A method and system for reducing greenhouse gas emissions through intercropping system design

Abstract

The present application belongs to the field of agricultural ecology and environmental engineering technology, and specifically relates to a method and system for reducing greenhouse gas emissions through intercropping system design, which comprises multi-source sensing, multi-scale coupled modeling and adaptive optimization control, collects environmental and crop data through a distributed sensing network, constructs a three-level model including crop interaction, soil microenvironment and gas generation-transmission, identifies key emission driving factors and their action windows, generates optimal intercropping configurations and dynamically controls them by an execution mechanism. The system integrates digital twin verification, causal inference atlas and safety decision engine, realizes emission reduction and stable yield collaboration, and forms a closed-loop intelligent control system.

Description

Technical Field

[0001] This invention belongs to the field of agricultural ecology and environmental engineering technology, specifically a method and system for reducing greenhouse gas emissions through intercropping system design. Background Technology

[0002] Against the backdrop of escalating global climate change and the advancement of the "dual-carbon" strategy, the emission reduction and carbon sequestration functions of agricultural ecosystems, as important sources of greenhouse gas emissions and potential carbon sinks, have attracted significant attention. Traditional monoculture farming, while ensuring food supply, easily leads to soil organic matter loss, excessive nitrogen fertilizer application, and the continuous release of non-carbon dioxide greenhouse gases such as methane and nitrous oxide, making it difficult to balance production efficiency and ecological sustainability. Intercropping systems, through the complementarity of different crops in resource utilization and physiological ecology, can maintain or increase yields while suppressing soil greenhouse gas emissions, becoming a key pathway for the green and low-carbon transformation of agriculture.

[0003] Existing technologies include intercropping of leguminous and gramineous crops (utilizing nitrogen fixation to reduce nitrogen fertilizer input) and pairing deep-rooted and shallow-rooted crops (improving soil aeration and inhibiting methanogenic bacteria activity). These schemes, based on the physiological and ecological complementarity of crops, have shown certain emission reduction potential in small-scale field trials. However, the design of existing intercropping systems mostly relies on empirical ratios or static crop combination rules, lacking a systematic analysis of the coupling relationship between the dynamic processes of greenhouse gas emissions and crop interaction mechanisms. A closed-loop technical framework of "emission source identification—quantification of key driving factors—dynamic optimization of intercropping parameters" has not yet been established.

[0004] Existing intercropping emission reduction strategies suffer from deep-seated technical contradictions: a modeling and control gap exists between the complexity of the intercropping system structure and the nonlinear and time-varying nature of greenhouse gas emission responses. Soil greenhouse gas emissions are influenced by structural parameters such as crop type and row ratio, and are also sensitive to dynamic environmental disturbances such as rainfall, fertilization, and temperature and humidity fluctuations. These disturbances generate complex cascading effects in the intercropping system. Existing technologies, based solely on static combinations or fixed row ratio designs, cannot capture critical emission windows during the growth period, leading to a misalignment between emission reduction measures and emission peaks. Furthermore, the lack of a mechanistic model supporting "crop configuration—soil microenvironment—gas flux" makes high-frequency monitoring and manual adjustments difficult to achieve precise intervention. Moreover, existing technologies do not actively embed intercropping into the prediction-feedback-optimization control loop for greenhouse gas emissions, resulting in a disconnect between system design and emission reduction targets.

[0005] Therefore, the present invention provides a method and system for reducing greenhouse gas emissions through intercropping system design. Summary of the Invention

[0006] This invention provides a method and system for reducing greenhouse gas emissions through intercropping system design, aiming to solve the technical challenge of a modeling and control gap between the system parameters of intercropping structures and the dynamic response to greenhouse gas emissions in existing technologies. To achieve the above-mentioned objective, this invention constructs a closed-loop intelligent design method for intercropping systems that integrates multi-source environmental sensing, crop interaction mechanism modeling, soil biogeochemical process simulation, and adaptive optimization control, and forms a supporting system architecture with real-time feedback and adjustment capabilities.

[0007] The method of this invention includes the following steps: First, a distributed multimodal sensor network is deployed in the target farmland area to collect key environmental variables and crop growth status parameters affecting greenhouse gas emissions; second, a multi-scale coupled model is constructed based on the collected data, including crop root architecture interaction, canopy microclimate coupling, soil nitrogen and carbon cycle fluxes, and dynamic response of microbial communities; third, the model is used to dynamically simulate and predict greenhouse gas emission fluxes under different intercropping configurations, and the key driving factors that contribute the most to emissions and their window of action under specific growth stages and environmental disturbance conditions are identified; finally, an optimal intercropping structure parameter set is generated based on the prediction results, and the field planting layout is dynamically adjusted or a preset intervention strategy is implemented through an execution mechanism to effectively suppress peak greenhouse gas emissions.

[0008] Furthermore, the distributed multimodal sensor network consists of a soil temperature and humidity sensor, a soil redox potential probe, an atmospheric carbon dioxide and methane concentration detection unit, a canopy photosynthetically active radiation receiver, a rainfall event trigger recorder, and a crop stem flow monitoring device. Each sensor node uploads data to the edge computing server via a low-power wide-area communication protocol. The soil temperature and humidity sensor and the redox potential probe are embedded in soil layers at different depths to obtain vertical profile information, and the crop stem flow monitoring device is installed at the base of the main crop plant to reflect the dynamics of water absorption.

[0009] In a preferred embodiment of the present invention, the multi-scale coupling model comprises three levels: the first level is a crop interaction sub-model, used to describe the influence of different crop species on light interception efficiency, transpiration competition intensity, and root exudate interaction patterns under spatial configuration; the second level is a soil microenvironment sub-model, used to characterize changes in soil pore structure, fluctuations in organic matter decomposition rate, and shifts in nitrogen form transformation pathways caused by crop interactions; and the third level is a greenhouse gas generation-transport sub-model, used to quantify the net emission flux of methanogens and denitrifying bacteria after their activities are regulated by the aforementioned microenvironment factors, and to consider the diffusion resistance and time lag effects of gases in unsaturated soil media.

[0010] Furthermore, the crop interaction sub-model adopts an individual-based modeling method, treating each crop as an intelligent agent with independent physiological behavior rules. Its light-harvesting capacity is dynamically updated based on the three-dimensional geometry of the canopy and the shading relationship with neighboring crops, while the root expansion direction is guided by the local nutrient gradient and the secretion signals of neighboring roots. The soil microenvironment sub-model introduces enzyme kinetic equations to describe the cellulose degradation and protein mineralization processes, and combines Monod-type reaction functions to characterize the dependence of nitrification and denitrification rates on dissolved oxygen concentration, ammonium nitrogen content, and soluble organic carbon levels. The greenhouse gas generation-transport sub-model is based on Fick's diffusion law, and introduces lag terms to reflect the time delay characteristics of gas release after rainfall infiltration.

[0011] In a preferred embodiment of the present invention, the key driving factor identification mechanism is implemented by constructing a causal inference graph. This graph expresses the conditional probability dependencies between variables in the form of a Bayesian network, where nodes represent environmental disturbance events, crop configuration characteristics, or soil biochemical indicators, and edge weights reflect the information transmission intensity between variables. When a new sensor data stream is received, the system automatically updates the network posterior distribution and quantifies the marginal contribution of each configuration parameter to the current emission flux according to the Shapley value decomposition algorithm, thereby identifying the dominant factor and its sensitive period of action.

[0012] Furthermore, the optimal intercropping structure parameter set includes five dimensions: crop type combination, row spacing ratio, sowing density, bandwidth configuration, and rotation sequence. Its generation process adopts a multi-objective constrained optimization algorithm to minimize the cumulative greenhouse gas emission equivalent throughout the entire growth period while meeting the minimum yield guarantee threshold. A Pareto frontier screening mechanism is introduced during the optimization process to ensure that the output scheme achieves a balance between emission reduction efficiency and production stability. All candidate schemes must be virtually verified by a digital twin platform, and can only enter the execution stage when the simulated emission curve is lower than the historical baseline and there is no significant risk of yield loss.

[0013] In a preferred embodiment of the present invention, the actuator includes a programmable seeding robot, a variable fertilizer controller, and a remote irrigation scheduling module. The three receive operation instructions from the central decision engine through a unified instruction bus. The programmable seeding robot completes precise positioning and seeding based on the optimized row spacing ratio and bandwidth configuration. The variable fertilizer controller regulates the amount of nitrogen fertilizer applied according to crop needs to avoid local excessive accumulation. The remote irrigation scheduling module dynamically adjusts the irrigation frequency and the amount of water per irrigation based on soil moisture feedback to prevent anaerobic environment induced by water saturation.

[0014] Furthermore, the system also includes a greenhouse gas emission early warning module, which continuously compares the deviation between the measured emission flux and the model prediction value. Once an abnormal upward trend exceeding the preset tolerance range is detected, an emergency response procedure is immediately activated, triggering intervention measures such as the deployment of temporary shading nets, local oxygenation injection, or targeted spraying of microbial inhibitors to block the development of potential emission pulses.

[0015] In a preferred embodiment of the present invention, the central decision engine runs within a secure chip with a trusted execution environment. All sensor data is processed by a hardware-level encryption module before transmission, and the model inference process is also completed in an isolated sandbox to ensure that sensitive agricultural data is not leaked to the general operating system. In addition, the system supports offline mode operation and can still complete basic judgment and control output based on the locally cached historical pattern library during communication interruption.

[0016] Furthermore, the digital twin platform integrates high-resolution terrain data, long-term meteorological sequences, and soil texture classification information to construct a virtual farmland mirror covering the county scale. This allows users to conduct multi-scenario stress tests before deploying physical intercropping systems, including non-steady-state disturbance scenarios such as extreme drought, heavy rainfall impacts, and pest and disease outbreaks, thereby improving the robustness and adaptability of the final design scheme.

[0017] As a preferred embodiment of the present invention, the multi-objective constrained optimization algorithm incorporates a knowledge-guided mechanism. This mechanism extracts effective intercropping pairing rules from published agronomic literature and field trial databases, encodes them as soft constraints, and embeds them into the search space to accelerate convergence and avoid biologically infeasible solutions. At the same time, the system periodically receives anonymous performance feedback from other deployment points and updates local model parameters through a federated learning framework to achieve cross-regional experience sharing without exposing the original data.

[0018] Furthermore, the greenhouse gas emission equivalent calculation adopts a unified global warming potential conversion standard, converting methane and nitrous oxide emissions into carbon dioxide equivalents at fixed multiples, and taking into account the indirect carbon sink effect caused by crop residue returning to the field, making the emission reduction assessment more comprehensive and policy compatible.

[0019] In a preferred embodiment of the present invention, the causal inference graph supports dynamic topology reconstruction. When data from multiple consecutive reproductive cycles indicate a structural change in the original variable relationships, the system automatically triggers the graph retraining process, using incremental learning techniques to update the node connection structure and edge weight parameters, ensuring that the model always reflects the actual operating rules of the current ecosystem.

[0020] Furthermore, the variable fertilizer controller is equipped with a dual-channel feeding system, one channel delivering conventional nitrogen fertilizer and the other channel delivering slow-release nitrification inhibitor compound granules. The mixing ratio of the two can be adjusted in real time according to the soil nitrate nitrogen concentration, so that the nitrogen supply rhythm is precisely matched with the crop absorption peak, minimizing ineffective residues.

[0021] In a preferred embodiment of the present invention, the remote irrigation scheduling module adopts a pulse drip irrigation strategy, and the water supply time is strictly controlled within the range of maintaining moderate moisture in the root zone but not enough to trigger a reducing environment. Combined with soil conductivity monitoring, it prevents salt accumulation, thereby saving water while maintaining a soil metabolic pattern dominated by aerobic microorganisms.

[0022] Furthermore, the targeted spraying of microbial inhibitors in the emergency response procedure is performed by a high-precision spraying system carried by a drone. This system plans its flight trajectory based on an emission hotspot map and uses lidar to correct nozzle height and wind speed compensation parameters in real time, ensuring that the agent only acts on the target micro-area and avoids large-scale diffusion.

[0023] The beneficial effects of this invention are as follows: This invention discloses a method and system for reducing greenhouse gas emissions through intercropping system design. By leveraging a multi-scale coupled model and a causal inference mechanism, it accurately identifies key drivers of greenhouse gas emissions and their sensitive windows in the intercropping system. It utilizes a digital twin platform and multi-objective optimization algorithms to generate dynamic intercropping configuration schemes that balance emission reduction efficiency and production stability. This represents a fundamental shift from static empirical matching to a closed-loop control paradigm driven by mechanistic models. It ensures the continued effectiveness of the system's emission reduction function under complex environmental disturbances. Furthermore, it safeguards agricultural data sovereignty and model evolution capabilities through a secure chip and federated learning architecture. Simultaneously, it achieves seamless integration from decision-making to implementation through programmable actuators, avoiding operational delays and error accumulation caused by manual intervention. Ultimately, it forms a complete technology chain integrating perception, modeling, prediction, optimization, and execution, enabling the intercropping system to truly become an intelligent ecological unit actively participating in greenhouse gas regulation. Attached Figure Description

[0024] The present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a schematic diagram of the overall structure of a system for reducing greenhouse gas emissions through intercropping system design in this invention; Figure 2 This is a flowchart of a method for reducing greenhouse gas emissions through intercropping system design according to the present invention; Figure 3 This is a diagram illustrating the hierarchical structure and information interaction relationship of the multi-scale coupling model used in this invention. Figure 4 This is a schematic diagram of the workflow for identifying key driving factors based on causal inference graphs in this invention. Figure 5 A schematic diagram of the virtual farmland mirror interface for multi-scenario stress testing of the digital twin platform of this invention; Figure 6 This is a schematic diagram illustrating the operational status of the actuator of this invention dynamically adjusting the field planting layout. Detailed Implementation

[0025] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0026] like Figure 1 - Figure 6 As shown in the embodiment of the present invention, a method and system for reducing greenhouse gas emissions through intercropping system design includes deploying a distributed multimodal sensor network in a target farmland area. This network consists of soil temperature and humidity sensors, soil redox potential probes, atmospheric carbon dioxide and methane concentration detection units, canopy photosynthetically active radiation receivers, rainfall event triggering recorders, and crop stem flow monitoring devices. Each sensor node uploads the collected data to an edge computing server via the LoRaWAN low-power wide-area communication protocol. The soil temperature and humidity sensors and redox potential probes are embedded at soil depths of 0–10 cm, 10–30 cm, and 30–60 cm, respectively, to obtain vertical profile information on moisture and redox status. The crop stem flow monitoring device, using the thermal pulse method, is installed at 5 cm above the ground at the base of the main crop (such as corn or rice) to continuously record daily water absorption dynamics. All sensors are set to a sampling frequency of once every 10 minutes, and the data is transmitted after AES-128 hardware encryption to ensure the security and integrity of the original sensing information.

[0027] After receiving the aforementioned multi-source sensor data, the edge computing server uses it as input to drive the multi-scale coupled model. This model comprises three levels: the first level is the crop interaction sub-model, the second level is the soil microenvironment sub-model, and the third level is the greenhouse gas generation-transport sub-model. The crop interaction sub-model employs an individual-based modeling approach (IBM), treating each crop as an intelligent agent with independent physiological and behavioral rules. The canopy geometry of each crop is generated by the L-system algorithm, with leaf tilt angle, leaf area index, and spatial distribution parameters preset based on variety characteristics and dynamically updated according to growth stages. Light capture capacity is calculated using a ray tracing algorithm, considering the shading effect caused by neighboring crops; root expansion direction is guided by both local nutrient gradients and root secretion signals. The nutrient gradient is fed back in real-time by the soil microenvironment sub-model, and the root secretion signal intensity is related to crop type and stress state, expressed as:

[0028] Current growth rate (unit: g / (plant·d)) This is the maximum biomass of the species under standard conditions (unit: g / plant).

[0029] A soil microenvironment sub-model describes soil physicochemical changes induced by crop interactions. This model incorporates enzyme kinetic equations to characterize the organic matter decomposition process.

[0030]

[0031] Deutsche constant (unit: mg / kg)

[0032] The rates of nitration and denitrification are described using Monod-type reaction functions, the expressions of which are:

[0033]

[0034] These are the half-saturation constants (units are mgN / kg, mgN / kg, mg / L, and mgC / kg, respectively). Oxygen inhibition constant (unit: mg / L).

[0035] The greenhouse gas generation-transport submodel is built on Fick's diffusion law, while introducing a time lag term to reflect the delayed effect of gas release after rainfall events.

[0036]

[0037]

[0038] (Usually set to 72h).

[0039] After the multi-scale coupled model is run, the system enters the key driving factor identification stage. This stage is achieved by constructing a causal inference graph, which expresses the conditional probability dependencies between variables in the form of a Bayesian network. The network nodes include 28 variables, such as crop row spacing ratio, bandwidth configuration, soil DOC concentration, redox potential (Eh), rainfall intensity, and canopy transmittance. The edge weights are obtained through training with historical data, and the conditional probability distribution is estimated using Gaussian process regression. When new sensor data streams arrive, the system updates the network posterior distribution using a variational inference algorithm and calls the Shapley value decomposition algorithm to quantify the marginal contribution of each configuration parameter to the current greenhouse gas emission flux. for:

[0040] If the duration is greater than 48 hours, it is determined to be a key driving factor, and its window of action is recorded.

[0041] Based on the identification results, the system initiates a multi-objective constrained optimization algorithm to generate the optimal intercropping structure parameter set. This parameter set includes five dimensions: crop combination (e.g., corn-soybean, rice-milk clover), row spacing ratio (the ratio of row spacing between the main crop and the intercropping crop, ranging from 1:1 to 3:1), sowing density (number of main crop plants per acre, ranging from 3000 to 6000 plants), bandwidth configuration (single strip width, ranging from 0.8 to 2.5 m), and rotation sequence (the number of days the intercropping crop is delayed relative to the main crop, ranging from 0 to 15 days). The optimization objective function is:

[0042] 28 and 265 are the conversion factors for the 100-year global warming potential (GWP) recommended by IPCCAR6. Carbon sequestration caused by returning crop residues to the field (unit: kgCO2-eq / ha). (unit: kg / ha) The minimum yield guarantee threshold was set at 90% of the local conventional single-crop yield. The optimization algorithm adopted the NSGA-III multi-objective evolutionary algorithm, with a population size of 200, 100 generations, and Pareto front screening to retain the top 10% of non-dominated solutions.

[0043] All candidate solutions must undergo virtual validation on a digital twin platform. This platform integrates 10m resolution DEM topographic data, 30-year meteorological sequences (interpolated from China Meteorological Administration stations), and FAO soil texture classification maps to construct a virtual farmland mirror covering a county scale. Users can set up multi-scenario stress tests on the platform, including: 30 consecutive days without rainfall (simulating extreme drought), 24-hour rainfall ≥150mm (simulating heavy rainfall impact), and non-steady-state disturbance scenarios such as an increase in pest and disease infection rate to 30%. A solution will only be validated if it meets the requirements under all test scenarios. Only then is it marked as "executable".

[0044] A feasible solution is confirmed and sent to the executing mechanism via a unified command bus. The executing mechanism includes a programmable seeding robot, a variable-rate fertilizer controller, and a remote irrigation scheduling module. The programmable seeding robot is equipped with an RTK-GNSS positioning system (positioning accuracy ±2cm) and completes precise seeding within 48 hours after land preparation, based on the optimized row spacing ratio and bandwidth configuration. The variable-rate fertilizer controller is equipped with a dual-channel feeding system: the first channel delivers urea (46% nitrogen content), and the second channel delivers slow-release composite granules containing 3,4-dimethylpyrazole phosphate (DMPP) (1.5% DMPP content). The mixing ratio of the two is controlled in real-time by the soil nitrate nitrogen concentration, and the control logic is as follows:

[0045] The mass fraction of total nitrogen application. The remote irrigation scheduling module adopts a pulse drip irrigation strategy, with a single irrigation duration controlled at 30–60 minutes, and the irrigation interval based on the average moisture content of the 0–30 cm soil layer.

[0046]

[0047] Meanwhile, the system monitors soil electrical conductivity (EC). When EC > 1.5 dS / m, it automatically extends the irrigation interval and increases the amount of leaching water to prevent salt accumulation.

[0048] The system also integrates a greenhouse gas emission early warning module. This module calculates the measured emission flux every 6 hours. If two consecutive exceedances occur, it is considered an abnormally rising trend, and the emergency response procedure will be immediately activated. Emergency measures include: deploying temporary shade nets (to reduce canopy temperature by 2–4°C and decrease transpiration-induced soil wetting and drying), injecting pure oxygen into the 0–20 cm soil layer through microporous aeration pipes (flow rate 0.5 L / min·m², continuous for 2 hours), or using drones to perform targeted spraying of microbial inhibitors. The drones are equipped with lidar and multispectral cameras, and use emission hotspot maps... The flight trajectory is planned, the spray system uses piezoelectric nozzles, the droplet diameter is controlled between 80 and 120 μm, and the wind speed compensation algorithm ensures that the agent sedimentation deviation is less than 0.5m.

[0049] The central decision engine runs within the Intel SGX security chip, and all model inference is performed in a Trusted Execution Environment (TEE). The system supports offline mode: when communication is interrupted for more than 30 minutes, it automatically switches to the locally cached "historical model library," which contains the 10 best configuration schemes and their corresponding environmental response characteristics from the past three reproductive cycles. The system then selects the most suitable scheme for execution of basic control through nearest neighbor matching.

[0050] The multi-objective constraint optimization algorithm incorporates a knowledge-guided mechanism. This mechanism extracts effective pairing rules from 127 intercropping agronomy articles indexed in Web of Science and the Ministry of Agriculture and Rural Affairs' field experiment database (containing 86 sites and 212 treatment combinations), such as "when intercropping maize with legumes, the row spacing ratio should be 2:1" and "when intercropping rice with green manure, the green manure sowing should be no later than 10 days after rice transplanting," and encodes these as soft constraints.

[0051] (Determined by expert scoring). This probability term, as a multiplicative factor in the fitness function, guides the search towards biologically plausible regions.

[0052] The system periodically receives anonymous performance feedback from other deployment sites. Feedback data includes metrics such as actual emission equivalents, production deviations, and execution success rates. The data is fed to the federated learning server. Local model parameters are updated weekly, and the FedAvg algorithm is used to aggregate global gradients, ensuring that the model's cross-regional generalization ability continues to improve without exposing the original farmland data.

[0053] The technical effects of the present invention are verified through three embodiments and three comparative examples below.

[0054] Example 1: Jingtai County, Baiyin City, Gansu Province (Yellow River Irrigation Area, dominated by irrigated agriculture) Regional background: Located in the upper reaches of the Yellow River, it has a temperate arid climate with an average annual precipitation of 180 mm. The soil type is irrigated desert soil. It is the main corn producing area in Gansu Province. Irrigation relies on the Yellow River diversion project. Traditional planting methods have the risks of low nitrogen fertilizer utilization and secondary soil salinization.

[0055] Crop combination: The main crop is corn (Zhengdan 958, the local main variety), and the intercropping crop is pea (Longwan 10, a nitrogen-fixing legume, suitable for irrigation area ecology).

[0056] Optimize parameters: Row spacing ratio: 2:1 (50cm for corn, 25cm for peas), bandwidth 2.0m; Sowing sequence: synchronous sowing (mid-April, the spring sowing window in Gansu irrigation area); Sowing density: 4,500 corn plants / mu, 100,000 pea plants / ha; Variable fertilization: The mixing ratio of DMPP and urea is dynamically adjusted according to soil nitrate nitrogen (DMPP accounts for 5% when nitrate nitrogen is <10mg / kg, 5%-15% when nitrate nitrogen is 10-30mg / kg, and 20% when nitrate nitrogen is >30mg / kg). Intelligent irrigation: pulse drip irrigation (50 minutes per irrigation, 6-day interval; when the average moisture content of the 0-30cm soil layer is <0.18, the interval is shortened to 4 days; when it is >0.35, the interval is extended to 8 days).

[0057] Implementation results: Greenhouse gas emissions: CH4 emissions < 3 kg / ha (soil aeration is good in arid irrigated areas, and methanogenic bacteria activity is low), N2O emissions 2.8 kg / ha (variable fertilization reduces nitrogen surplus), carbon sink 55 kg CO2-eq / ha (pea stubble returned to the field). Emission equivalent: (3×28 + 2.8×265 - 55 = 763) kgCO2-eq / ha; Yield: Corn 12.0 t / ha (local average monoculture yield 13.2 t / ha, meeting 90% yield threshold), peas 2.5 t / ha (intercropped pea fresh pod yield).

[0058] Example 2: Ganzhou District, Zhangye City, Gansu Province (core irrigation area of ​​Hexi Corridor, arid climate) Regional background: The average annual precipitation is 130 mm, with abundant sunshine (more than 3,000 hours of sunshine per year). The soil is gray-brown desert soil. It is a national corn seed production base. The crop relies on drip irrigation for water conservation. Traditional monoculture has led to significant nitrogen loss and N2O emissions.

[0059] Crop combination: The main crop is corn (Xianyu 335), and the intercropping crop is arrowhead pea (a common green manure variety in local dryland / irrigated areas).

[0060] Optimize parameters: Row spacing ratio: 3:1 (corn 60cm, arrowhead pea 20cm), bandwidth 2.5m; Sowing time: Arrowhead peas should be sown 3 days later than corn (to avoid competition during the seedling stage); Sowing density: 5,000 corn plants / mu, 150,000 arrowhead pea plants / ha; Key driving factors: The system identified "soil soluble organic carbon (DOC) concentration" as the key driving factor, with the window of action being the maize V5-R2 stage (jointing to silking stage).

[0061] Implementation results: Greenhouse gas emissions: CH4 emissions < 2 kg / ha (extreme drought, no anaerobic environment), N2O emissions 2.5 kg / ha (nitrogen fixation by arrowhead peas reduces nitrogen fertilizer input), carbon sink 60 kg CO2-eq / ha; Emission equivalent: (2×28 + 2.5×265 - 60 = 641) kgCO2-eq / ha; Yield: 13.5 t / ha for corn (higher than the local average of 12.8 t / ha for monoculture), and 3.2 t / ha for fresh arrowhead pea grass (which can be used as fodder or green manure).

[0062] Example 3: Wudu District, Longnan City, Gansu Province (Southern Terraced Fields Area, Humid Climate) Regional background: It has a subtropical humid climate with an average annual rainfall of 800 mm, with rainfall concentrated in July to September. It has many mountain terraces (soil is yellow soil), and the traditional cropping is wheat monoculture. Heavy rainfall can easily lead to soil waterlogging and a surge in methane emissions.

[0063] Crop combination: The main crop is wheat (Lantian 29, a winter wheat variety), and the intercropping crop is sesbania (a local green manure variety with strong waterlogging resistance).

[0064] Optimize parameters: Digital twin simulation: Simulates an extreme scenario of "180mm of rainfall in 72 hours", optimizes bandwidth to 1.5m, and uses high ridges (ridge height 30cm, suitable for terraced drainage). Irrigation and emergency response: Irrigation should be suspended before rainfall. On the second day after rainfall, microporous aeration pipes should be activated to increase oxygen (0-20cm soil layer, flow rate 0.6L / min・m²). Sowing density: 300 kg / ha for wheat, 3.0 kg / ha for sesbania, with a row spacing ratio of 1:1 (25 cm for wheat, 25 cm for sesbania).

[0065] Implementation results: Greenhouse gas emissions: CH4 emissions 220 kg / ha (reduced by 38% compared to traditional farming due to less waterlogging in terraced fields), N2O emissions 2.4 kg / ha, carbon sequestration 38 kg CO2-eq / ha; Emission equivalent: (220×28 + 2.4×265 - 38 = 6718) kgCO2-eq / ha; Yield: Wheat 7.8 t / ha (local average of 8.2 t / ha for monoculture, yield loss of only 4.9%), fresh sesame grass yield 3.0 t / ha.

[0066] Comparative Example 1: Same area as Example 1 (Jingtai Yellow River Irrigation Area, Baiyin County) Planting pattern: Traditional corn-pea intercropping (without intelligent control), row spacing 1:1 (corn 40cm, pea 40cm), synchronous sowing, no variable fertilization (single application of urea, nitrogen application rate 240kg / ha), conventional drip irrigation (single irrigation for 90 minutes, 10-day interval).

[0067] result: Greenhouse gas emissions: CH4 emissions < 4 kg / ha, N2O emissions 4.2 kg / ha, carbon sink 32 kg CO2-eq / ha; Emission equivalent: (4×28 + 4.2×265 - 32 = 1143) kgCO2-eq / ha; Yield: 11.2 t / ha for corn and 2.0 t / ha for peas (a 20% reduction compared to Example 1).

[0068] Comparative Example 2: Same region as Example 2 (Zhangye Ganzhou Hexi Corridor Irrigation Area) Planting pattern: Corn monoculture (without interruption), density 6000 plants / mu, conventional fertilization (nitrogen application rate 270kg / ha, no nitrification inhibitor), ordinary flood irrigation (6 irrigations during the growing season, with an average water volume of 80m³ per mu each time).

[0069] result: Greenhouse gas emissions: CH4 emissions < 3 kg / ha, N2O emissions 5.0 kg / ha, carbon sink 25 kg CO2-eq / ha; Emission equivalent: (3×28 + 5.0×265 - 25 = 1344) kgCO2-eq / ha; Yield: 12.8 t / ha of corn (a 5.1% decrease compared to Example 2).

[0070] Comparative Example 3: Same region as Example 3 (Wudu Terraced Fields Area, Longnan) Planting pattern: wheat monoculture (without interruption), conventional sowing (density 280 kg / ha), no drainage or aeration measures, and manual application of nitrogen fertilizer (nitrogen application rate 180 kg / ha).

[0071] result: Greenhouse gas emissions: CH4 emissions after heavy rainfall are 350 kg / ha (due to anaerobic conditions caused by water accumulation), N2O emissions are 2.8 kg / ha, and carbon sequestration is 29 kg CO2-eq / ha; Emission equivalent: (350×28 + 2.8×265 - 29 = 10413) kgCO2-eq / ha; Yield: Wheat 6.5 t / ha (16.7% reduction compared to Example 3, due to waterlogging).

[0072] The above data is summarized in the table below:

[0073]

[0074] Data shows that this invention significantly reduces greenhouse gas emission equivalents in different ecological zones (an average reduction of 38.7% compared to the comparative example), while maintaining or increasing crop yields. Especially under environmental disturbances (such as heavy rainfall), the system effectively suppresses emission pulses through dynamic regulation, demonstrating its robustness.

[0075] In summary, this invention, through a closed-loop architecture of distributed sensing, multi-scale modeling, causal inference, digital twin verification, and intelligent execution, realizes the transformation of intercropping systems from experience-based allocation to mechanism-driven intelligent regulation, providing a quantifiable, reproducible, and scalable technical path for reducing greenhouse gas emissions from agricultural sources.

[0076] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A system for reducing greenhouse gas emissions through intercropping system design, characterized by, include: Distributed multimodal sensor networks, edge computing servers, central decision engines, digital twin platforms, and actuators; The distributed multimodal sensor network is used to collect data on soil temperature and humidity, soil redox potential, atmospheric carbon dioxide and methane concentrations, canopy photosynthetically active radiation, rainfall events, and crop stem flow in the target farmland area, and encrypts and uploads the data to the edge computing server. The edge computing server is configured with a multi-scale coupling model, which includes a crop interaction sub-model, a soil microenvironment sub-model, and a greenhouse gas generation-transport sub-model, and is used to dynamically simulate greenhouse gas emission fluxes under different intercropping configurations based on sensor data. The central decision engine is used to identify key driving factors and their window of action based on simulation results, and to generate the optimal intercropping structure parameter set that meets the minimum yield guarantee threshold and minimizes the cumulative greenhouse gas emission equivalent throughout the entire growth period through a multi-objective constraint optimization algorithm. The digital twin platform is used to perform multi-scenario virtual verification of the optimal intercropping structure parameter set, and is only authorized to execute when the simulated emission curve is lower than the historical baseline and there is no significant risk of yield loss. The actuator is used to receive a verified set of parameters and implement dynamic adjustments or preset intervention strategies for field planting layout to suppress greenhouse gas emission peaks.

2. The system for reducing greenhouse gas emissions through intercropping design according to claim 1, characterized in that, The soil temperature and humidity sensor and soil redox potential probe in the distributed multimodal sensor network are embedded in three soil depths of 0–10 cm, 10–30 cm, and 30–60 cm to obtain vertical profile information; the crop stem flow monitoring device is installed at the base of the main crop plant and continuously records the dynamics of water absorption using the thermal pulse method; each sensor node uploads data with AES-128 hardware encryption via a low-power wide-area communication protocol.

3. A system for reducing greenhouse gas emissions through intercropping design according to claim 1, characterized in that, The crop interaction sub-model employs an individual-based modeling approach, treating each crop as an intelligent agent. Its light-harvesting capacity is dynamically updated based on the canopy's three-dimensional geometry and neighboring shading relationships. Root expansion direction is guided by both local nutrient gradients and secretion signals from neighboring roots. The intensity of the root secretion signal is determined by the formula... For the current biomass, This represents the maximum biomass.

4. A system for reducing greenhouse gas emissions through intercropping design according to claim 1, characterized in that, The soil microenvironment sub-model introduces enzyme kinetic equations to describe the cellulose degradation and protein mineralization processes, and combines Monod-type reaction functions to characterize the dependence of nitrification and denitrification rates on dissolved oxygen concentration, ammonium nitrogen content, and soluble organic carbon levels. The greenhouse gas generation-transport sub-model is constructed based on Fick's diffusion law and introduces an exponentially decaying hysteresis weighted kernel function to reflect the time delay characteristics of gas release after rainfall events.

5. A system for reducing greenhouse gas emissions through intercropping design according to claim 1, characterized in that, The key driver identification mechanism is implemented through a causal inference graph in the form of a Bayesian network. This graph includes environmental disturbance events, crop configuration characteristics, and soil biochemical indicators as nodes. When new sensor data is received, the system updates the network posterior distribution and quantifies the marginal contribution of each configuration parameter to the current emission flux according to the Shapley value decomposition algorithm. When the Shapley value of a parameter exceeds 0.15 and the duration is greater than 48 hours, it is determined to be a key driver and its window of action is recorded.

6. A system for reducing greenhouse gas emissions through intercropping design according to claim 1, characterized in that, The optimal intercropping structure parameter set includes five dimensions: crop type combination, row spacing ratio, sowing density, bandwidth configuration, and rotation sequence. The multi-objective constrained optimization algorithm adopts the NSGA-III evolutionary algorithm to minimize emission equivalents while ensuring that the simulated yield is not less than 90% of the local conventional monoculture yield. 28 and 265 are the global warming potential conversion factors recommended by IPCCAR6.

7. A system for reducing greenhouse gas emissions through intercropping design according to claim 1, characterized in that, The actuators include a programmable seeding robot, a variable fertilizer controller, and a remote irrigation scheduling module. The programmable seeding robot performs precise seeding based on an RTK-GNSS positioning system with optimized row spacing and bandwidth configuration. The variable fertilizer controller is equipped with a dual-channel feeding system, one channel delivering conventional nitrogen fertilizer and the other channel delivering slow-release granules containing the nitrification inhibitor DMPP, and dynamically adjusting the mixing ratio according to the soil nitrate nitrogen concentration. The remote irrigation scheduling module adopts a pulse drip irrigation strategy, with the duration of a single irrigation controlled at 30–60 minutes, and dynamically adjusting the irrigation interval based on the average moisture content of the 0–30cm soil layer.

8. A system for reducing greenhouse gas emissions through intercropping design according to claim 1, characterized in that, It also includes a greenhouse gas emission early warning module, which continuously compares the relative deviation between the measured emission flux and the model prediction. When the deviation exceeds 30% twice in a row, an emergency response procedure is initiated, triggering the deployment of temporary shade nets, local oxygenation injection through microporous aeration pipes, or the spraying of microbial inhibitors by drones at designated points. The drones plan their flight trajectories based on the emission hotspot map generated by the mobile flux vehicle and use lidar to correct the spray height and wind speed compensation parameters in real time.

9. A system for reducing greenhouse gas emissions through intercropping design according to claim 1, characterized in that, The central decision engine runs within a secure chip with a trusted execution environment. All sensor data is hardware-encrypted before transmission, and model inference is completed in an isolated sandbox. The system supports offline mode, and during communication interruptions, it completes basic judgments and control outputs based on a locally cached historical pattern library. The digital twin platform integrates high-resolution terrain, long-term meteorological sequences, and soil texture classification information, and supports stress testing in non-steady-state disturbance scenarios such as extreme drought, heavy rainfall impacts, and pest and disease outbreaks.

10. A method for reducing greenhouse gas emissions through intercropping system design, applicable to the system for reducing greenhouse gas emissions through intercropping system design as described in any one of claims 1-9, characterized in that, Includes the following steps: Deploy a distributed multimodal sensor network and collect multi-source environmental and crop status data; A multi-scale coupled model was used to dynamically simulate greenhouse gas emission fluxes under different intercropping configurations. Identify key driving factors and their sensitive action windows through causal inference mapping; A multi-objective constrained optimization algorithm is used to generate the optimal set of intercropping structure parameters that balances emission reduction efficiency and production stability; Multi-scenario virtual verification of candidate solutions is performed in a digital twin platform; The verified solutions will be distributed to the implementing agencies for precision sowing, variable fertilization, and intelligent irrigation. The emission early warning module operates synchronously, and emergency intervention is initiated for abnormally rising trends, thereby achieving closed-loop control of greenhouse gas emissions.

Images