A method for assessing the carbon footprint of wheat-corn rotation with straw return to the field

Abstract

The application discloses a kind of wheat corn rotation straw field returning carbon footprint evaluation method, it is related to carbon emission evaluation technical field, including, according to dynamic rotation crop time axis division growth stage, fusion regional crop carbon flux time series data and soil microbial activity index, using interpretable machine learning method, identify carbon sink stage and carbon source stage of wheat and corn in different growth stages, dynamically generate phase carbon state sequence;Based on phase carbon state sequence, fusion soil basic attribute data and real-time regional climate data, using ensemble learning method, fit the nonlinear response relationship of straw mineralization rate, dynamically generate carbon conversion factor sequence;Carbon conversion factor sequence and phase carbon state sequence are coupled in time and space, calculate CO2 and N2O emission and soil organic carbon change amount according to time period and spatial range, generate carbon dynamic curve by time series integration.The application improves the timeliness and accuracy of carbon dynamic simulation.

Description

Technical Field

[0001] This invention relates to the field of carbon emission assessment technology, and in particular to a method for assessing the carbon footprint of wheat-corn rotation with straw return to the field. Background Technology

[0002] With increasing global awareness of climate change and environmental protection, agricultural carbon footprint assessment technology has gradually become a research hotspot, especially for the dynamic monitoring and analysis of carbon emissions and carbon sequestration processes in crop rotation systems. Current technologies mainly rely on traditional carbon flux observations and empirical models to estimate greenhouse gas emissions from rotated farmland, combining fixed-point measurements of soil respiration and crop growth to establish static or semi-dynamic carbon balance models. Some studies employ multiple regression models based on environmental factors and soil parameters to attempt to reveal the relationship between carbon mineralization rates and climate and soil conditions during straw return to the field.

[0003] While existing technologies have attempted to estimate carbon emissions by incorporating multiple environmental factors, they generally fail to dynamically reflect the phased changes in carbon flux during wheat-maize rotation, particularly lacking the ability to identify carbon sink and carbon source phases. This hinders a deeper understanding and optimization of straw return effects. Furthermore, current technologies struggle to achieve spatiotemporal coupling analysis based on dynamic climate and soil heterogeneity, making it difficult to accurately simulate the nonlinear response relationships of carbon conversion processes. Consequently, this impacts the dynamic correction of carbon footprints and efficient decision support. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a carbon footprint assessment method for wheat-corn rotation with straw return to the field, solving the problem of difficulty in dynamically and accurately assessing changes in carbon emissions and carbon sequestration during crop rotation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for assessing the carbon footprint of wheat-corn rotation with straw return to the field, which includes collecting preliminary rotation time data, real-time regional climate data, and basic soil property data, and performing fusion analysis to generate a dynamic rotation crop timeline;

[0008] Based on the dynamic crop rotation timeline, the growth stages are divided, and the time series data of crop carbon flux in this region are integrated with soil microbial activity indicators. An interpretable machine learning method is used to identify the carbon sink stage and carbon source stage of wheat and corn at different growth stages, and dynamically generate a stage carbon state sequence.

[0009] Based on the phased carbon state sequence, soil basic property data and real-time regional climate data are integrated, and an ensemble learning method is used to fit the nonlinear response relationship of straw mineralization rate and dynamically generate carbon conversion factor sequence.

[0010] The carbon transformation factor sequence was spatiotemporally coupled with the stage carbon state sequence. The CO2 and N2O emissions and soil organic carbon changes were calculated according to time period and spatial range. The carbon dynamic curve was generated by time series integration.

[0011] Emission rates and soil carbon sequestration rates are extracted from the carbon dynamic curve, and periodic cumulative analysis is performed to calculate the total greenhouse gas emissions and total soil carbon sequestration, thereby obtaining the net carbon footprint value. A dynamic correction mechanism is constructed using meteorological anomaly factors and soil heterogeneity factors as inputs to perform spatiotemporal adaptive correction of the net carbon footprint value and generate net carbon footprint data.

[0012] Identify peak carbon emission periods and straw treatment phases from net carbon footprint data, and optimize straw return time, tillage depth, and crop rotation sequence based on carbon dynamic curves and net carbon footprint data to generate closed-loop agronomic management strategies.

[0013] As a preferred embodiment of the carbon footprint assessment method for wheat-corn rotation with straw return to the field as described in this invention, the specific steps for generating the dynamic rotation crop timeline are as follows:

[0014] The preliminary crop rotation time data, real-time regional climate data, and basic soil property data were standardized and weighted and fused to generate a fused feature set.

[0015] The temporal response characteristics of crop growth stages to environmental conditions are extracted from the fusion feature set, the rotation time nodes are dynamically adjusted, and the data are fused with the preliminary rotation time data to generate a dynamic rotation crop time axis.

[0016] As a preferred embodiment of the carbon footprint assessment method for wheat-corn rotation with straw return to the field as described in this invention, the specific steps for dynamically generating the staged carbon state sequence are as follows:

[0017] Based on the dynamic crop rotation timeline, the growth stage time periods are determined, and the measured carbon flux data of wheat and corn at each corresponding time period are compiled to form a carbon flux time series dataset.

[0018] The carbon flux time series dataset and the soil microbial activity index of the same period were standardized and feature fused to form a multidimensional environmental response feature set.

[0019] An integrated algorithm is used to dynamically identify carbon sink and carbon source characteristics at different growth stages from a multidimensional environmental response feature set, and generate a staged carbon state label sequence.

[0020] The phased carbon state label sequence is processed by time series smoothing and anomaly detection to generate a phased carbon state sequence.

[0021] As a preferred embodiment of the carbon footprint assessment method for wheat-maize rotation with straw return to the field as described in this invention, the method involves: standardizing and fusing the carbon flux time-series dataset with soil microbial activity indicators from the same period to form a multidimensional environmental response feature set. The specific steps are as follows:

[0022] The carbon flux time series dataset was time-aligned, interpolated, and smoothed with soil microbial activity indices from the same period to generate a joint data table.

[0023] The joint data table is standardized, and the time-series statistical features of carbon flux and microbial activity features are extracted, weighted and fused and restructured to form a multidimensional environmental response feature set.

[0024] As a preferred embodiment of the carbon footprint assessment method for wheat-corn rotation with straw return to the field as described in this invention, the specific steps for dynamically generating the carbon conversion factor sequence are as follows:

[0025] Based on the phased carbon state sequence, real-time regional climate data and basic soil property data are matched and feature fused according to the time dimension to generate a coupled regulation feature set;

[0026] Based on the coupled regulation feature set, time-series training samples are constructed, and ensemble learning is used to characterize the nonlinear response of straw mineralization rate, generating carbon conversion factor sequences for each time period.

[0027] As a preferred embodiment of the carbon footprint assessment method for wheat-corn rotation with straw return to the field as described in this invention, the specific steps for generating a carbon dynamic curve through time-series integration are as follows:

[0028] By combining the carbon transformation factor sequence with the stage carbon state sequence, and matching the corresponding farmland areas according to time and spatial coordinates, a spatiotemporal coupled dataset is formed.

[0029] Relevant variables were selected and analyzed from the spatiotemporally coupled dataset, and a multivariate response function was constructed.

[0030] Using a multivariate response function, and according to time step and spatial resolution, CO2 and N2O emission rates and soil organic carbon sequestration are calculated to generate a distributed emission and carbon sequestration rate matrix.

[0031] By integrating and accumulating the distributed emissions and carbon sequestration rate matrix, the total greenhouse gas emissions and soil carbon sink changes over a time period are calculated, generating a carbon dynamic curve.

[0032] As a preferred embodiment of the carbon footprint assessment method for wheat-corn rotation with straw return to the field as described in this invention, the specific steps for obtaining the net carbon footprint value are as follows:

[0033] The CO2 and N2O emission rates and soil carbon fixation rates for each time period were extracted from the carbon dynamic curve and summarized to form a periodic carbon flux dataset.

[0034] A time-series cumulative analysis was performed on the periodic carbon flux dataset to summarize the total greenhouse gas emissions and total soil carbon sequestration within the crop rotation cycle.

[0035] The net carbon footprint is calculated by subtracting the total greenhouse gas emissions from the total soil carbon sequestration.

[0036] As a preferred embodiment of the carbon footprint assessment method for wheat-corn rotation with straw return to the field as described in this invention, the specific steps for generating net carbon footprint data are as follows:

[0037] Based on the net carbon footprint value, a spatiotemporal perturbation factor system is constructed by combining perturbation factors extracted from real-time regional climate data and basic soil property data. This system is then feature-aligned and integrated with the net carbon footprint value to generate a high-dimensional input sample set.

[0038] Using a high-dimensional input sample set, the nonlinear response of the net carbon footprint value to spatiotemporal perturbation changes is fitted to generate a dynamic adjustment function;

[0039] The net carbon footprint value is adaptively calibrated using a dynamic adjustment function to generate net carbon footprint data.

[0040] As a preferred embodiment of the carbon footprint assessment method for wheat-corn rotation with straw return to the field as described in this invention, the method involves: using a high-dimensional input sample set to fit the nonlinear response of the net carbon footprint value to spatiotemporal disturbances, generating a dynamic adjustment function. The specific steps are as follows:

[0041] By using a nonlinear correlation measurement method, the complex dependency relationship between net carbon footprint and spatiotemporal disturbance factors is identified from a high-dimensional input sample set, and variable correlation characteristics are obtained.

[0042] Based on the characteristics of variable association, nonlinear regression is used to fit the dynamic response curve of net carbon footprint value relative to spatiotemporal disturbance factors, forming a preliminary adjustment function;

[0043] The initial adjustment function is optimized through cross-validation and residual analysis to generate a dynamic adjustment function.

[0044] As a preferred embodiment of the carbon footprint assessment method for wheat-corn rotation with straw return to the field as described in this invention, the specific steps for generating a closed-loop agronomic management strategy are as follows:

[0045] Analyze net carbon footprint data, extract carbon emission fluctuation information within crop rotation cycles, identify peak carbon emission periods and key straw treatment stages, and generate a table identifying peak carbon emission periods and a list of key straw treatment periods.

[0046] Based on the carbon emission peak period identification table and the list of key periods for straw treatment, and by aligning the records of management measures for the corresponding time periods in the carbon dynamic curve, we analyze the sensitivity response of management measures to CO2 and N2O emissions and extract a set of carbon emission response sensitivity factors.

[0047] Based on the set of carbon emission response sensitive factors, the timing of straw return to the field, tillage depth and crop rotation sequence are adjusted to generate a set of optimized agronomic management strategies.

[0048] For each set of optimized agronomic management strategies, carbon footprint simulation assessment and comparative analysis are conducted to select the combination with the lowest greenhouse gas emissions and the optimal soil carbon sequestration capacity, thereby generating a closed-loop agronomic management strategy.

[0049] The beneficial effects of this invention are as follows: by dynamically identifying the carbon sink and carbon source status of crops at different growth stages, a detailed temporal characterization of carbon emissions and carbon sequestration behavior is achieved, improving the timeliness and accuracy of carbon dynamic simulation; by effectively integrating crop physiological cycles and environmental response characteristics, it accurately reflects the comprehensive impact of multiple factors on the carbon cycle, providing a high spatiotemporal resolution input basis for dynamic carbon footprint assessment, and promoting the scientificity and rationality of carbon footprint assessment. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart of the carbon footprint assessment method for wheat-corn rotation straw return to the field in this invention.

[0052] Figure 2 This is a flowchart illustrating the generation of the dynamic crop rotation timeline in this invention.

[0053] Figure 3 This is a flowchart of the carbon dynamic curve generation process in this invention.

[0054] Figure 4 This is a flowchart illustrating the generation of the closed-loop agronomic management strategy in this invention. Detailed Implementation

[0055] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0056] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0057] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0058] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a method for assessing the carbon footprint of wheat-corn rotation with straw return to the field, comprising the following steps:

[0059] S1. Collect preliminary crop rotation time data, real-time regional climate data, and basic soil property data, and perform fusion analysis to generate a dynamic crop rotation timeline.

[0060] S1.1. Perform time-series data standardization and weighted fusion on the preliminary crop rotation time data, real-time regional climate data, and basic soil property data to generate a fused feature set.

[0061] Specifically, preliminary crop rotation time data, real-time regional climate data, and basic soil property data are uniformly resampled along the time dimension to ensure consistent time steps and alignment windows. The start and end points of the crop rotation in the preliminary crop rotation time data are normalized and encoded, for example, using the earliest crop sowing time as the starting point of the time axis and converting the remaining time points into standardized time values. For the sub-items in the real-time regional climate data, such as temperature, precipitation, radiation, and humidity, z-score standardization is performed to ensure a mean of 0 and a standard deviation of 1. For parameters in the basic soil property data, such as organic matter content, texture composition, pH value, and bulk density, min-max normalization is performed, scaling all variables to the [0,1] interval. After standardization, the standardized data are weighted and fused, with weighting coefficients set to reflect the relative contribution of each data type to the impact of crop rotation. A linear combination of the weighting coefficients is then performed to generate a fused feature vector, ultimately constructing a fused feature set.

[0062] It should also be explained that the specific steps for setting the weighting coefficient group are as follows: Based on the degree of influence of preliminary crop rotation time data, real-time regional climate data, and basic soil property data on crop growth stages, and in conjunction with historical experimental data, determine the relative importance. Set initial weighting coefficients, for example, 0.3 (preliminary crop rotation time data), 0.4 (real-time regional climate data), and 0.3 (basic soil property data).

[0063] S1.2 Extract the temporal response features of crop growth stages to environmental conditions from the fusion feature set, dynamically adjust the rotation time nodes, and merge them with the preliminary rotation time data to generate a dynamic rotation crop time axis.

[0064] Specifically, based on the changing trends of real-time regional climate data and basic soil property data at different time points in the fusion feature set, the temporal distribution characteristics of key indicators of environmental factors, including temperature, precipitation, soil temperature and soil moisture, are extracted at typical growth stages such as crop sowing, jointing, heading and maturity.

[0065] Using the sliding window statistical method, the moving mean and gradient rate of change of each environmental factor in the fused feature set are calculated over a continuous period. The expression is as follows:

[0066] ;

[0067] in, Indicates a point in time The moving average of environmental factors, This indicates the width of the sliding window. This represents the offset within the sliding window. Indicates environmental factors at a given time point The value of , Indicates the current point in time;

[0068] ;

[0069] in, Indicates a point in time The gradient rate of change of environmental factors Indicates environmental factors at a given time point The value of , Indicates environmental factors at a given time point The value of , Indicates the time interval for calculating the difference;

[0070] By combining the inflection points of the phased changes in crop carbon flux time series data, the time period that best matches the current climate conditions and soil conditions is located, and the start and end nodes of each growth stage in the original preliminary crop rotation time data are adjusted.

[0071] For example, if the initial crop rotation time data shows that the corn planting period is in early May, and the fused features show that the temperature is more stable and the soil moisture is more suitable in mid-May, then the corn planting node will be dynamically adjusted to mid-May.

[0072] The adjusted time points are weighted and merged with the initial crop rotation time data, and arranged in chronological order to generate a dynamic crop rotation timeline.

[0073] S2. Based on the dynamic crop rotation timeline, the growth stages are divided. By integrating the time series data of crop carbon flux in this region with soil microbial activity indicators, and using interpretable machine learning methods, the carbon sink stages and carbon source stages of wheat and corn at different growth stages are identified, and a staged carbon state sequence is dynamically generated.

[0074] It should be noted that by accurately dividing the growth stages of wheat and maize based on a dynamic crop rotation timeline, and combining regional crop carbon flux time-series data with soil microbial activity indicators, interpretable machine learning methods are introduced at the time-series scale to quantitatively identify the carbon budget characteristics of different growth stages. This enables automatic classification and dynamic identification of carbon sink and carbon source stages, thereby constructing a staged carbon state sequence with stage continuity and response accuracy. This overcomes the limitations of traditional methods that rely on empirical segmentation or static classification to determine carbon flow characteristics, and enhances the spatiotemporal analytical capability for changes in the carbon cycle state in the crop rotation system.

[0075] The beneficial effects are as follows: by integrating regional crop carbon flux time-series data and soil microbial activity indicators, an interpretable machine learning method is used to dynamically identify carbon sink and carbon source stages at different growth stages. This differs from existing technologies that commonly use fixed growth cycle divisions and empirical parameter settings for carbon emission calculations. Existing technologies often infer carbon budget status based on crop planting calendars or single climate parameters, making it difficult to accurately capture the impact of climate fluctuations and changes in soil activity on carbon cycle stages, resulting in insufficient spatiotemporal adaptability and dynamic response capabilities of the assessment results. In contrast, by constructing a staged carbon state sequence, high-resolution identification of the carbon flow transformation status of rotation crops is achieved, making carbon footprint assessment more consistent with actual farming conditions and environmental disturbances, and improving the accuracy of carbon emission identification and the reliability of dynamic control.

[0076] S2.1. Determine the time period of the growth stage based on the dynamic crop rotation time axis, organize the measured carbon flux data of wheat and corn in each corresponding time period, and form a carbon flux time series dataset.

[0077] Specifically, based on the dynamic crop rotation timeline, and according to the key growth stages marked on the timeline, such as wheat sowing, jointing, heading, grain filling, and maturity, and maize sowing, seedling stage, jointing, tasseling, grain filling, and maturity, the entire crop rotation cycle is divided into continuous growth stage periods. Using each growth stage period as an index, the corresponding measured carbon flux data for wheat and maize are retrieved. Time alignment is performed uniformly by day or hour, and missing and outlier values ​​are removed before linear interpolation is used to complete the data. Finally, the wheat carbon flux data and maize carbon flux data within each stage period are categorized and summarized to construct a carbon flux time-series dataset.

[0078] S2.2. Standardize and fuse the carbon flux time series dataset with soil microbial activity indicators from the same period to form a multidimensional environmental response feature set.

[0079] S2.2.1. The carbon flux time series dataset is time-aligned, interpolated, and smoothed with the soil microbial activity index of the same period to generate a joint data table.

[0080] Specifically, when processing the carbon flux time series dataset with soil microbial activity indicators of the same period, a time index is established according to a uniform time resolution (e.g., daily), and the carbon flux time series dataset and soil microbial activity indicators are matched with the time index respectively, and non-overlapping data points are removed.

[0081] For missing data points, linear interpolation was used to fill in the gaps. The interpolated values ​​were calculated as a weighted average of adjacent valid observations. After interpolation, the carbon flux time series dataset and soil microbial activity index were smoothed using a three-point moving average method to reduce the interference of short-term fluctuations on feature representation.

[0082] The time-series dataset of carbon flux, after time alignment, interpolation, and smoothing, is concatenated with soil microbial activity indicators by time column to generate a joint data table.

[0083] S2.2.2 Standardize the joint data table and extract the time-series statistical features of carbon flux and the features of microbial activity. Perform weighted fusion and structural reconstruction to form a multidimensional environmental response feature set.

[0084] Specifically, the carbon flux time series dataset and soil microbial activity index data column in the joint data table are subjected to min-max standardization in sequence, and each column is uniformly scaled to a fixed range.

[0085] For the standardized carbon flux data, a sliding window statistical method was used to extract time-series statistical features of carbon flux, such as mean, range, and rate of change; for the standardized soil microbial activity index data, the mean and standard deviation were extracted as microbial activity features.

[0086] The time-series statistical characteristics of carbon flux and the characteristics of microbial activity are assigned weighting coefficients respectively, and then linearly weighted and fused to generate a weighted feature vector group.

[0087] The weighted feature vector groups are concatenated and rearranged in chronological order to form a multidimensional environmental response feature set with a unified structure.

[0088] S2.3. An integrated algorithm is used to dynamically identify the carbon sink and carbon source characteristics at different growth stages from a multidimensional environmental response feature set, and generate a staged carbon state label sequence.

[0089] Specifically, the data for each time node in the multidimensional environmental response feature set are used as input variables, including crop growth stage, soil moisture, soil temperature, precipitation, air temperature, and soil organic matter content corresponding to the time node. Multiple decision tree-based classification models are constructed, including random forest model, gradient boosting tree model, and extreme random tree model. Each set of time node data is sequentially input into multiple decision tree-based classification models, and the predicted probability values ​​of carbon sink stage labels and carbon source stage labels corresponding to each classification model are output. The label prediction probabilities of all classification models at the same time node are integrated and fused to form the final label probability distribution of the time node.

[0090] Based on the weighted voting rules, the prediction results of multiple decision tree models are integrated to determine the final stage carbon state label;

[0091] It should also be noted that the weighted voting rule is based on the model fusion concept in ensemble learning. It achieves a comprehensive judgment by assigning different weights to the prediction results of multiple base classifiers. The weights are typically determined based on the classification accuracy or other performance metrics of each base classifier on the validation set. Base classifiers with higher accuracy receive greater weights, thereby enhancing the stability and accuracy of the overall prediction. The weighted voting rule effectively reduces the prediction bias of a single model, improves the robustness and generalization ability of the classification results, and is widely used in the result fusion process of ensemble algorithms such as random forests and gradient boosting trees.

[0092] By organizing the data in chronological order, a phased carbon state label sequence is generated, enabling dynamic identification of carbon sink and carbon source characteristics at different stages of growth.

[0093] S2.4 Perform time series smoothing and anomaly detection processing on the phased carbon state label sequence to generate a phased carbon state sequence.

[0094] Specifically, for the phased carbon state label sequence, the sliding window method is used for time series smoothing, and short-term fluctuations are eliminated by calculating the weighted average of the labels within the window;

[0095] Based on statistical anomaly detection methods, outliers are identified in the smoothed sequence, and possible abnormal carbon states are marked. The outliers are then replaced by neighboring labels to correct the abnormal data, ultimately generating a continuous and stable phased carbon state sequence.

[0096] S3. Based on the phased carbon state sequence, the system integrates basic soil property data and real-time regional climate data, and uses an ensemble learning method to fit the nonlinear response relationship of straw mineralization rate, dynamically generating a carbon conversion factor sequence.

[0097] It should be noted that, based on a phased carbon state sequence, a dynamic process control logic is introduced. The carbon sink and carbon source states of crops at different growth stages are used as limiting conditions. Combined with basic soil property data and real-time regional climate data, an ensemble learning method is employed to dynamically and nonlinearly fit the straw mineralization rate, thereby generating a carbon conversion factor sequence that adjusts in real-time with changes in time, climate, and soil conditions. Compared to existing technologies that treat the straw mineralization rate as a static constant or rely on a single environmental factor, this approach, by introducing multidimensional time-series variables and a nonlinear modeling mechanism, achieves a high-precision expression of the dynamics of carbon flow in the straw mineralization process, effectively improving the spatiotemporal adaptability and scientific rigor of carbon conversion estimation.

[0098] The beneficial effects are as follows: Compared with existing methods that commonly estimate carbon footprint based on fixed carbon emission coefficients or empirical statistical models, this invention constructs a dynamic crop rotation timeline and combines it with staged carbon state sequences and carbon transformation factor sequences. Using ensemble learning methods, it achieves spatiotemporal dynamic calculations of CO2 and N2O emissions and changes in soil organic carbon, further generating a carbon dynamic curve and calculating the net carbon footprint value. This allows for real-time reflection of the comprehensive impact of crop growth stages, climate conditions, and soil property changes on the carbon emission process, significantly improving the timeliness and accuracy of carbon footprint assessment results. Compared with existing methods that roughly estimate carbon footprints based on annual emission factors, this invention can accurately identify carbon source-carbon sink transformation nodes under short-cycle, unstable climate, and heterogeneous soil conditions, demonstrating significant innovation in dynamic modeling and regulation mechanisms of carbon footprint.

[0099] S3.1 Based on the phased carbon state sequence, real-time regional climate data and basic soil property data are matched and feature fused according to the time dimension to generate a coupled regulation feature set.

[0100] Specifically, based on the phased carbon state sequence, real-time regional climate data and soil basic attribute data are matched according to the time dimension. First, the timestamps of real-time regional climate data and soil basic attribute data are unified to the same time resolution, and time alignment method is used to handle time series differences.

[0101] The time-matched data is fused to combine the time periods in the phased carbon state sequence with the corresponding real-time regional climate data and soil basic property data to form a coupled regulation feature set, which includes key features such as temperature, precipitation, soil texture and organic matter content, ensuring complete correspondence in time order, and finally outputting a spatiotemporally corresponding and fused coupled regulation feature set.

[0102] S3.2 Based on the coupled regulation feature set, construct time-series training samples, use ensemble learning to characterize the nonlinear response of straw mineralization rate, and generate carbon conversion factor sequences for each time period.

[0103] Specifically, based on the coupling regulation feature set, the training samples are divided according to time order. Each sample contains the climate characteristics, basic soil properties and stage carbon state sequence of the corresponding time period, forming a time-series training sample set.

[0104] For each time node in the time series training samples, the coupling regulation feature corresponding to the time node is extracted as the input variable, and the straw mineralization rate is taken as the target variable. Random forest model, gradient boosting tree model and extreme random tree model are constructed and trained respectively. The prediction ability is evaluated by the performance on the training set. Then, the prediction results are fused by the weighted ensemble method to finally generate an ensemble learning model that reflects the nonlinear relationship between straw mineralization rate and coupling regulation feature.

[0105] After training, for each time period, the coupling regulation features within that time period are input into the trained ensemble learning model. The prediction results of the carbon conversion factor values ​​from the random forest model, gradient boosting tree model, and extreme random tree model are obtained respectively. The prediction outputs are then integrated using a weighted fusion strategy to obtain the final carbon conversion factor prediction value for each time period. The carbon conversion factor sequence prediction for all time periods is completed in sequence to generate the carbon conversion factor sequence.

[0106] S4. Couple the carbon transformation factor sequence with the stage carbon state sequence in time and space, calculate the CO2 and N2O emissions and the change in soil organic carbon according to time period and spatial range, and generate a carbon dynamic curve through time series integration.

[0107] S4.1 Combine the carbon transformation factor sequence with the stage carbon state sequence, and match the corresponding farmland areas according to time and spatial coordinates to form a spatiotemporally coupled dataset.

[0108] Specifically, the carbon conversion factor sequence and the phased carbon state sequence are matched according to time stamps for each time period. Combined with the geospatial coordinate information of farmland in each time period, the carbon conversion factor and carbon state data in the corresponding area are selected.

[0109] Spatial aggregation and temporal arrangement of data from different time periods within the same space are performed to form a spatiotemporally coupled data table;

[0110] Using a spatiotemporal coupled data table as a unified data structure, including time dimension, spatial coordinates, corresponding carbon conversion factor values, and staged carbon state sequences, complete coupling of time and space information is achieved.

[0111] S4.2. Select relevant variables from the spatiotemporal coupled dataset and analyze them to construct a multivariate response function.

[0112] Specifically, based on preset carbon emission and carbon sequestration related indicators from the spatiotemporal coupled dataset, variables such as CO2 emission rate, N2O emission rate, soil organic carbon fixation rate, and influencing factors such as carbon conversion factor and stage carbon state sequence are selected.

[0113] Calculate the mean, standard deviation, and distribution characteristics of the selected variables to assess their representativeness; use Pearson correlation coefficient or Spearman correlation coefficient to analyze the correlation between variables, eliminate redundant or low-correlation variables, and ensure the representativeness and stability of the variable set.

[0114] Based on the selected variable set, multiple regression analysis or weighted regression methods are used to model changes in greenhouse gas emissions and soil carbon sinks as response variables. By minimizing the sum of squared residuals and fitting the regression coefficients, a quantitative relationship between each response variable and the response index is established, forming a multivariate response function that can reflect the comprehensive impact of multiple variables.

[0115] It should also be noted that when setting up carbon emission and carbon sequestration related indicators, key indicators such as CO2 emission rate, N2O emission rate, and soil organic carbon fixation are selected based on relevant industry standards and authoritative literature. The measurement frequency and spatial resolution of the indicators are determined by considering the characteristics of actual agricultural production and the environmental conditions of the study area, ensuring that the indicators can reflect the impact of different growth stages and management measures on carbon dynamics during crop rotation. Finally, based on historical observation data and experimental results, reasonable numerical ranges for each indicator are set for subsequent calibration and dynamic adjustment, enabling effective quantitative evaluation of carbon emission and carbon sequestration processes.

[0116] S4.3 Using a multivariate response function, calculate the CO2 and N2O emission rates and soil organic carbon fixation according to the time step and spatial resolution, and generate a distributed emission and carbon sequestration rate matrix.

[0117] Specifically, using a multivariate response function, the variables in the spatiotemporally coupled dataset are segmented according to a preset time step and spatial resolution. Data from each time period and space are then input and calculated to obtain the corresponding CO2 emission rate, N2O emission rate, and soil organic carbon fixation. The expressions are as follows:

[0118] ;

[0119] in, Indicates a point in time and spatial location CO2 emission rate, Indicates a point in time and spatial location The set of environment variables, Indicates spatial location, Represents a multivariate response function;

[0120] ;

[0121] in, Indicates a point in time and spatial location The N2O emission rate;

[0122] ;

[0123] in, Indicates a point in time and spatial location Soil organic carbon fixation;

[0124] The calculation results are arranged sequentially according to the time step and spatial unit location to form a distributed emission and carbon sequestration rate matrix.

[0125] It should also be explained that the specific steps for setting the time step and spatial resolution are as follows: Based on the actual size of the study area and the frequency of data collection, determine the time step to meet the time sensitivity of crop growth cycles and climate change; typically, a daily or hourly time step is selected. Based on the spatial heterogeneity of the region and the data resolution requirements, determine the spatial resolution, such as by using a fixed grid partitioning method to divide the area into several spatial units, each corresponding to a specific geographical area. Combining the time step and spatial resolution, a spatiotemporal grid is formed, providing a basic data framework for the subsequent dynamic calculation of carbon emissions and carbon sequestration rates.

[0126] S4.4 Integrate and accumulate the distributed emissions and carbon sequestration rate matrix to calculate the total greenhouse gas emissions and soil carbon sink changes over the time period, generating a carbon dynamic curve.

[0127] Specifically, based on the distributed emission and carbon sequestration rate matrix, the CO2 emission rate and N2O emission rate are summed in the spatial dimension according to the time step and spatial resolution to obtain the total emission rate at each time step.

[0128] The total carbon sequestration of soil organic carbon is summed in the spatial dimension to obtain the total carbon sequestration at each time step. For the greenhouse gas emission rate and soil carbon sequestration rate at each time step, the emission rate and carbon sequestration rate within each spatial unit are first spatially weighted and summed according to the spatial resolution to obtain the total emission rate and total carbon sequestration at each time step. Based on the time step length, the total emission rate and total carbon sequestration at all time steps are cumulatively summed to calculate the changes in total greenhouse gas emissions and soil carbon sink within a specified time period. The cumulative emissions and carbon sequestration within the time period are plotted as a carbon dynamic curve changing with time.

[0129] S5. Extract emission rates and soil carbon sequestration rates from the carbon dynamic curve, perform periodic cumulative analysis, calculate total greenhouse gas emissions and total soil carbon sequestration, obtain net carbon footprint value, and construct a dynamic correction mechanism with meteorological anomaly factors and soil heterogeneity factors as inputs to perform spatiotemporal adaptive correction of net carbon footprint value and generate net carbon footprint data.

[0130] S5.1 Extract the CO2 and N2O emission rates and soil carbon fixation rates for each time period from the carbon dynamic curve, and summarize them to form a periodic carbon flux dataset.

[0131] Specifically, based on the carbon dynamic curve, according to the preset time period, data on CO2 emission rate, N2O emission rate and soil carbon fixation rate are extracted for each time period, and the average rate or total amount within the time period is calculated to form CO2 emission data, N2O emission data and soil carbon fixation data for each time period. The data are then summarized in chronological order to generate a periodic carbon flux dataset.

[0132] It should also be noted that the preset time period division is based on the crop growth stage division standard determined in the dynamic crop rotation time axis, combined with the characteristics of crop physiological cycle and changes in environmental conditions, to ensure that the time period corresponds to the key growth stage of the crop and can accurately reflect the carbon flux change pattern in different periods.

[0133] S5.2 Perform time-series cumulative analysis on the periodic carbon flux dataset to summarize the total greenhouse gas emissions and total soil carbon sequestration within the crop rotation cycle.

[0134] Specifically, for the periodic carbon flux dataset, the CO2 emission rate and N2O emission rate at each time point are accumulated and summed according to the preset crop rotation cycle time period to obtain the total greenhouse gas emissions within the crop rotation cycle.

[0135] Meanwhile, the soil organic carbon fixation rate at each time point is cumulatively summed to obtain the total soil carbon sequestration within the crop rotation cycle, ultimately forming a summary of total greenhouse gas emissions and total soil carbon sequestration data.

[0136] It should also be noted that the basis for pre-setting the crop rotation cycle time period includes the annual planting system arrangement of the dominant crops (wheat and corn) in the region, combined with the key agronomic nodes such as sowing, jointing, heading, grain filling and maturity recorded in agricultural meteorological data, to divide the crop rotation cycle into multiple biologically significant continuous stages, such as the sowing to jointing stage, the jointing to grain filling stage and the grain filling to harvest stage, to ensure that there is a correlation between carbon flux changes and crop growth process. The division criteria are based on many years of historical field observation records and regional dominant agronomic procedures.

[0137] S5.3 Calculate the difference between the total greenhouse gas emissions and the total soil carbon sequestration to obtain the net carbon footprint value.

[0138] Specifically, the total greenhouse gas emissions within the crop rotation cycle and the total soil organic carbon sequestration are extracted from the periodic carbon flux dataset. Units are then uniformly converted according to the time step to ensure consistency, for example, to kgC equivalent / ha. A difference calculation is then performed, expressed as:

[0139] ;

[0140] in, Indicates the net carbon footprint value. Indicates total greenhouse gas emissions. This indicates the total amount of organic carbon fixed in the soil;

[0141] Example: When kgC equivalent / ha kgC equivalent / ha, then kgC equivalent / ha; finally, The corresponding crop rotation cycle is marked according to the time dimension to form a net carbon footprint data sequence.

[0142] S5.4 Based on the net carbon footprint value, a spatiotemporal perturbation factor system is constructed by combining the perturbation factors extracted from real-time regional climate data and basic soil property data, and then feature-aligned and integrated with the net carbon footprint value to generate a high-dimensional input sample set.

[0143] Specifically, within each crop rotation cycle, the net carbon footprint value is extracted as the target variable, and the corresponding time period and spatial location are recorded.

[0144] Environmental disturbance factors such as wind speed, temperature, precipitation, humidity and solar radiation are extracted from real-time regional climate data. The average value of each time period is calculated according to the time step and spatial resolution consistent with the net carbon footprint value to form a subset of environmental disturbance factors.

[0145] Soil disturbance factors such as organic matter content, pH value, density, bulk density and moisture are extracted from basic soil property data and matched to corresponding farmland areas according to spatial coordinates to form a subset of soil disturbance factors.

[0146] A subset of environmental disturbance factors and a subset of soil disturbance factors are jointly matched in time and space to generate a complete set of spatiotemporal disturbance factors.

[0147] The set of spatiotemporal disturbance factors and the net carbon footprint value are matched one-to-one according to the same time period and spatial location to form a high-dimensional input sample set for each cycle.

[0148] S5.5. Using a high-dimensional input sample set, fit the nonlinear response of the net carbon footprint value to spatiotemporal disturbances to generate a dynamic adjustment function.

[0149] S5.5.1 Using a nonlinear correlation measurement method, the complex dependency relationship between net carbon footprint value and spatiotemporal disturbance factors is identified from a high-dimensional input sample set, and variable correlation characteristics are obtained.

[0150] Specifically, all perturbation factors are extracted from the high-dimensional input sample set as analysis variables, and the net carbon footprint value is taken as the target variable. The data sequence of each perturbation factor and the net carbon footprint value is discretized to construct a joint distribution matrix. The mutual information of the joint distribution is calculated, and then the maximum mutual information value between each variable is found based on the maximum information coefficient algorithm to obtain the nonlinear correlation value.

[0151] By traversing different discretization grid partitioning schemes, the partitioning result corresponding to the maximum mutual information is selected to ensure that the strongest nonlinear dependence between variables is captured, and finally the maximum information coefficient value between each perturbation factor and the net carbon footprint value is output.

[0152] Then, the perturbation factors are sorted in descending order according to the magnitude of the nonlinear correlation values. The perturbation factors with nonlinear correlation values ​​greater than the set maximum information coefficient discrimination threshold are selected as perturbation factors with significant dependencies, thus forming the variable association features.

[0153] It should also be explained that the specific steps for setting the maximum information coefficient discrimination threshold are as follows: Based on the historical high-dimensional input sample set, calculate the distribution of the maximum information coefficient between the perturbation factor and the net carbon footprint value; combined with the statistical characteristics of the sample data, select the critical value in the distribution that can distinguish between significantly correlated and weakly correlated variables as the initial maximum information coefficient discrimination threshold; evaluate the variable screening effect and predictive performance under different maximum information coefficient discrimination thresholds through cross-validation, and determine the optimal maximum information coefficient discrimination threshold; the maximum information coefficient discrimination threshold is generally set between 0.2 and 0.4, and in the example, it is set to 0.3, which can effectively filter noisy variables and retain key perturbation factors; finally, the maximum information coefficient discrimination threshold ensures the accuracy and generalization ability of variable association features and improves the stability and interpretability of the dynamic adjustment function.

[0154] S5.5.2 Based on the characteristics of variable correlation, nonlinear regression is used to fit the dynamic response curve of the net carbon footprint value relative to the spatiotemporal disturbance factor to form a preliminary adjustment function.

[0155] Specifically, based on the variable correlation characteristics, the corresponding spatiotemporal perturbation factors and net carbon footprint values ​​are selected from the high-dimensional input sample set and arranged in time series to form training samples;

[0156] Using a nonlinear regression method, the input features in the training samples are fitted with the net carbon footprint value. The model parameters are iteratively adjusted to minimize the loss function and optimize the fitting effect of the net carbon footprint value on the response curve of spatiotemporal disturbance factors.

[0157] After fitting is complete, a preliminary adjustment function is generated that reflects the dynamic response of the net carbon footprint value.

[0158] For example, existing nonlinear regression algorithms such as support vector regression or random forest regression are used in the fitting process to determine the optimal one to improve fitting accuracy and ensure that the adjustment function can accurately reflect the nonlinear change of net carbon footprint value to perturbation factors.

[0159] S5.5.3 Optimize the initial adjustment function through cross-validation and residual analysis to generate the dynamic adjustment function.

[0160] Specifically, based on the initial adjustment function, the training sample set is divided. The k-fold cross-validation method is used to divide the sample set into k subsets. One subset is selected as the validation set and the rest are used as the training set. The fitting effect of the adjustment function is trained and validated respectively.

[0161] Calculate the residual value for each round of validation, including the error between the predicted value and the actual net carbon footprint value, analyze the residual distribution and trend, and identify deviations and outliers in the fitting.

[0162] By combining the results of cross-validation and residual analysis, the nonlinear regression parameters are adjusted or a more suitable nonlinear regression algorithm is selected. The adjustment function is iteratively optimized until the residuals meet the preset residual threshold and are stable on each validation set. A high-precision dynamic adjustment function that reflects the dynamic changes of net carbon footprint is generated.

[0163] It should also be noted that when setting the residual threshold, the mean and standard deviation of the residuals are calculated based on the statistical analysis of the error distribution between historical net carbon footprint values ​​and predicted values ​​to determine the normal fluctuation range of the residuals. A reasonable error tolerance range is set in combination with the carbon flux monitoring accuracy and actual measurement error in the study area. Empirical rules are adopted, such as setting the residual threshold as the residual mean plus or minus two standard deviations, to ensure that residuals exceeding the residual threshold are considered abnormal. Through comparison and verification of the prediction effect under different residual thresholds, a value that is neither too lenient nor too strict is selected as the final preset residual threshold, under the premise of ensuring prediction accuracy and generalization ability.

[0164] S5.6. Adaptively calibrate the net carbon footprint value using a dynamic adjustment function to generate net carbon footprint data.

[0165] Specifically, based on the dynamic adjustment function, the net carbon footprint value and the corresponding spatiotemporal disturbance factor data are read sequentially according to the time series, and the net carbon footprint value is mapped to the dynamic adjustment function to obtain the calibration adjustment amount;

[0166] The calibration adjustment is weighted and superimposed with the original net carbon footprint value to generate calibrated net carbon footprint data; the steps are repeated for all time points to form complete net carbon footprint data.

[0167] For example, the weights of the calibration adjustment can be dynamically adjusted based on the confidence level of the adjustment function to ensure the adaptability and dynamic response capability of the calibration process.

[0168] S6. Identify the peak carbon emission period and straw treatment period from the net carbon footprint data, and optimize the straw return time, tillage depth and crop rotation sequence based on the carbon dynamic curve and net carbon footprint data to generate a closed-loop agronomic management strategy.

[0169] S6.1 Analyze net carbon footprint data, extract carbon emission fluctuation information within the crop rotation cycle, identify peak carbon emission periods and key straw treatment stages, and generate a peak carbon emission period identification table and a key straw treatment period list.

[0170] Specifically, based on net carbon footprint data, the data is segmented according to the time period of the crop rotation cycle, the carbon emission change rate at each time point is calculated, the carbon emission rate is smoothed using the sliding window method, the local maximum value in the smoothing curve is identified as the carbon emission peak point, and the peak carbon emission period is matched and identified by combining the growth stage of the rotation crop and the time of agricultural operation.

[0171] Based on the time periods corresponding to peak carbon emissions and combined with the time records of straw treatment operations, the critical period for straw treatment was identified as the stage in which operations such as straw return to the field, crushing and covering occur in concentrated areas and cause significant fluctuations in greenhouse gas emissions.

[0172] The identified peak carbon emission periods are compiled into a peak carbon emission period identification table, and the key straw treatment period time points are compiled into a key straw treatment period list.

[0173] For example, the sliding window width can be set to 7 days (example), and the threshold for determining peak carbon emissions can be set to the average carbon emission rate plus 1.5 times the standard deviation of the example.

[0174] S6.2 Based on the carbon emission peak period identification table and the list of key periods for straw treatment, the management measures records for the corresponding time periods in the carbon dynamic curve are aligned with time, and the sensitivity response of management measures to CO2 and N2O emissions is analyzed to extract the set of carbon emission response sensitivity factors.

[0175] Specifically, based on the carbon emission peak period identification table and the straw treatment key period list, the management measures records for the corresponding time periods in the carbon dynamic curve are precisely matched according to the time axis. The characteristics of the management measures and CO2 and N2O emission rate data for each time period are extracted. When using correlation analysis, the time series data of management measures and the CO2 and N2O emission rate data for the corresponding time periods are collected. By calculating the Pearson correlation coefficient or Spearman rank correlation coefficient, the linear or nonlinear relationship strength between changes in management measures and changes in greenhouse gas emissions is quantified to obtain the response intensity index of management measures to CO2 and N2O emissions. Management measures with response intensity exceeding the preset sensitivity threshold are selected as carbon emission response sensitivity factors and summarized to form a carbon emission response sensitivity factor set.

[0176] It should also be noted that the preset sensitivity threshold is based on the statistical distribution characteristics of historical greenhouse gas emission data and the response intensity of management measures to determine the specific sensitivity threshold range. Typically, a value with a response intensity significantly higher than the random fluctuation level is selected as the sensitivity threshold to ensure that the selected management measures have a real and stable impact on CO2 and N2O emissions, while avoiding misjudgments caused by noise.

[0177] S6.3 Based on the set of carbon emission response sensitive factors, adjust the straw return time, tillage depth and crop rotation sequence to generate a set of optimized agronomic management strategies.

[0178] Specifically, based on the set of sensitive factors for carbon emission response, sensitive factors related to straw return time, tillage depth, and crop rotation sequence are extracted. These sensitive factors are then classified and prioritized. Combining historical agronomic management data and crop rotation timelines, multiple sets of straw return time adjustment schemes, tillage depth optimization schemes, and crop rotation sequence adjustment schemes are developed. Agronomic management strategy evaluation indicators are used to predict and screen the effects of each scheme. Finally, a diverse set of agronomic management strategies that meet the preset emission reduction and carbon sequestration targets are selected and compiled into an optimized agronomic management strategy scheme set.

[0179] S6.4. For each set of optimized agronomic management strategies, conduct carbon footprint simulation assessment and comparative analysis, select the combination with the lowest greenhouse gas emissions and the best soil carbon sequestration capacity, and generate a closed-loop agronomic management strategy.

[0180] Specifically, for each scheme in the set of optimized agronomic management strategies, based on net carbon footprint data and agronomic management strategy influencing factors, the carbon flux calculation method is used to cumulatively calculate the CO2 and N2O emission rates and soil organic carbon fixation rates of the optimized agronomic management strategies within the crop rotation cycle according to the time step and spatial resolution for each set of optimized agronomic management strategies. The emission rates and carbon fixation rates at each time step are integrated over time to obtain the total greenhouse gas emissions and the total soil carbon sequestration.

[0181] By comparing and analyzing the total greenhouse gas emissions and the total soil carbon sequestration, and setting multiple evaluation criteria based on greenhouse gas emissions and soil carbon sequestration capacity, a combination of schemes with the lowest greenhouse gas emissions and the highest soil carbon sequestration capacity is selected. The selected schemes are then verified to confirm their stability and applicability, ultimately forming a closed-loop agronomic management strategy as the basis for decision-making on agronomic management optimization.

[0182] For example, greenhouse gas emissions are defined as annual emissions of less than 5 tons of CO2 equivalent, and soil carbon sequestration capacity is defined as annual carbon sequestration of more than 2 tons of carbon equivalent.

[0183] It should also be noted that the multi-indicator evaluation standard is based on the relevant industry standards and technical guidelines on greenhouse gas emissions and soil carbon sequestration. It combines the goals of agricultural ecological environmental protection with the specific needs of crop rotation agronomy practices. Through research and experience summarization, key indicators such as total greenhouse gas emissions, total soil carbon sequestration, carbon emission intensity, and crop yield are selected as evaluation standards to ensure that the evaluation system is scientific, reasonable, and meets the requirements of practical application.

[0184] This embodiment also provides a computer device applicable to the carbon footprint assessment method of wheat-corn rotation straw return to the field, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the carbon footprint assessment method of wheat-corn rotation straw return to the field as proposed in the above embodiment.

[0185] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0186] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the carbon footprint assessment method for wheat-corn crop rotation with straw return to the field as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0187] In summary, this invention achieves a detailed temporal characterization of carbon emissions and carbon sequestration behavior by dynamically identifying the carbon sink and carbon source status of crops at different growth stages, thereby improving the timeliness and accuracy of carbon dynamic simulation; it effectively integrates crop physiological cycles and environmental response characteristics to accurately reflect the comprehensive impact of multiple factors on the carbon cycle, providing a high spatiotemporal resolution input basis for dynamic carbon footprint assessment and promoting the scientific and rational nature of carbon footprint assessment.

[0188] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for evaluating the carbon footprint of wheat-corn rotation straw returning to field, characterized in that: comprising, Collect preliminary rotation time data, real-time regional climate data and soil basic attribute data, and perform fusion analysis to generate a dynamic rotation crop time axis; According to the dynamic rotation crop time axis, the growth stages are divided, the carbon flux time series data and the soil microbial activity index of crops in the region are fused, and an interpretable machine learning method is used to identify the carbon sink stage and the carbon source stage of wheat and corn in different growth stages, and a stage carbon state sequence is dynamically generated; Based on the stage carbon state sequence, the soil basic attribute data and the real-time regional climate data are fused, an integrated learning method is used to fit the nonlinear response relationship of straw mineralization rate, and a carbon conversion factor sequence is dynamically generated; The carbon conversion factor sequence and the stage carbon state sequence are spatio-temporally coupled, the CO2 and N2O emissions and the soil organic carbon change are calculated according to the time period and the spatial range, and the carbon dynamic curve is generated by time series integration; From the carbon dynamic curve, the emission rate and the soil carbon fixation rate are extracted, the periodic accumulation analysis is performed, the total greenhouse gas emission and the total soil carbon fixation are calculated, the net carbon footprint value is obtained, and the meteorological anomaly factor and the soil heterogeneity factor are input to build a dynamic correction mechanism to adaptively correct the net carbon footprint value in space and time, and generate the net carbon footprint data; From the net carbon footprint data, the carbon emission peak period and the straw treatment stage are identified, and based on the carbon dynamic curve and the net carbon footprint data, the straw returning time, the tillage depth and the rotation sequence are optimized to generate a closed-loop agricultural management strategy.

2. The method for evaluating carbon footprint of wheat-maize crop rotation straw returning according to claim 1, wherein: The specific steps of generating the dynamic rotation crop time axis are as follows, The preliminary rotation time data, real-time regional climate data and soil basic attribute data are standardized and weighted fused to generate a fusion feature set; From the fusion feature set, the time sequence response characteristics of crop growth stages to environmental conditions are extracted, the rotation time nodes are dynamically adjusted, and the dynamic rotation crop time axis is generated by fusing with the preliminary rotation time data.

3. The method for evaluating carbon footprint of wheat-maize crop rotation straw returning according to claim 1, wherein: The specific steps of dynamically generating the stage carbon state sequence are as follows, According to the dynamic rotation crop time axis, the growth stage time period is determined, the measured carbon flux data of wheat and corn in each corresponding period is arranged to form a carbon flux time series data set; The carbon flux time series data set and the soil microbial activity index at the same period are standardized and feature fused to form a multi-dimensional environmental response feature set; An integrated algorithm is used to dynamically identify the carbon sink and carbon source characteristics of different growth stages from the multi-dimensional environmental response feature set to generate a stage carbon state label sequence; The stage carbon state label sequence is time series smoothed and abnormally detected to generate a stage carbon state sequence.

4. The method for evaluating carbon footprint of wheat-maize crop rotation straw returning according to claim 3, wherein: The specific steps of standardizing and feature fusing the carbon flux time series data set and the soil microbial activity index at the same period to form a multi-dimensional environmental response feature set are as follows, The carbon flux time series data set and the soil microbial activity index at the same period are time-aligned, interpolated and smoothed to generate a joint data table; The joint data table is standardized, and the carbon flux time series statistical features and the microbial activity features are extracted, weighted fused and structure reconstructed to form a multi-dimensional environmental response feature set.

5. The method for evaluating carbon footprint of wheat-maize crop rotation straw return to field according to claim 1, wherein: The specific steps of dynamically generating the carbon conversion factor sequence are as follows, Based on the phased carbon state sequence, real-time regional climate data and soil basic attribute data are matched and characteristic fused in time dimension to generate a coupling adjustment characteristic set; Based on the coupling adjustment characteristic set, a time series training sample is constructed, an integrated learning is used to depict the nonlinear response of straw mineralization rate, and a carbon conversion factor sequence is generated in each time period.

6. The method for evaluating carbon footprint of wheat-maize crop rotation straw return to field according to claim 1, wherein: The carbon dynamic curve is generated by time series integration, and the specific steps are as follows, The carbon conversion factor sequence is combined with the phased carbon state sequence, and the corresponding farmland area is matched according to the time and space coordinates to form a spatio-temporal coupling data set; From the spatio-temporal coupling data set, relevant variables are screened and analyzed to construct a multivariate response function; Using the multivariate response function, the CO2 and N2O emission rate and the soil organic carbon fixation amount are calculated according to the time step and the spatial resolution, and a distributed emission and carbon fixation rate matrix is generated; Integrate the distributed emission and carbon fixation rate matrix to calculate the total greenhouse gas emissions and soil carbon sink changes in the time period, and generate a carbon dynamic curve.

7. The method for evaluating carbon footprint of wheat-maize crop rotation straw return to field according to claim 1, wherein: The specific steps of obtaining the net carbon footprint value are as follows, From the carbon dynamic curve, the CO2 and N2O emission rate and the soil carbon fixation rate in each time period are extracted to form a periodic carbon flux data set; The periodic carbon flux data set is analyzed by time series accumulation, and the total greenhouse gas emissions and soil carbon fixation amount in the crop rotation period are summarized respectively; The total greenhouse gas emissions and soil carbon fixation amount are calculated by difference to obtain the net carbon footprint value.

8. The method for evaluating carbon footprint of wheat-maize crop rotation straw mulching according to claim 1, wherein: The specific steps of generating the net carbon footprint data are as follows, According to the net carbon footprint value, the disturbance factors extracted from the real-time regional climate data and the soil basic attribute data are combined to construct a spatio-temporal disturbance factor system, and the net carbon footprint value is characteristic aligned and integrated to generate a high-dimensional input sample set; Using the high-dimensional input sample set, the nonlinear response of the net carbon footprint value to the spatio-temporal disturbance change is fitted to generate a dynamic adjustment function; Using the dynamic adjustment function, the net carbon footprint value is adaptively calibrated to generate the net carbon footprint data.

9. The method for evaluating carbon footprint of wheat-maize crop rotation straw mulching according to claim 8, characterized in that: The specific steps of fitting the nonlinear response of the net carbon footprint value to the spatio-temporal disturbance change by using the high-dimensional input sample set are as follows, Using a nonlinear correlation measurement method, the complex dependence relationship between the net carbon footprint value and the spatio-temporal disturbance factor is identified from the high-dimensional input sample set to obtain variable association characteristics; Based on the variable association characteristics, the dynamic response curve of the net carbon footprint value relative to the spatio-temporal disturbance factor is fitted by using nonlinear regression to form a preliminary adjustment function; The preliminary adjustment function is optimized by cross-validation and residual analysis to generate a dynamic adjustment function.

10. The method for evaluating carbon footprint of wheat-maize crop rotation straw return to field according to claim 1, wherein: The specific steps of generating the closed-loop agricultural management strategy are as follows, The net carbon footprint data is analyzed to extract carbon emission fluctuation information in the crop rotation period, identify the carbon emission peak period and the key straw processing stage, and generate a carbon emission peak period identification table and a straw processing key period list; Based on the carbon emission peak period identification table and the straw processing key period list, the time alignment is performed with the management measure records in the corresponding time period in the carbon dynamic curve, the sensitivity response of the management measures to CO2 and N2O emission is analyzed, and a carbon emission response sensitivity factor set is extracted; According to the carbon emission response sensitive factor set, the straw returning time, the tillage depth and the rotation sequence are adjusted to generate a set of optimized agricultural management strategies; For each set of optimized agricultural management strategies, carbon footprint simulation evaluation and comparative analysis are performed to screen out the combination with the lowest greenhouse gas emission and the optimal soil carbon sequestration capacity, and a closed-loop agricultural management strategy is generated.

Images