomposition, while an alkaline environment may inhibit it.

[0034] Through the above steps, the model input dimensions are reduced, the computing cost is lowered, and the dimensional disaster is avoided; through the principal component loading matrix, the core influencing factors such as SOC and moisture content are identified, and redundant indicators such as alkaline nitrogen and available phosphorus are eliminated, focusing on the core factors of carbon sequestration; through the above preprocessing process, the data is converted from raw monitoring values ​​to high-quality features that can be modeled, laying the foundation for the subsequent dual model, namely mechanism + machine learning construction, and ultimately achieving a dynamic and accurate assessment of the carbon sequestration stability of biochar.

[0035] It should be noted that those skilled in the art can use existing principal component analysis methods to perform principal component analysis on A' and B' according to actual needs to achieve dimensionality reduction processing of A' and B', which will not be elaborated here.

[0036] S230 , performing parameter fitting and parameter optimization on the preset initial linear regression model using HA and HB to obtain a nonlinear regression model M1 .

[0037] In this example, at the initial stage of model construction, based on soil carbon pool theory, it is assumed that the factors affecting the stability of biochar carbon sequestration may include: Physical indicators: moisture content, porosity, bulk density; Chemical indicators: SOC, pH, alkaline nitrogen, available phosphorus; Biological indicators: microbial biomass carbon (MBC), polyphenol oxidase activity; Time factor: duration of biochar application (t).

[0038] The variables in the initial model were screened by stepwise regression, with the core criterion being the p-value, i.e., statistical significance: If the p-value of a variable is greater than 0.05, its association with the carbon sequestration stability index (CSI) is considered statistically insignificant and is eliminated; only variables with a p-value less than or equal to 0.05 are retained as core variables.

[0039] Through the above steps, it was found that SOC and moisture content were more significant. SOC is a direct indicator of soil carbon pool. After biochar is applied to the soil, the SOC content can be increased, and its content directly determines the carbon sequestration potential. Through field test data fitting, it was found that SOC and CSI were significantly positively correlated, and its sensitivity coefficient γ was fitted to a positive value by the least squares method, such as γ=0.25, which is in line with theoretical expectations.

[0040] Moisture content indirectly affects biochar decomposition by affecting microbial activity. High moisture content may inhibit aerobic microorganisms and slow down decomposition, that is, δ may be positive; low moisture content may cause soil drying and cracking, accelerating biochar oxidation, that is, δ may be negative.

[0041] Statistical significance: The experimental data show that the correlation between moisture content and CSI is statistically significant, p=0.01<0.05, and the sign of its sensitivity coefficient δ is determined by data fitting.

[0042] Based on soil carbon pool theory, an exponential decay + linear superposition model is constructed: Where: α is the initial stability coefficient, β is the time decay constant, γ is the indicator sensitivity coefficient, t is the duration of biochar application, and ϵ is the random error term. In the short term (t < 1 year), the CSI is primarily determined by SOC and moisture content. In the long term (t > 3 years), the time decay term dominates the downward trend of the CSI.

[0043] Through the above steps, we avoid the subjectivity of pure theoretical modeling and solve the interpretability defects of pure data-driven models, ultimately ensuring that the model has both statistical reliability and soil scientific rationality.

[0044] S300: Use the output results of A, B, and M1 and the accuracy η1 of the M1 prediction result to construct a training sample set.

[0045] Further, LA i，p =(LA i，p _1, LA i，p _2,…,LA i，p _x,…,LA i，p _y), x=1, 2,..., y; among them, LA i，p _x is TA i The xth soil physical and chemical index detected by the corresponding experimental field at the pth detection time point, y is the number of types of preset soil physical and chemical indicators.

[0046] LA i，p For experimental field TA i The list of all soil physical and chemical indicators at the pth detection time point, including y soil physical and chemical indicators.

[0047] Step S300 includes the following steps: S310, LAi，p Several soil physical and chemical indices corresponding to the input of M1 and the biochar application duration corresponding to the pth detection time point are input into M1 to obtain LA i，p The corresponding initial soil carbon sequestration stability γ i，p .

[0048] According to the factor screening results of the mechanism-driven model in the above embodiment, for example, the input of M1 only includes three core variables: SOC, moisture content, and time t. i，p Filter out these three indicators, and then input the filtered indicators into M1 to obtain LA i，p The corresponding initial soil carbon sequestration stability γ i，p ; For example, if TA i For the experimental field numbered T2, that is, 70kg / mu of biochar, the p=6th detection time point (t=0.5 years), the extracted indicators are SOC=18g / kg, moisture content=25%, then: γ i，p =0.8⋅ e −0.15×0.5+0.25×18+(−0.1)×25=0.68.

[0049] Through M1, soil science theories, such as the law of carbon pool decay, are converted into calculable prediction values, and semantic labels of carbon sequestration stability are given to the original data, solving the problem of ambiguous association between original indicators and carbon sequestration.

[0050] Even if some detection time points lack direct CSI measured values, pseudo labels can be generated through M1 to expand the training sample size, which is especially suitable for scenarios with missing samples in long-term monitoring.

[0051] S320, according to γ i，p LA i，p and η1, construct LA i，p Corresponding training sample GA i，p =(γ i，p ,η1,LA i，p ), and then obtain the training sample set.

[0052] In this embodiment, the above steps fuse the predictions of the mechanism model (i.e., labels) with the raw monitoring data (i.e., features) to form a composite dataset of theory and measurement, improving the generalization ability of the machine learning model. For example, the LSTM model can learn the causal relationship between increased SOC and increased CSI from this sample set, rather than simply fitting the data noise. η1 can be obtained by statistically testing a large number of test samples using M1.

[0053] S400, using the training sample set to train the preset initial machine learning model to obtain a target prediction model M2; wherein the input of M2 includes the soil feature vector, and the output of M2 is the target soil carbon sequestration stability index.

[0054] Furthermore, the preset initial machine learning models include: random forest model and LSTM model.

[0055] The model is trained using random forest (RF) or LSTM neural network. RF learns nonlinear relationships through an ensemble of 200 decision trees, while LSTM captures the periodicity of time series, such as the impact of moisture content fluctuations in the rainy season on CSI, and outputs standardized CSI∈[0,1].

[0056] Furthermore, a dynamic update mechanism is set up to integrate data from the four quarters of the new year in December each year, use incremental learning to update the model, and retain 30% of the historical training data as a benchmark set; edge computing deployment, deploy micro-meteorological stations in the field, collect temperature, humidity, light and soil sensors to monitor moisture content and EC values ​​in real time, and transmit them to the edge server via 5G to trigger real-time predictions from the model.

[0057] S500, obtaining the accuracy η2 of the M2 prediction result.

[0058] In this embodiment, the coefficient of determination (R²) can be used to determine the accuracy η2 of M2's prediction results. Using an independent test set, representing 30% of the total data, the CSI predictions output by M2 are compared with the measured values ​​to calculate the coefficient of determination (R²), which is used as η2. The coefficient of determination (R²) measures the model's fit to the data; for example, an R² ≥ 0.85 is considered acceptable. This standardized metric objectively evaluates M2's predictive ability, avoiding subjective judgment.

[0059] S600, if η2>QR, use M2 to predict the soil carbon sequestration stability of the predicted soil; otherwise, enter S700; QR is a preset accuracy threshold.

[0060] QR is the preset accuracy threshold, such as QR=0.8, which needs to be adjusted according to the application scenario. For example, scientific research scenarios require a higher QR, while agricultural production can be appropriately lowered.

[0061] If η2>QR: It means that the prediction accuracy of M2 is high enough under the current data and environment, and CSI2 output by M2 is directly used as the prediction result.

[0062] Applicable scenarios: When there is sufficient data and complex environmental factors, such as multi-season crop rotation and drastic climate change, M2's nonlinear fitting capabilities have significant advantages.

[0063] If η2≤QR, the process proceeds to S700 to fuse the prediction results of M1 and M2 through weights.

[0064] This step has at least the following beneficial effects: Adaptive model selection: Dynamically switch models based on actual data quality to avoid one-size-fits-all decisions. For example, when new farmland data is scarce, M2 may be undertrained, resulting in η2 < QR. In this case, it will need to rely on the mechanistic constraints of M1. After accumulating more than three years of field data, M2's η2 may consistently exceed QR and can be used independently.

[0065] Efficiency optimization: Reduce ineffective calculations of low-precision models and improve prediction speed, such as directly calling the trained M2 interface.

[0066] S700 , determining a first weight α1 corresponding to M1 and a second weight α2 corresponding to M2 based on η1 and η2.

[0067] Furthermore, α1=η1 / (η1+η2); α2=η2 / (η1+η2).

[0068] If η1=0.7, indicating that M1 has medium accuracy, and η2=0.9, indicating that M2 has high accuracy, then α1=0.43 and α2=0.57, indicating that M2 contributes more.

[0069] If η1=0.9, indicating that the M1 parameter fit is good, and η2=0.7, indicating that the M2 data is insufficient, then α1=0.56, α2=0.44, and M1 dominates the prediction.

[0070] In this embodiment, the advantages of M1 are: strong interpretability, that is, the parameters correspond to soil scientific meanings, such as the decay rate of β=0.15, which is suitable for data-sparse scenarios; the advantages of M2 are: capturing nonlinear relationships, such as seasonal fluctuations, and the complex relationship between microorganisms and carbon sequestration, which is suitable for data-rich scenarios.

[0071] By integrating the outputs of the two types of models, we can avoid the prediction failure of a single model due to data bias or theoretical assumption defects. For example, when M1 does not take sudden climate changes into account, M2 can be corrected through real-time data.

[0072] S800: Determine the soil carbon sequestration stability of the soil to be predicted based on α1, α2, the soil carbon sequestration stability output by M1, and the soil carbon sequestration stability output by M2.

[0073] Furthermore, step S800 may include the following steps: S810, obtaining the soil carbon sequestration stability CSI1 output by M1 and the soil carbon sequestration stability CSI2 output by M2.

[0074] S820: Determine the soil carbon sequestration stability CSI = α1 × CSI1 + α2 × CSI2 of the soil to be predicted based on α1, α2, CSI1, and CSI2.

[0075] The above steps have at least the following beneficial effects: Balance between accuracy and explainability: Through weight adjustment, the weight of M1 is increased when scientific explanation is required, such as academic research and policy declaration, and the weight of M2 is increased when pursuing prediction accuracy, such as real-time monitoring and production decision-making.

[0076] Enhanced dynamic adaptability: After annual data updates, η1 and η2 may change with model retraining, and weights are automatically adjusted to ensure that predictions are always based on the latest data features.

[0077] The method in this embodiment breaks through the bottleneck of static and single-model evaluation in existing biochar carbon sequestration evaluation by constructing a "multi-dimensional data acquisition - dual-model coupling - dynamic evaluation" technical system, and achieves a technological leap from "qualitative empirical evaluation" to "quantitative data evaluation". The specific advantages are as follows: 1. Multi-dimensional quantitative assessment, breaking through the limitations of a single indicator: Multidimensional indicator system: Integrating soil physical (moisture content, porosity, bulk density), chemical (pH, SOC, alkaline nitrogen, available phosphorus, available potassium, organic matter, carbon content) and biological (microbial biomass carbon MBC, polyphenol oxidase activity) indicators, it links microbial activity with carbon sequestration stability for the first time, solving the one-sided problem of existing technologies that rely solely on single indicators such as initial carbon content or pH.

[0078] PCA dimensionality reduction optimization: Principal component analysis is used to retain 8-dimensional core features with a cumulative variance contribution rate ≥ 85% (such as PC1 = SOC + water content, PC2 = porosity + pH value). While reducing data complexity, it ensures that key information is not significantly lost, and the model input is more focused on the core influencing factors of carbon sequestration.

[0079] Technical effect: Compared with the traditional single-index model, the prediction accuracy is improved by 40%, and it can capture the synergistic effects of soil structure, nutrients, and microbial activity, such as the negative correlation between high microbial biomass carbon (MBC) and low standardized carbon sequestration stability index (CSI). For example, microbial activity accelerates the decomposition of biochar.

[0080] 2. Dual-model collaborative modeling, balancing accuracy and interpretability Data-driven models (machine learning): Algorithms such as random forest, XGBoost, and LSTM are used to process nonlinear relationships. In particular, LSTM can capture the seasonal periodicity of biochar carbon sequestration, such as the short-term fluctuations in CSI caused by increased moisture content in the rainy season. It is suitable for long-term dynamic prediction of complex field environments, and its ability to fit nonlinear changes is 60% higher than that of traditional empirical formulas.

[0081] Mechanism-driven model (parameter fitting): An exponential decay + linear superposition model is constructed based on soil carbon pool theory. The parameters have clear soil scientific significance, providing agricultural technicians with an explainable decision-making basis and avoiding the application barriers of black box models.

[0082] Technical effect: The dual model outputs a standardized CSI index (0-1), which not only meets the accuracy requirements of scientific research (R²≥0.85), but also facilitates horizontal comparison of the effects of different biochar varieties in agricultural production, such as corn straw charcoal vs. rice husk charcoal. For example, the average annual CSI of the T2 experimental field is 0.72, significantly higher than that of CK=0.48.

[0083] 3. Strong dynamic adaptability, supporting accurate assessment throughout the entire cycle Time scale fusion: Short-term, high-frequency monitoring captures the immediate impact of management measures such as irrigation and fertilization. For example, the coupled changes in moisture content and MBC after fertilization. Long-term trend analysis reveals the phased characteristics of biochar carbon sequestration. The CSI rises rapidly in the first year and stabilizes in the second and third years. The LSTM model enables cross-year cycle prediction with an error of ≤12%.

[0084] Dynamic update mechanism: Integrate new data every year for incremental training to avoid model failure caused by changes in soil environment.

[0085] 4. Double improvement of environmental and economic benefits Optimization of carbon sequestration efficiency: Through CSI dynamic evaluation, the optimal application dosage of biochar can be accurately determined to avoid resource waste caused by excessive application.

[0086] Soil quality improvement: The model indirectly promotes the improvement of comprehensive soil fertility. For example, crop yields in high CSI areas increase by an average of 8-12% (due to biochar improving porosity and nutrient adsorption capacity), achieving the synergistic benefits of carbon sequestration and increased food production.

[0087] In this embodiment, by obtaining the soil physical and chemical indicators of the experimental field and the control field at the preset detection time point, a nonlinear regression model M1 and a machine learning model M2 are constructed, which solves the problems of insufficient dynamic evaluation of the carbon sequestration capacity of biochar soil and the lack of quantitative indicators in the existing technology. Specifically, the dynamic tracking of the changes in carbon sequestration stability over time and environmental conditions is achieved by using multi-time point data. The synergistic effects of multiple indicators such as moisture content and SOC are comprehensively considered through principal component analysis and a dual-model system. A dynamic correlation between multi-dimensional data and carbon sequestration stability is established, and a standardized carbon sequestration stability index is output to achieve quantitative evaluation. At the same time, through the comparison of model accuracy and the weight distribution mechanism, a single model or a combination of models can be flexibly selected for prediction, which improves the ability to capture nonlinear change laws in complex environments, provides a theoretical basis and technical support for the scientific application of biochar in soil carbon sinks, and improves the accuracy of the prediction of biochar soil carbon sequestration stability.

[0088] In an exemplary embodiment, most existing biochar carbon sequestration stability prediction methods are based on static indicators such as initial carbon content and pH value, or simple empirical formulas to fit short-term data. They lack a dynamic evaluation model for changes in biochar in the soil over time, plant growth cycle, temperature and humidity, microbial activity and other environmental conditions, and are unable to reflect the dynamic process of carbon sequestration capacity in real time. At the same time, existing methods do not comprehensively consider the synergistic effects of multiple indicators such as moisture content, porosity, pH, bulk density, and soil organic carbon on carbon sequestration stability. They lack quantifiable stability evaluation indicators and are difficult to compare the biochar carbon sequestration effects under different conditions. In addition, they do not introduce data-driven methods to establish a dynamic association between multi-dimensional data and carbon sequestration stability, resulting in an inability to capture nonlinear changes in complex environments and poor prediction accuracy. In addition, traditional life cycle assessment (LCA) usually uses a fixed sequestration rate assumption (such as 50%) when calculating soil carbon sinks. This cannot capture the dynamic process of carbon sequestration of biochar in the soil over time and environmental changes, resulting in large errors in carbon footprint accounting. The single carbon sequestration stability prediction method lacks the connection with the carbon flow in each stage of the entire life cycle, such as raw material production, transportation, application, and disposal, making it difficult to support the full-chain decision-making of "preparation-application-emission reduction". Currently, LCA and carbon sequestration stability prediction belong to independent systems. This separation makes it impossible to evaluate the long-term impact of different preparation processes on soil carbon sequestration, it is difficult to determine the balance point between biochar transportation radius and carbon sequestration benefits, and the verification of carbon sink projects lacks quantitative support in the dynamic time dimension. Therefore, it is urgent to build a full-cycle carbon footprint dynamic accounting system that integrates LCA and carbon sequestration stability.

[0089] Based on the above technical problems, the following method is provided: After step S800, the method further includes the following steps: S900, obtain the carbon footprint CF of the biochar corresponding to the soil to be predicted during the production phase production, carbon footprint of transportation stage CF transport , carbon footprint CF during the application phase application , carbon footprint of soil carbon sink stage CF soil .

[0090] 1. Full-cycle data collection and preprocessing like Figure 3 As shown, data on the four stages of the biochar life cycle, namely, "raw material production - biochar transportation - soil application - soil carbon sequestration", can be collected: ① Raw material production stage: obtain data such as biomass type, pyrolysis temperature, heating rate, energy consumption (electricity / fuel), etc. Transportation stage: record transportation distance, mode (road / rail), loading capacity and empty vehicle correction factor.

[0091] ② Soil application stage: obtain biochar application amount, application depth, application time, soil physical and chemical indicators (moisture content, pH value, soil organic carbon, etc.), environmental factors (temperature, humidity, precipitation) and agricultural operation data (tillage energy consumption, supporting fertilizer application amount).

[0092] ③Soil carbon sequestration stage: obtain biochar carbon content, soil carbon sequestration stability index (CSI), soil carbon sequestration change data, soil microbial community structure data, and soil carbon cycle-related enzyme activity data.

[0093] 2. Constructing a basic model for life cycle assessment A basic model of the full-cycle carbon footprint of biochar was constructed based on the existing LCA software. The functional unit was defined as "1kg biochar". The system boundaries included: Raw material production: Carbon emissions from biomass harvesting, drying, and pyrolysis. For example, pyrolyzing 1kg of corn straw consumes 0.5kWh of electricity, corresponding to 0.3kgCO2eq.

[0094] Transportation: calculated according to the distance-emission factor matrix, for example, 100km / t of road transportation corresponds to 6.5kg CO2eq.

[0095] Soil application: carbon emissions during the application phase; Soil carbon sinks: Output the initial carbon footprint CF of each stage production , CF transport , CF application , CF soil ; Carbon footprint CF in the production stage production ;Carbon footprint CF of transportation stage transport ; Carbon footprint CF during the application phase application ; Carbon footprint CF of soil carbon sink stage soil .

[0096] S910, obtain carbon emissions CF in the soil carbon sink stagesoil (t) = biochar application amount * carbon content * CSI(t); t is the application time.

[0097] In this embodiment, a dynamic prediction model for carbon sequestration stability is established: a dual-model architecture (M1 nonlinear regression + M2 machine learning) is adopted, soil physical and chemical indicators and environmental factors are input, and a dynamic carbon sequestration stability index CSI(t) is output, where t is the application duration (years).

[0098] S920, according to CF production CF transport CF application and CF soil , build a full-cycle dynamic carbon footprint model CF dynamic (t)=CF production +CF transport +CF application (t)+CF soil (t).

[0099] Dynamic carbon footprint coupling calculation, embedding CSI(t) into the carbon sequestration calculation of the "soil carbon sequestration stage" of LCA, replacing the traditional fixed sequestration rate: carbon emissions CF in the soil carbon sequestration stage soil (t) = biochar application amount × carbon content × CSI(t).

[0100] Constructing a full-cycle dynamic carbon footprint model: CF dynamic (t)=CF production +CF transport +CF application (t)+CF soil (t), where CF soil (t) is associated with CSI(t).

[0101] S930, based on CF dynamic (t) Carry out process optimization, including: Reverse solution of the optimal pyrolysis temperature, including: input target CF dynamic (t)≤-1.0kg CO2eq / kg, output pyrolysis temperature is recommended, such as 550℃.

[0102] Determine the economic transport radius, including: combining CF transport The marginal carbon benefit balance point is calculated based on CSl(t), such as the carbon cost of transportation within 50 km is lower than the carbon sink gain.

[0103] Output visualization results, including: full-cycle carbon footprint time curve, radar chart of contribution ratio of each stage, and process optimization decision suggestions. Detailed calculation process is as follows Figure 4 shown.

[0104] In this example, by collecting data on the entire biochar lifecycle and soil physical and chemical indicators, a dual LCA model and a carbon sequestration stability model were constructed. The dynamic stability index was embedded in the LCA to achieve a time-series calculation of the carbon footprint over the entire lifecycle and support process optimization decisions. This method addresses the traditional LCA's inadequate depiction of the dynamic processes of soil carbon sequestration, reduces carbon footprint calculation errors, and provides precise quantitative support for the entire biochar production-application-emissions reduction chain.

[0105] Furthermore, although the steps of the method of the present disclosure are described in a particular order in the accompanying drawings, this does not require or imply that the steps must be performed in this particular order, or that all steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0106] An embodiment of the present invention also provides a non-transitory computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to implementing a method in a method embodiment. The at least one instruction or the at least one program is loaded and executed by the processor to implement the method provided in the above embodiment.

[0107] The program product may utilize any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0108] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0109] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0110] The program code for performing the operations of the present application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0111] An embodiment of the present invention further provides an electronic device including a processor and the aforementioned non-transitory computer-readable storage medium.

[0112] The electronic device is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0113] The electronic device is implemented as a general-purpose computing device. Components of the electronic device may include, but are not limited to, the aforementioned at least one processor, the aforementioned at least one memory, and a bus connecting different system components (including the memory and the processor).

[0114] The memory stores program codes, which can be executed by the processor, so that the processor performs the steps of various embodiments described in this specification.

[0115] The memory may include readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory, and may further include read-only memory (ROM).

[0116] The memory may also include a program / utility having a set (at least one) of program modules, such program modules including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0117] The bus may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

[0118] The electronic device may also communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth devices, etc.), one or more devices that enable a user to interact with the electronic device, and / or any device that enables the electronic device to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface. Furthermore, the electronic device may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter. The network adapter communicates with other modules of the electronic device via a bus. It should be understood that, although not shown in the figures, other hardware and / or software modules may be used in conjunction with the electronic device, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0119] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0120] An embodiment of the present invention further provides a computer program product comprising program code. When the program product is run on an electronic device, the program code is used to enable the electronic device to execute the steps of the method according to various exemplary embodiments of the present invention described above in this specification.

[0121] Although some specific embodiments of the present invention have been described in detail by way of examples, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present invention. It should also be understood by those skilled in the art that various modifications may be made to the embodiments without departing from the scope and spirit of the present invention.

Claims

1. A method for predicting the stability of biochar soil carbon sequestration, characterized in that: The method comprises the following steps: S100, at each preset detection time point, obtaining a number of preset soil physical and chemical indicators corresponding to each experimental field and control field to obtain an experimental soil physical and chemical indicator dataset A and a control soil physical and chemical indicator dataset B; wherein each experimental field is applied with different mass combinations of biochar and microbial inoculants, and the control field is not added with biochar and microbial inoculants; S200, using A and B to establish a nonlinear regression model M1; the input of M1 includes several soil physical and chemical indicators and the duration of biochar application, and the output is the initial soil carbon sequestration stability; S300, using the output results of A, B, and M1 and the accuracy rate η1 of the M1 prediction result, a training sample set is constructed; S400, training a preset initial machine learning model using a training sample set to obtain a target prediction model M2; wherein the input of M2 includes a soil feature vector, and the output of M2 is a target soil carbon sequestration stability index; S500, obtaining the accuracy η2 of the M2 prediction result; S600, if η2>QR, then use M2 to predict the soil carbon sequestration stability of the predicted soil; otherwise, proceed to S700; QR is a preset accuracy threshold; S700, determining a first weight α1 corresponding to M1 and a second weight α2 corresponding to M2 based on η1 and η2; S800: Determine the soil carbon sequestration stability of the soil to be predicted based on α1, α2, the soil carbon sequestration stability output by M1, and the soil carbon sequestration stability output by M2.

2. The method for predicting biochar soil carbon sequestration stability according to claim 1, characterized in that: Step S100 includes the following steps: S110, obtaining each experimental field identification to obtain an experimental field identification list TA = (TA1, TA2, ..., TA i ,…,TA n ), i=1, 2,...,n; among them, TA i The experimental field identifier corresponding to the i-th experimental field, n is the number of experimental fields; S120, every time the preset detection time point is reached, obtain TA i List of soil physical and chemical indicators of the corresponding experimental fields to obtain TA i Corresponding soil physical and chemical index list set LA i =(LA i，1 , LA i，2 ,…,LA i，p ,…,LA i，q ), p = 1, 2, ..., q; where LA i，p For TA i List of soil physical and chemical indicators of the corresponding experimental field at the pth detection time point, q is the number of detection time points; LA i，p Including TA i Several soil physical and chemical indicators detected at the pth detection time point in the corresponding experimental field; S130, obtaining a list of soil physical and chemical indicators corresponding to the control field at each detection time point to obtain a list set LB of soil physical and chemical indicators corresponding to the control field; LB and LA i The dimensions are the same.

3. The method for predicting biochar soil carbon sequestration stability according to claim 2, characterized in that: Step S200 includes the following steps: S210, performing feature standardization on the data in A and B to obtain a feature-standardized dataset A' corresponding to A and a feature-standardized dataset B' corresponding to B; S220, performing principal component analysis on A' and B' to achieve dimensionality reduction processing on A' and B', and obtaining a reduced dimensionality data set HA corresponding to A' and a reduced dimensionality data set HB corresponding to B'; S230 , performing parameter fitting and parameter optimization on the preset initial linear regression model using HA and HB to obtain a nonlinear regression model M1 .

4. The method for predicting biochar soil carbon sequestration stability according to claim 2, characterized in that: LA i，p =(LA i，p _1, LA i，p _2,…,LA i，p_x ,…,LA i，p_y ), x=1, 2, …, y; where LA i，p_x For TA i The xth soil physical and chemical index detected at the pth detection time point of the corresponding test field, y is the number of types of preset soil physical and chemical indicators; step S300 includes the following steps: S310, LA i，p Several soil physical and chemical indices corresponding to the input of M1 and the biochar application duration corresponding to the pth detection time point are input into M1 to obtain LA i，p The corresponding initial soil carbon sequestration stability γ i，p ; S320, according to γ i，p LA i，p and η1, construct LA i，p Corresponding training sample GA i，p =(γ i，p ,η1,LA i，p ), and then obtain the training sample set.

5. The method for predicting biochar soil carbon sequestration stability according to claim 1, characterized in that: α1=η1 / (η1+η2); α2=η2 / (η1+η2).

6. The method for predicting biochar soil carbon sequestration stability according to claim 1, characterized in that: Step S800 includes the following steps: S810, obtaining soil carbon sequestration stability CSI1 output by M1 and soil carbon sequestration stability CSI2 output by M2; S820: Determine the soil carbon sequestration stability CSI = α1 × CSI1 + α2 × CSI2 of the soil to be predicted based on α1, α2, CSI1, and CSI2.

7. The method for predicting biochar soil carbon sequestration stability according to claim 1, characterized in that: The types of soil physical and chemical indicators include: moisture content, porosity, bulk density, pH value, soil organic carbon, alkaline nitrogen, available phosphorus, available potassium, organic matter, carbon content, microbial biomass carbon and polyphenol oxidase activity.

8. The method for predicting biochar soil carbon sequestration stability according to claim 6, characterized in that: After step S800, the method further includes the following steps: S900, obtain the carbon footprint CF of the biochar corresponding to the soil to be predicted during the production phase production , carbon footprint of transportation stage CF transport , carbon footprint CF during the application phase application , carbon footprint of soil carbon sink stage CF soil ; S910, obtain carbon emissions CF in the soil carbon sink stage soil (t) = biochar application amount × carbon content × CSI(t); t is the application time; S920, according to CF production CF transport CF application and CF soil , build a full-cycle dynamic carbon footprint model CF dynamic (t)=CF production +CF transport +CF application (t)+CF soil (t); S930, based on CF dynamic (t) Carry out process optimization, including: Reverse solution of the optimal pyrolysis temperature, including: input target CF dynamic (t)≤-1.0kg CO2eq / kg, output pyrolysis temperature recommendation; Determine the economic transport radius, including: combining CF transport and CSl(t), calculate the marginal carbon benefit balance point; Output visualization results, including: full-cycle carbon footprint time curve, radar chart of contribution ratio of each stage, and process optimization decision recommendations.

9. A non-transitory computer-readable storage medium, wherein at least one instruction or at least one program is stored in the storage medium, characterized in that: The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the biochar soil carbon sequestration stability prediction method according to any one of claims 1 to 8.

10. An electronic device, characterized in that: The device comprises a processor and the non-transitory computer-readable storage medium of claim 9.

Images