A method for dynamically reconstructing a paddy field greenhouse gas emission stage structure

Abstract

This invention discloses a method for dynamic reconstruction of the stage structure of greenhouse gas emissions in paddy fields. The method acquires multi-source time-series data, including remote sensing observations, water processes, and meteorological drivers, to reconstruct the crop growth trajectory of the target paddy field. It constructs a set of intermediate channel state variables, including substrate supply potential, plant transport potential, oxidation inhibition potential, and wet-dry transition pulses. The method identifies the boundaries of dynamic stages of greenhouse gas emissions and adaptively switches between multiple candidate kinetic sub-models to achieve continuous reconstruction of the stage structures of methane, nitrous oxide, and carbon dioxide emissions. It outputs continuous flux curves, stage contribution rates, peak times, event pulse contributions, and field-scale spatial distribution results. This invention elevates the assessment of greenhouse gas emissions in paddy fields from total emission estimation to dynamic reconstruction of stage structures, and is applicable to greenhouse gas emission monitoring, stage structure analysis, field-scale risk identification, and refined water management in paddy fields.

Description

Technical Field

[0001] This invention belongs to the field of agricultural remote sensing, farmland ecological process simulation and greenhouse gas emission monitoring technology. Specifically, it relates to a method for dynamic reconstruction of the stage structure of greenhouse gas emissions in paddy fields, and more particularly to a method for dynamic identification, structural switching and spatial mapping of continuous processes and stage contribution characteristics of greenhouse gas emissions in paddy fields at the field scale using multi-source time-series observation data. Background Technology

[0002] Greenhouse gas emissions from paddy fields are influenced by multiple factors, including crop growth, water management, meteorological conditions, and soil environment, exhibiting significant time-varying, phased, and event-pulsating characteristics. Existing technologies primarily target cumulative emissions, average fluxes, or discrete monitoring values. While these can characterize emission intensity to some extent, they struggle to further identify phase boundaries, peak times, and phase contribution rates under different growth stages and wet-dry transition events.

[0003] Existing methods generally suffer from two shortcomings. First, crops are often treated as background factors, lacking explicit expression of the mediating role of crops in the formation, transport, and suppression of greenhouse gases. Second, most methods use a single fixed model or a uniform parameter system, making it difficult to characterize the structural dynamic differences under different ecological process states such as continuous flooding, desiccation response, rehydration pulse, and growth decline.

[0004] Meanwhile, the development of synthetic aperture radar, optical remote sensing, UAV remote sensing, and continuous field moisture observation has enabled the continuous acquisition of key variables such as paddy field flooding status, wet-dry transition processes, leaf area index, and growth progress. Therefore, it is necessary to propose a new technical method to upgrade greenhouse gas emissions from paddy fields from traditional total estimation to dynamic reconstruction of stage structures. This method, through joint modeling of crop growth trajectories and intermediate channel state variables, can achieve continuous identification and representation of emission contributions at different stages, peak times, event impulses, and their spatial distribution at the field scale. Summary of the Invention

[0005] The purpose of this invention is to provide a method for dynamic reconstruction of the stage structure of greenhouse gas emissions in paddy fields, in order to solve the problems in the prior art that make it difficult to continuously identify the boundaries, peak times and stage contribution rates of greenhouse gas emissions in paddy fields, difficult to express the mediating role of crops in the formation, transport and suppression of greenhouse gases, difficult to characterize the structural dynamic differences under different growth stages and the dry-and-re-water event, and difficult to achieve field-scale spatial mapping and risk feedback.

[0006] To address the above problems, this invention provides a method for dynamic structural reconfiguration of greenhouse gas emission stages in paddy fields, comprising the following steps:

[0007] S1: Acquire multi-source time-series data of the target paddy field within the observation period and preprocess it under a unified spatiotemporal reference. The multi-source time-series data includes at least time-series remote sensing observation data for characterizing the flooding state and wet-dry transition process of the paddy field, water process data for characterizing the field water change process, and meteorological data for characterizing environmental driving conditions.

[0008] S2: Based on the multi-source time-series data preprocessed in step S1, construct crop growth trajectory variable groups and water process variable groups for the target paddy field. The crop growth trajectory variable group includes at least two or more continuous time series from leaf area index, aboveground biomass, plant height, and growth progress indicators, and extracts one or more of their rate of change, inflection point location, and stage transition characteristics. The water process variable group includes at least one or more of the following: field surface water level, soil moisture content, flooding state, continuous flooding duration, continuous dry-out duration, flooding state change rate, dry-out event indicator, and re-watering event indicator.

[0009] S3: Based on the crop growth trajectory variable group and the water process variable group, construct a set of intermediate channel state variables to characterize the formation, transport and inhibition of greenhouse gases in paddy fields. The intermediate channel state variable group includes at least the substrate supply potential index, plant transport potential index, oxidation inhibition potential index and dry-wet transition pulse index.

[0010] S4: Based on the dynamic change characteristics of the crop growth trajectory, the change characteristics of the intermediate channel state variable group, and the field dry-wet transition events, identify the dynamic stage boundaries of greenhouse gas emissions and form a stage sequence that changes over time.

[0011] S5: Construct multiple candidate dynamic sub-models for different stages and event states, including at least a stage sub-model for the continuous flooding growth stage, an event sub-model for the desiccation response stage, an event sub-model for the rehydration pulse stage, and a stage sub-model for the growth decline stage; based on the stage sequence obtained in step S4 and the joint state vector at the current moment, calculate the matching degree of each candidate dynamic sub-model to the current stage or event state, and determine the corresponding switching weight accordingly; wherein, the joint state vector includes at least one or more of crop growth trajectory variables, intermediate channel state variables, water process variables, and meteorological driving variables, and the switching weight is determined by threshold discrimination, probability discrimination, gating function, state machine rules, or a combination thereof; according to the switching weight, perform unique calling or weighted fusion output on the corresponding candidate dynamic sub-model, complete the adaptive structural switching between each candidate dynamic sub-model on the time series, and establish a dynamic reconstruction model for greenhouse gas emissions;

[0012] S6: Using the greenhouse gas emission dynamic reconstruction model, the emission process of at least one greenhouse gas in the target paddy field is continuously reconstructed to obtain the corresponding stage structure reconstruction results. The stage structure reconstruction results include at least two or more of the following: continuous flux curve, contribution rate of each stage, peak occurrence time and event pulse contribution.

[0013] S7: Based on one or more of the following: a small amount of ground-based greenhouse gas flux observation data, historical sample database data, or local UAV observation data, perform deviation correction, parameter constraint, or migration correction on the greenhouse gas emission dynamic reconstruction model established in step S5 and the stage structure reconstruction results obtained in step S6 to obtain the corrected stage structure reconstruction results.

[0014] S8: Map the corrected stage structure reconstruction results obtained in step S7 to field-scale grid cells to generate the spatial distribution results of greenhouse gas emission stage structure of the target paddy field, and output the corresponding stage risk warning information and water management feedback information.

[0015] Preferably, in step S1, the time-series remote sensing observation data includes Sentinel-1 synthetic aperture radar remote sensing data, Sentinel-2 optical remote sensing data, and UAV remote sensing data; the moisture process data includes field surface water level data, soil moisture content data, and flooding state sequence data; and the meteorological data includes air temperature, precipitation, radiation, humidity, and wind speed. This scheme can characterize the greenhouse gas emission process of paddy fields from three levels: field surface moisture state, crop canopy state, and environmental driving conditions, improving the completeness of input information. It is particularly suitable for observation in rice-growing areas under cloudy and rainy conditions, and can enhance the continuous identification capability of field flooding state and crop growth state.

[0016] Preferably, in step S2, the construction of the crop growth trajectory variable set is achieved through a combination of remote sensing inversion, temporal smoothing, and growth process constraints. The crop growth trajectory variable set further includes one or more dynamic features among trajectory inflection points, growth rates, accelerations, peak times, and stage inflection points. The water process variable set includes at least one or more of the following: field surface water level, soil volumetric water content, flooding index, duration of continuous flooding, duration of continuous drying, rate of change of flooding state, drying event indicator, and rewatering event indicator. This scheme integrates crop growth status and field water processes into a unified basic state variable system, providing parallel and complementary inputs for subsequent intermediate channel construction and stage identification. Compared to using only a single crop variable or a single water variable, this scheme is more conducive to capturing the coupling characteristics of biological and water-driven processes in the emission process.

[0017] Preferably, in step S3, the intermediate channel state variable set is not directly given by a single greenhouse gas flux observation, but is constructed from crop growth trajectory variables, water process variables, and their derived features. Specifically, the substrate supply potential index is characterized by one or more of leaf area index, aboveground biomass, biomass growth rate, and growth progress indicators; the plant transport potential index is characterized by one or more of leaf area index, plant height, and growth progress indicators; the oxidation inhibition potential index is characterized by one or more of flooding state, flooding duration, field surface water level, or soil moisture content; and the dry-wet transition pulse index is characterized by one or more of flooding state change rate, desiccation event indicator, and rehydration event indicator. The intermediate channel state variable set is obtained through normalization calculation, weighted combination, rule mapping, or a combination thereof. This scheme can explicitly express the mediating role of crops in the formation, transport, and suppression of greenhouse gases, improving the emission process from directly fitting fluxes to reconstructing fluxes based on channel states. This scheme helps improve the model's mechanistic interpretability and cross-stage adaptability.

[0018] Preferably, in step S4, the dynamic stage boundary is not divided according to a fixed calendar period, but is jointly identified based on at least two of the following: crop growth trajectory turning point information, water state change information, and intermediate channel state variable mutation information. This scheme can avoid the obscuring of the dynamic changes in the actual emission process by the fixed growth period division method, and is more conducive to identifying stage mutations triggered by key events such as drying and rehydration. This can improve the sensitivity of stage boundary identification and the authenticity of stage contribution analysis.

[0019] Preferably, in S5, the adaptive structure switching is a model structure-level switching, rather than parameter updates within a single fixed model; the candidate kinetic sub-models include at least two or more of the following: stage sub-models for the continuous flooding growth phase, event sub-models for the drying response phase, event sub-models for the re-flooding pulse phase, and stage sub-models for the growth-and-decline phase. This scheme can invoke different kinetic structures for different growth stages and different wet-dry transition events, characterizing the differences in emission mechanisms at the model structure level. Compared to single-model parameter tuning, this scheme is more suitable for expressing complex emission processes where stages and events coexist.

[0020] Preferably, in step S5, the joint state vector includes at least one or more of the following: crop growth trajectory variables, intermediate channel state variables, water process variables, and meteorological driving variables. The switching weight is used to characterize the degree of matching between each candidate kinetic sub-model and the current stage or event state, and is determined using threshold discrimination, probability discrimination, gating function, state machine rules, or a combination thereof. When the switching weight of a candidate kinetic sub-model is greater than a preset threshold, the candidate kinetic sub-model is invoked to output the reconstruction result, or a weighted fusion output is performed according to the switching weights of each candidate kinetic sub-model. This scheme introduces the switching weight into the model structure selection process, making the switching between candidate kinetic sub-models have a calculable, interpretable, and scalable basis. By having both unique invocation and weighted fusion output methods coexisting, the stability and flexibility of model switching can be balanced.

[0021] Preferably, in step S6, the stage structure reconstruction results target at least one or more of the formation of methane, nitrous oxide, and carbon dioxide; wherein, the contribution rate of each stage is calculated as the proportion of the reconstructed flux of each stage in the cumulative flux of the entire observation period, and the event pulse contribution is calculated as the incremental flux within a preset time window before and after desiccation and rehydration. This scheme allows the output results to go beyond cumulative emissions or average fluxes, and further reveal the structural differences between different stages, different events, and different greenhouse gases. This enhances the supporting capabilities of the results for emission reduction diagnosis, mechanism analysis, and time-series management.

[0022] Preferably, in step S7, model calibration uses one or more of the following as constraint information: a small amount of ground-based greenhouse gas flux observation data, historical sample database data, or local UAV observation data. This constraint information is applied to the greenhouse gas emission dynamic reconstruction model established in step S5 and the stage structure reconstruction results obtained in step S6, performing one or more of the following processing: linear deviation correction, staged residual correction, parameter constraints, posterior shrinkage, or migration correction, to obtain the corrected stage structure reconstruction results. This scheme can perform targeted deviation correction and consistency improvement on the model output without relying on large-scale high-frequency measured flux data. This scheme is particularly suitable for application scenarios with limited field-scale observation samples but requiring high output reliability.

[0023] Preferably, in step S8, the target paddy field is divided into multiple grid units. Steps S2 to S7 are performed on each grid unit to obtain the corrected stage structure reconstruction result at the grid unit level. One or more of the following are generated by spatial stitching: a stage contribution rate distribution map, a peak risk distribution map, and a stage high-risk area distribution map at the field scale. A risk score is constructed based on peak flux, event stage contribution ratio, and intermediate channel state variables, and corresponding stage risk warning information and water management feedback information are output. This scheme can extend the stage structure reconstruction result from the time series level to the spatial distribution level, realizing field-scale risk identification and differentiated management. This not only supports the location of stage high-risk areas but also provides spatial decision-making basis for irrigation regulation, re-watering timing optimization, and emission reduction management.

[0024] Compared with related technologies, the method for dynamic structural reconfiguration of greenhouse gas emission stages in paddy fields provided by this invention has the following beneficial effects:

[0025] (1) This invention no longer targets only cumulative emissions or average flux, but rather focuses on the greenhouse gas emission stage structure as the reconstruction object, which can identify the contribution rate of each growth stage, peak time point and event pulse contribution.

[0026] (2) By constructing intermediate channel state variables such as substrate supply potential, plant transport potential, oxidation inhibition potential and dry-wet transition pulse, this invention couples the crop growth trajectory with the emission process, which can more accurately express the role mechanism of crops in emission formation, transport and inhibition.

[0027] (3) The present invention dynamically calls different candidate dynamic sub-models according to the crop growth stage and the dry-wet transition event to realize the model structure level switching, rather than single model parameter tuning, thereby improving the ability to characterize the stage and event-based emission process.

[0028] (4) This invention unifies remote sensing observation, water processes and crop trajectories under the same spatiotemporal framework, enabling grid unit-level stage structure reconstruction and spatial splicing, and is suitable for continuous expression and spatial mapping of greenhouse gas emissions at the field scale.

[0029] (5) The present invention can output peak risk warning, distribution of high-risk areas in stages and water management feedback information based on the stage structure reconstruction results, providing a basis for refined emission reduction management of paddy fields. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the overall process of the method of the present invention;

[0031] Figure 2 This is a schematic diagram illustrating the integrated construction and dynamic stage identification of crop growth trajectory variable groups, water process variable groups, and intermediate channel state variable groups in this invention.

[0032] Figure 3 This is a schematic diagram illustrating the calling of candidate dynamic sub-models, determination of switching weights, and adaptive structure switching in this invention;

[0033] Figure 4 This is a schematic diagram of the structural reconstruction results during the greenhouse gas emission stage in this invention; (a) continuous reconstruction flux curve of greenhouse gases; (b) contribution rate of greenhouse gases at different stages;

[0034] Figure 5 This is a schematic diagram of the greenhouse gas emission risk diagnosis output in this invention; (a) CH4 peak risk warning map; (b) distribution map of high-risk areas in different stages;

[0035] Figure 6 This is a schematic diagram of the peak time point diagnosis and water management feedback output at the field scale in this invention; (a) Spatial distribution map of CH4 peak time point; (b) Regional map of water management feedback at the field scale. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] 1. Multi-source time series data acquisition and preprocessing

[0038] This embodiment selects a standardized paddy field in the experimental area of ​​Taihe County, Ji'an City, Jiangxi Province as the target object. The field area is 300m×400m and divided into 10m×10m grid units for dynamic structural reconstruction of greenhouse gas emissions at the field scale. This embodiment uses April 15, 2025 as the sowing start date and the observation period covers the entire growing season.

[0039] like Figure 1 As shown, the data used in this embodiment includes Sentinel-1 synthetic aperture radar data, Sentinel-2 optical remote sensing data, local UAV remote sensing data, continuous field water level monitoring data, daily meteorological data, and a small amount of static box flux correction data. Specifically, Sentinel-1 data is used to identify field flooding status, duration of water accumulation, and wet-dry transition events; Sentinel-2 data is used to retrieve leaf area index, canopy growth status, and growth process; UAV data is used for local high-resolution auxiliary calibration; continuous field water level monitoring data is used to characterize water processes; and meteorological data is used to characterize environmental driving conditions such as temperature, precipitation, radiation, humidity, and wind speed.

[0040] The Sentinel-1 data underwent sequential orbit correction, radiometric calibration, terrain correction, and speckle noise suppression. The Sentinel-2 data underwent atmospheric correction, cloud shadow identification and removal, resampling, and temporal registration. The UAV data underwent orthorectified stitching and spatial registration. The field water level and meteorological data underwent outlier removal, missing value imputation, and temporal scale unification. After processing, all data were uniformly mapped to a 10m×10m spatial grid and a daily time step, forming a unified spatiotemporal dataset.

[0041] 2. Construction of crop growth trajectory, water processes, and intermediate channel states

[0042] like Figure 2 As shown, this embodiment first constructs a set of crop growth trajectory variables:

[0043]

[0044] in, For a moment Leaf area index, For a moment Aboveground biomass. For a moment Plant height, For a moment The crop growth trajectory variables are used to characterize the crop growth status of the target field throughout the entire growing season, and further extract leaf area index change rate, biomass growth rate, and trajectory inflection point for subsequent stage identification.

[0045] Furthermore, construct the set of variables for the water process:

[0046]

[0047] in, For a moment The water level on the field surface; For a moment Soil volumetric water content; For a moment The flood index; Deadline The duration of continuous flooding; Deadline The duration of continuous drying; The rate of change of the flooded state; This is an indicator quantity for the event of a dry fall; This is an indicator of the refilling event.

[0048] Based on this, construct the intermediate channel state variable group:

[0049]

[0050] Among them, the substrate supply potential index The plant transport potential index is composed of a weighted average of leaf area index, biomass growth rate, and growth progress indicators. The index is composed of a weighted average of leaf area index, plant height, and growth progress indicators; the oxidative inhibition potential index... Composed of the flooding index and the duration of continuous flooding; wet-dry transition pulse index. It consists of the flooding state change rate, the desiccation event indicator, and the refill event indicator. Preferably, the state variables of each intermediate channel can be obtained through normalization calculation, weighted combination, or rule mapping.

[0051] Subsequently, based on crop trajectory inflection points, changes in water state, and abrupt changes in intermediate channel state variables, the boundaries of dynamic stages for greenhouse gas emissions are identified. Preferably, when there is a significant change in the leaf area index or biomass growth rate, a drying-off event or a rehydration event occurs, or an abrupt change occurs in an intermediate channel state variable, the corresponding time point is marked as a candidate boundary point. After screening and merging the candidate boundary points, a dynamic stage sequence is generated. Preferably, the dynamic stage sequence includes one or more of the following: seedling stage, tillering stage, jointing and booting stage, drying-off event sub-stage, rehydration event sub-stage, heading and grain-filling stage, and maturity stage.

[0052] 3. Construction of candidate dynamic sub-models and adaptive structure switching

[0053] like Figure 3 As shown, after obtaining the dynamic stage sequence, this embodiment constructs multiple candidate dynamic sub-models for different stages and event states, including at least a stage sub-model for the continuous flooding growth stage, an event sub-model for the drying response stage, an event sub-model for the re-watering pulse stage, and a stage sub-model for the growth and decline stage.

[0054] To automatically invoke different sub-models at different times, a joint state vector is constructed:

[0055]

[0056] in, For the crop growth trajectory variable group, For the intermediate channel state variable group, For the water process variable group, This is a set of meteorological driving variables.

[0057] Furthermore, for the first The switching weights of the candidate dynamic sub-models are defined as follows:

[0058]

[0059] in, The number of candidate dynamic sub-models. For the first The gating score function for each candidate dynamic sub-model is determined by the joint state vector and the dynamic stage sequence, and is used to reflect the degree of matching of each candidate dynamic sub-model at the current time step.

[0060] For any greenhouse gas At any moment The reconstructed flux is represented as:

[0061]

[0062] in For the first The candidate dynamic sub-models at time... The output function.

[0063] Therefore, it can be seen that the present invention adopts model structure-level switching rather than parameter updating within a single fixed model, thus enabling a more accurate description of emission dynamics differences under different reproductive stages and different dry-wet transition events.

[0064] 4. Output of Stage Structure Reconstruction Results and Model Correction

[0065] like Figure 4 As shown, after completing the adaptive structure switching of the candidate dynamic sub-model, the preliminary continuous reconstructed flux of the target greenhouse gas at each grid cell and time step can be obtained. To improve the consistency between the stage structure reconstruction results and ground observations, a model calibration process is set up in this embodiment.

[0066] For any greenhouse gas In grid cells and time The initial reconstructed flux is denoted as Its corrected flux can be expressed as:

[0067]

[0068] in, Indicates the corresponding dynamic stage. and Greenhouse gases In the stage The phased correction coefficients and bias terms are as follows.

[0069] The aforementioned correction coefficients can be obtained by inversion from a small number of ground flux observation samples, or by migration from historical sample databases of samples with conditions similar to those of the current experimental area. For cases with a small sample size, posterior shrinkage or residual smoothing can be used to constrain the correction coefficients at different stages to avoid overfitting in local stages.

[0070] After obtaining the corrected continuous reconstructed flux, the cumulative flux, stage contribution rate, peak flux, peak time, and event pulse contribution within different stages are further calculated. For any greenhouse gas... In grid cells No. The cumulative flux within a stage is denoted as:

[0071]

[0072] in, Represents grid cells In the The time set corresponding to the stage For the time step, a daily time step is preferred in this embodiment. Grid cell The total cumulative flux over the entire observation period is:

[0073]

[0074] in, The total number of stages indicates the number of grid cells. In the The contribution rate of the stage is:

[0075]

[0076] For any greenhouse gas , grid cell The peak flux and peak time point are expressed as follows:

[0077]

[0078]

[0079] To characterize the pulse discharge caused by the dewatering-refilling event, the event window can be further defined. Incremental flux within the pulse:

[0080]

[0081] in, This represents the background flux before the event. From this, the impulse contribution rate of the corresponding event can be obtained.

[0082] Furthermore, to generate field-scale statistical results, spatial aggregation can be performed based on the grid cell-level results. For any greenhouse gas... Its field-scale average flux curve can be expressed as:

[0083]

[0084] in, This represents the set of all grid cells within the target field. This indicates the total number of grid cells.

[0085] Correspondingly, the first The stage cumulative flux at the field scale can be expressed as:

[0086]

[0087] The total cumulative flux at the field scale over the entire observation period can be expressed as:

[0088]

[0089] Then the first The contribution rate of a stage at the field scale can be expressed as:

[0090]

[0091] In this embodiment, Figure 4 (a) presents the continuous remodeling flux curves for methane and nitrous oxide. Figure 4 (b) presents statistical results of greenhouse gas contribution rates at different stages. Figure 4 This invention provides a clear visual demonstration of the output effect of the continuous process reconstruction and stage structure analysis. Methane's contribution is mainly concentrated during the tillering and heading / grain-filling stages, with a certain contribution also observed during the jointing and booting stages. Nitrous oxide, on the other hand, exhibits a higher structural contribution in the drying process and later stages, thus reflecting the differentiated emission mechanisms of different greenhouse gases at different stages.

[0092] 5. Risk Diagnosis Output

[0093] like Figure 5 As shown, after obtaining the corrected stage structure reconstruction results, the field-scale risk is further diagnosed and output. In this embodiment, the peak methane risk is used as the main risk diagnosis object, and the peak risk of different grid units is comprehensively scored by combining the contribution ratio of the rehydration event stage and the intermediate channel state variables.

[0094] For grid cells Its methane peak risk score can be expressed as:

[0095]

[0096]

[0097] in, Represents grid cells peak methane flux; This represents the methane pulse contribution of the grid cell within the rehydration event window, calculated from the aforementioned rehydration event pulse incremental flux. Relative to the total cumulative flux of grid cells Normalization yields the result; This represents a comprehensive risk indicator comprised of intermediate channel state variables; , and Let the risk score weights be , and satisfy the following conditions:

[0098]

[0099] Furthermore, the comprehensive risk indicator... It can be represented as:

[0100]

[0101] in, , , and Representing grid cells respectively The aggregated results of substrate supply potential, plant transport potential, oxidative retardation potential, and dry-wet transition pulse index within the risk diagnosis window; , , and For the corresponding weights, and satisfying:

[0102]

[0103] The risk diagnosis window A preset time window can be taken before and after the methane peak occurrence date, or a combination of the rehydration event window and the preset time window before and after the peak occurrence date can be taken. , and The risk diagnosis window contains the corresponding intermediate channel state variables. The normalized time average within the range is determined, that is:

[0104]

[0105]

[0106]

[0107] in, Represents grid cells The corresponding time steps within the risk diagnosis window, , and These represent the normalized substrate supply potential, plant transport potential, and oxidation inhibition potential indices, respectively.

[0108] The The amplification effect of wet-dry transition events on peak risk can be characterized by the weighted aggregate value of the wet-dry transition impulse index within the risk diagnosis window, i.e.:

[0109]

[0110] in, This represents the normalized dry-to-wet transition pulse index. Represents the time weight at the corresponding moment, and satisfies:

[0111]

[0112] Within the dry-out event window and the rehydration event window It can be set to have a higher weight than non-event periods to enhance the overall risk indicator's responsiveness to event impulses.

[0113] The normalization of the intermediate channel state variables can be achieved using range standardization, quantile standardization, or a combination thereof. Taking range standardization as an example, for any intermediate channel state variable... Its normalization result can be expressed as:

[0114]

[0115] in, Represents grid cells At any moment The values ​​of the intermediate channel state variables, and These represent the minimum and maximum values ​​of the corresponding variables in the target field and during the observation period, respectively; Preferred , , and One of them.

[0116] The , , and The value can be determined using methods such as equal-weighting, empirical assignment, or fitting based on historical samples, or based on the normalized results of the correlation between the state variables of each intermediate channel and the peak methane flux. Therefore, the comprehensive risk indicator... It is no longer a single empirical item, but a calculable indicator driven by four types of processes: substrate supply, plant transport, oxidative arrest, and wet-dry transition pulses. This improves the interpretability and transferability of methane peak risk diagnosis.

[0117] After obtaining the methane peak risk score for each grid cell, it can be divided according to the quantile threshold or preset rules. It is divided into four levels: low, attention, alert, and high alert. (Grid unit) The risk level is Then, at the field scale, the first The area proportion of each risk level can be expressed as:

[0118]

[0119] in, This represents the set of all grid cells within the target field. Indicates the total number of grid cells. Indicative functions are represented. By spatially stitching together the risk level results of all grid cells, a... Figure 5 (a) shows the methane peak risk warning diagram.

[0120] Furthermore, to identify which stage dominates the high-risk area, this embodiment compares the stage-cumulative flux of each stage in the high-risk grid cell and defines the grid cell. The high-risk dominant phase is:

[0121]

[0122] in, Represents grid cells Inner methane in The cumulative flux of the stage. The set of stages for high-risk identification is preferred to include the tillering stage, jointing and booting stage, drying event sub-stage, rehydration event sub-stage, and heading and grain-filling stage.

[0123] No. The area proportion of the high-risk-dominated phase can be expressed as:

[0124]

[0125] in, This represents a set of grid cells indicating a risk level that has reached the warning or high-level warning stage. This represents the total number of high-risk grid units. From this, the high-risk grid can be further divided into high-risk zones dominated by the tillering stage, the jointing and booting stage, the drying event, the rehydration event, and the heading and grain-filling stage. By spatially stitching together the dominant stage results of the high-risk grid units, a... Figure 5 (b) shows the distribution map of high-risk areas at each stage.

[0126] 6. Peak Hour Diagnosis and Moisture Management Feedback Output

[0127] like Figure 6 As shown, after completing the risk diagnosis, the differences in peak methane time points at the field scale and the corresponding water management feedback are further output. The methane peak time point diurnal sequence can be expressed as:

[0128]

[0129] in, Represents grid cells The date sequence of the occurrence of internal methane peaks Represents grid cells At any moment The corrected methane flux. The date sequence is numbered with the sowing start date as day 1. Thus, a... Figure 6 (a) shows the spatial distribution of methane peak times, used to describe the differences in the arrival time of methane peaks in different grid units.

[0130] To characterize the constraint effect of nitrous oxide pulses on management feedback during the drying stage, a grid cell is defined. Nitrous oxide pulse sensitivity index during the drying stage for:

[0131]

[0132] in, Represents grid cells The peak increase of nitrous oxide within the drying event window Represents grid cells The cumulative increment of nitrous oxide pulses within the drying event window Represents grid cells Aggregated results of the wet-dry transition pulse index within the drying event window. , and For the corresponding weights, and satisfying:

[0133]

[0134] The event window for the drying event is denoted as The refill event window is recorded as The background reference window before the event is denoted as The peak increment of nitrous oxide It can be represented as:

[0135]

[0136] in, Represents grid cells At any moment Corrected nitrous oxide flux, This represents the median value of nitrous oxide flux within the background reference window.

[0137] The cumulative increment of the nitrous oxide pulse It can be represented as:

[0138]

[0139] in, Represents grid cells The background nitrous oxide flux within the background reference window is preferably selected as follows: Mean or median internal nitrous oxide flux Indicates the time step.

[0140] The Available from the event window The dry-wet transition pulse index after normalization of the internal classics is aggregated to obtain the following:

[0141]

[0142] in, Represents grid cells At any moment Normalized dry-to-wet transition pulse index This indicates the number of time steps within the event window. , and The values ​​can be determined using equal weighting, empirical assignment, historical sample fitting, or based on the normalized results of the correlation between each indicator and the intensity of the nitrous oxide pulse during the drying stage. Therefore, the aforementioned... The computable metrics, driven by peak increment, cumulative increment, and dry-wet transition pulse status, can improve the interpretability and reusability of management feedback rules.

[0143] After completing the peak time point diagnosis, the methane peak risk score is further combined. High-risk-dominated stage And the pulse sensitivity index of nitrous oxide during the drying stage Generate a differentiated water management feedback zoning map. Define the management feedback categories as:

[0144]

[0145] in, This represents the management feedback decision rule function. Represents grid cells Management feedback categories.

[0146] The management feedback rule is set as follows: when When the level is low or of concern, the grid is classified as a regular stable irrigation zone; when When the area is a high-risk zone dominated by reclaimed water events, the grid is divided into a shortened reclaimed deep-water zone; when To be the main high-risk area for the dry incident, or When the exposure exceeds a preset threshold, the grid is divided into zones to avoid overexposure and stagger nitrogen application; when When the area is a high-risk zone dominated by the jointing and booting stage or the heading and grain-filling stage, the grid is divided into a lightly intermittently irrigated zone; when When the tillering stage is the dominant high-risk area, the grid is divided into a shallow water stable irrigation area during the tillering stage.

[0147] Furthermore, at the field scale, the first The area percentage of each management feedback category can be expressed as:

[0148]

[0149] By analyzing all grid cells The results can be spatially stitched together to form Figure 6 (b) shows the field-scale water management feedback zoning map. The preset threshold can be determined using a quantile threshold, an empirical threshold, or based on historical sample statistical results. Preferably, when When the threshold of the upper quartile is exceeded, it can be determined that the grid cell has a high nitrous oxide pulse sensitivity during the drying stage.

[0150] Figure 6 (a) reflects when the peak occurs. Figure 6 (b) reflects the management strategies to be adopted for different grids. Together, these two aspects constitute the application output chain of this invention, from point-in-time diagnosis to management feedback. In summary, based on multi-source time-series observation data, this invention integrates crop growth trajectory, water process characteristics, and intermediate channel state variables to construct a dynamic reconstruction method for the stage structure of greenhouse gas emissions in paddy fields. This method achieves continuous reconstruction, risk diagnosis, and management feedback output of the greenhouse gas emission process, and can be used for monitoring greenhouse gas emissions in paddy fields, analyzing stage structure, identifying field-scale risks, and refining water management.

[0151] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for dynamic structural reconfiguration of greenhouse gas emission stages in paddy fields, characterized in that, Includes the following steps: S1: Acquire multi-source time-series data of the target paddy field within the observation period and preprocess it under a unified spatiotemporal reference. The multi-source time-series data includes at least time-series remote sensing observation data for characterizing the flooding state and wet-dry transition process of the paddy field, water process data for characterizing the field water change process, and meteorological data for characterizing environmental driving conditions. S2: Based on the multi-source time-series data preprocessed in step S1, construct the crop growth trajectory variable group and water process variable group for the target paddy field respectively; S3: Based on the crop growth trajectory variable group and the water process variable group, construct a set of intermediate channel state variables to characterize the formation, transport and inhibition of greenhouse gases in paddy fields. The intermediate channel state variable group includes at least the substrate supply potential index, plant transport potential index, oxidation inhibition potential index and dry-wet transition pulse index. S4: Based on the dynamic change characteristics of the crop growth trajectory, the change characteristics of the intermediate channel state variable group, and the field dry-wet transition events, identify the dynamic stage boundaries of greenhouse gas emissions and form a stage sequence that changes over time. S5: Construct multiple candidate dynamic sub-models for different stages and event states, including stage sub-models for the continuous flooding growth stage, event sub-models for the drying response stage, event sub-models for the re-watering pulse stage, and stage sub-models for the growth and decline stage; based on the stage sequence obtained in step S4 and the joint state vector at the current moment, calculate the matching degree of each candidate dynamic sub-model to the current stage or event state, and determine the corresponding switching weight accordingly. S6: Using the aforementioned greenhouse gas emission dynamic reconstruction model, the emission process of at least one greenhouse gas in the target paddy field is continuously reconstructed to obtain the corresponding stage structure reconstruction results. S7: Based on one or more of the following: a small amount of ground-based greenhouse gas flux observation data, historical sample database data, or local UAV observation data, perform deviation correction, parameter constraint, or migration correction on the greenhouse gas emission dynamic reconstruction model established in step S5 and the stage structure reconstruction results obtained in step S6 to obtain the corrected stage structure reconstruction results. S8: Map the corrected stage structure reconstruction results obtained in step S7 to field-scale grid cells to generate the spatial distribution results of greenhouse gas emission stage structure of the target paddy field, and output the corresponding stage risk warning information and water management feedback information.

2. The method for dynamic reconstructing the greenhouse gas emission stage structure of paddy fields according to claim 1, characterized in that, In step S1, the remote sensing observation data includes Sentinel-1 synthetic aperture radar remote sensing data, Sentinel-2 optical remote sensing data, and UAV remote sensing data. Field water process data include field surface water level data, soil moisture content data, and flooding status sequence data; meteorological data include air temperature, precipitation, radiation, humidity, and wind speed; Sentinel-1 data is used to identify field surface flooding status, waterlogging duration, and wet-dry transition events; Sentinel-2 data is used to retrieve leaf area index, canopy growth status, and growth process; UAV data is used for local high-resolution auxiliary calibration; and meteorological data is used to characterize environmental driving conditions such as temperature, precipitation, radiation, humidity, and wind speed.

3. The method for dynamic reconstructing the greenhouse gas emission stage structure of paddy fields according to claim 1, characterized in that, In step S1, preprocessing includes multi-source data time registration, spatial registration, coordinate unification, outlier removal, noise suppression, missing data completion, scale conversion, and standardization. Specifically, speckle noise suppression is performed on synthetic aperture radar remote sensing data, cloud and shadow identification and removal are performed on optical remote sensing data, and data from different sources are resampled to a unified field grid.

4. The method for dynamic reconstructing the greenhouse gas emission stage structure of paddy fields according to claim 1, characterized in that, In step S2, the reconstruction of the crop growth trajectory is achieved by combining remote sensing inversion, temporal smoothing, and growth process constraints. The crop growth trajectory further includes one or more dynamic features from trajectory inflection points, growth rates, accelerations, peak times, and stage inflection points.

5. The method for dynamic reconstructing the greenhouse gas emission stage structure of paddy fields according to claim 1, characterized in that, In step S3, the intermediate channel state variable set is not directly given by a single greenhouse gas flux observation, but is constructed from crop growth trajectory variables, water process variables, and their derived characteristics. Specifically, the substrate supply potential index is characterized by one or more of the following: leaf area index, aboveground biomass, biomass growth rate, and growth progress indicators; the plant transport potential index is characterized by one or more of the following: leaf area index, plant height, and growth progress indicators; the oxidation inhibition potential index is characterized by one or more of the following: flooding state, flooding duration, field surface water level, or soil moisture content; and the dry-wet transition pulse index is characterized by one or more of the following: flooding state change rate, desiccation event indicator, and rehydration event indicator. The intermediate channel state variable set is obtained through normalization calculation, weighted combination, rule mapping, or a combination thereof.

6. The method for dynamic reconfiguration of greenhouse gas emission stages in paddy fields according to claim 1, characterized in that, In step S4, the dynamic stage boundary is not divided according to a fixed calendar period, but is jointly identified based on at least two of the following: crop growth trajectory turning point information, water state change information, and intermediate channel state variable mutation information.

7. The method for dynamic structural reconfiguration of greenhouse gas emission stages in paddy fields according to claim 1, characterized in that, In step S5, the adaptive structure switching is a model structure-level switching, rather than parameter updating within a single fixed model. The candidate dynamic sub-models include at least two or more of the following: stage sub-models for the continuous flooding growth phase, event sub-models for the drying response phase, event sub-models for the re-flooding pulse phase, and stage sub-models for the growth and decline phase.

8. The method for dynamic reconfiguration of greenhouse gas emission stages in paddy fields according to claim 1, characterized in that, In step S5, the joint state vector includes at least one or more of the following: crop growth trajectory variables, intermediate channel state variables, water process variables, and meteorological driving variables; the switching weight is used to characterize the degree of matching of each candidate kinetic sub-model with the current stage or event state, and is determined by threshold discrimination, probability discrimination, gating function, state machine rule, or a combination thereof; when the switching weight of a candidate kinetic sub-model is greater than a preset threshold, the candidate kinetic sub-model is called to output the reconstruction result, or the weighted fusion output is performed according to the switching weights of each candidate kinetic sub-model.

9. The method for dynamic reconfiguration of greenhouse gas emission stages in paddy fields according to claim 1, characterized in that, In step S6, the stage structure reconstruction results are for at least one or more of the generation of methane, nitrous oxide, and carbon dioxide. The contribution rate of each stage is calculated as the proportion of the reconstructed flux of each stage to the cumulative flux over the entire observation period, and the event pulse contribution is calculated as the incremental flux within a preset time window before and after the drying-refilling process.

10. The method for dynamic reconstructing the stage structure of greenhouse gas emissions in paddy fields according to claim 1, characterized in that, In step S7, the model calibration uses one or more of the following as constraint information: a small amount of ground-based greenhouse gas flux observation data, historical sample database data, or local UAV observation data. The model calibration performs one or more of the following on the greenhouse gas emission dynamic reconstruction model established in step S5 and the stage structure reconstruction results obtained in step S6: linear deviation correction, stage residual correction, parameter constraint, posterior shrinkage, or migration correction, in order to obtain the corrected stage structure reconstruction results.

Images