Regional scale crop nitrogen dynamic regulation and control method and device

Abstract

The invention discloses a regional scale crop nitrogen dynamic regulation and control method and device. The method comprises the following steps of obtaining multi-source data and constructing a space-time database; selecting a crop growth model, and calibrating and verifying the crop growth model based on the space-time database to obtain the crop growth model and generate a localized model parameter set; dividing a target area into space units, inputting soil data, meteorological data and a localized model parameter set of each space unit to a crop growth model, simulating crop growth processes of different nitrogen gradients, extracting optimal biomass and critical nitrogen concentration data point pairs of each space unit, and performing fitting to generate a spatialized pCNDC parameter map layer; inverting an actual biomass and nitrogen concentration spatial distribution map based on remote sensing data, calculating critical nitrogen concentration and nitrogen nutrition index of each spatial unit in combination with a pCNDC parameter map layer, and generating a spatial distribution map; when the nitrogen nutrition index is lower than a preset threshold value, the nitrogen deficiency amount is calculated, and a variable fertilization prescription map is generated.

Description

Technical Field

[0001] This invention relates to the field of agricultural technology, and more specifically, to a method and equipment for dynamic regulation of crop nitrogen at a regional scale. Background Technology

[0002] Nitrogen is one of the essential macronutrients for crop growth and development, playing a decisive role in photosynthesis, protein synthesis, and ultimately, yield and quality. In global agricultural production, nitrogen fertilizer application is a key measure to ensure food supply and increase yield per unit area. However, traditional nitrogen fertilizer management methods have long been based on experience or regional average recommendations, failing to fully consider the significant variability in the actual nitrogen requirements of crops over time and in spatial distribution.

[0003] This extensive management model often leads to insufficient nitrogen fertilizer application in some areas, limiting the full potential of crop yields; while in other areas, excessive application not only increases agricultural production costs and wastes valuable nitrogen fertilizer resources, but also causes serious environmental problems, such as soil acidification, groundwater nitrate pollution, eutrophication of water bodies, and increased emissions of greenhouse gases (such as nitrous oxide, N2O). Therefore, developing and applying precision nitrogen management technology for crops to achieve intelligent fertilization "on demand, in quantity, on time, and according to location" has become an inevitable requirement and an important direction for the sustainable development of modern agriculture.

[0004] The Critical Nitrogen Dilution Curve (CNDC) theory provides an important physiological basis for crop nitrogen nutrition diagnosis. This theory states that during crop growth, to maintain maximum growth rate, the critical nitrogen concentration (N) within the aboveground parts of the plant is [a specific value, likely related to nitrogen concentration]. c The minimum nitrogen concentration required to ensure maximum growth rate (N0) exhibits a power-law dilution relationship with increasing aboveground biomass (W), typically expressed as N0. c =a×W -b , where a and b are parameters related to crop type and environmental conditions.

[0005] Based on the CNDC theory, the actual measured nitrogen concentration (N) of crops can be compared. actual ) and the critical nitrogen concentration (N) corresponding to the current biomass c ), calculate the Nitrogen Nutrition Index (NNI), i.e., NNI = N actual / N c NNI is widely considered a reliable indicator for real-time assessment of whether a crop is in a state of nitrogen deficiency (NNI<1), sufficiency (NNI≈1), or surplus (NNI>1).

[0006] Although CNDC theory and its derived NNI index have shown great application potential in scientific research, their promotion and application in actual agricultural production, especially at the regional scale, face serious challenges. Most existing CNDC studies are based on site trials or conducted on relatively uniform small plots. Obtaining CNDC through experimental research not only incurs high time and economic costs, but the obtained CNDC parameters (a,b) are often only applicable to specific experimental sites, crop types, and environmental conditions.

[0007] However, real agricultural regions (such as a county, an irrigation district, or a large farm) typically exhibit significant spatial heterogeneity. This includes differences in climatic conditions (e.g., uneven spatial distribution of precipitation, temperature, and sunlight), soil diversity (e.g., spatial variations in soil type, organic matter content, texture, and fertility levels), regionalized crop types, and differences in field management practices (e.g., sowing time and irrigation methods). This high degree of spatial heterogeneity leads to significant differences in crop growth processes and nitrogen requirements at different locations within the region, making it difficult to directly and accurately generalize CNDC parameters obtained from site studies to the entire region.

[0008] Currently, research is actively underway that combines CNDC theory, crop growth models, and remote sensing technology for the dynamic and spatial assessment of crop nitrogen status at the regional scale, or for variable fertilization decision-making. Related Chinese patent literature also reflects R&D activities in CNDC and NNI diagnostic applications, remote sensing nitrogen monitoring, and variable fertilization technology. For example, CN106841051A proposes using crop canopy image greenness and coverage information combined with a critical nitrogen dilution model to calculate NNI; CN107505271A focuses on plant nitrogen estimation based on a nitrogen component radiative transfer model; while CN102487644A and CN103959973A involve variable fertilization devices and precision application methods.

[0009] However, the relevant technologies have not provided any technical insights into how to overcome the barriers of spatial heterogeneity, accurately, dynamically, and spatially generate and apply nitrogen diagnostic indicators, realize variable fertilization based on real-time crop needs, and achieve a novel nitrogen fertilizer management method that can self-optimize through feedback to adapt to long-term changing environments. Summary of the Invention

[0010] To address the aforementioned problems in existing technologies, this invention provides a regional-scale crop nitrogen dynamic regulation method and device, which can achieve precise variable fertilization based on crop physiological needs, environmental potential, and real-time growth status, based on real-time monitoring and closed-loop learning mechanisms.

[0011] The objective of this invention is achieved through the following technical solutions.

[0012] As a first aspect of this application, some embodiments of this application provide a method for dynamic regulation of crop nitrogen at a regional scale, comprising the following steps:

[0013] To acquire multi-source data of the target area to construct a spatiotemporal database, the multi-source data includes meteorological data, soil data, crop growth dynamic data, crop management measures data, and remote sensing data;

[0014] Select a crop growth model with a nitrogen response module, calibrate and validate the crop growth model based on a spatiotemporal database, obtain a crop growth model that reflects the dynamic response of crop nitrogen in the target area, and generate a localized model parameter set.

[0015] The target area is divided into spatial units. Soil data, meteorological data and localized model parameter sets of each spatial unit are input into the crop growth model. The crop growth process under nitrogen gradient is simulated under no water stress conditions. The optimal biomass and critical nitrogen concentration data point pairs of each spatial unit are extracted and the spatialized pCNDC parameter layer of the target area is fitted.

[0016] Based on remote sensing data, the actual biomass and nitrogen concentration spatial distribution map are inverted. The critical nitrogen concentration and nitrogen nutrient index of each spatial unit are dynamically calculated by combining the pCNDC parameter layer and spatial distribution maps are generated respectively. The nitrogen deficit is calculated based on the generated spatial map. The variable fertilizer prescription map is generated by combining the soil nitrogen supply potential and nitrogen fertilizer use efficiency to guide fertilization and realize the dynamic regulation of crop nitrogen at the regional scale.

[0017] Furthermore, after fertilization is carried out according to the variable fertilization prescription map, feedback data is collected and added to the spatiotemporal database. The feedback data includes crop yield, crop nitrogen uptake data at harvest, fertilization cost data, and control data.

[0018] Based on the feedback data, the crop growth model is optimized, the remote sensing data inversion is optimized, and the threshold is adjusted to form a closed-loop control mechanism, thereby achieving continuous optimization of the regional-scale crop nitrogen dynamic control method.

[0019] Furthermore, a quantitative evaluation system was used to screen candidate crop growth models. The crop growth models possessed a nitrogen response module and were applicable to the main varieties cultivated in the target area. The quantitative evaluation system included:

[0020] The data fit dimension is used to evaluate the degree of matching between the model's input requirements and the spatiotemporal database;

[0021] Accessibility and scalability are dimensions used to evaluate the openness of the model code and the ease of parameter adjustment;

[0022] The case studies and literature support dimension is used to count the number of application cases of the model in the target crop type.

[0023] Furthermore, the steps for calibrating a crop growth model include: setting the parameters of the crop growth model;

[0024] The sensitivity of crop growth model parameters to output variables was analyzed using a global sensitivity analysis method, and the key model parameters with the greatest impact on nitrogen response simulation were identified.

[0025] By utilizing crop growth dynamic data from a spatiotemporal database and employing a multi-objective optimization algorithm to calibrate key model parameters, the overall performance of the crop growth model is optimized across the entire nitrogen response range, from nitrogen deficiency to nitrogen abundance.

[0026] Furthermore, during the simulation of crop growth under different nitrogen gradients, the treatments of different nitrogen gradients were divided into the following groups based on the biomass values: the low biomass group with the smallest value, the high biomass group or plateau group with the largest value, and the rest being the medium biomass group.

[0027] Within the high biomass group or plateau period group, the treatment with the lowest nitrogen application rate is selected, representing the minimum nitrogen supply required for the crop to reach its maximum growth rate during the current growth stage. The corresponding aboveground biomass is recorded as the optimal biomass, and the plant nitrogen concentration is recorded as the critical nitrogen concentration, thus obtaining the optimal biomass and critical nitrogen concentration data point pairs for the spatial unit.

[0028] Furthermore, based on the data point pairs of optimal biomass and critical nitrogen concentration for each spatial unit, a power function equation for the critical nitrogen dilution curve is fitted to obtain the pCNDC parameters a and b for each spatial unit. The power function equation for the critical nitrogen dilution curve is expressed as:

[0029] N c =a×W optimal -b ;

[0030] In the formula, W optimal N represents biomass. c Indicates the critical nitrogen concentration;

[0031] The parameters a and b of all spatial units are collected and stored as raster layers according to their spatial location, forming a spatialized pCNDC parameter a layer and a spatialized pCNDC parameter b layer, which together constitute a spatialized pCNDC parameter layer covering the target area.

[0032] Furthermore, the steps for retrieving the actual biomass and nitrogen concentration spatial distribution map based on remote sensing data include: acquiring remote sensing data covering the key growth stages of crops in a spatiotemporal database, and calculating vegetation indices sensitive to biomass.

[0033] A biomass inversion model is established based on vegetation index and ground-measured biomass to generate a spatial distribution map of actual biomass covering the target area and corresponding to the time phase of remote sensing data.

[0034] Based on the spectral features of plants sensitive to nitrogen in remote sensing data, including red edge parameters, chlorophyll index, nitrogen reflectance index or specific band combination spectral information, an actual nitrogen concentration inversion model is established by combining the measured nitrogen concentration of plants on the ground, and the spatial distribution map of the actual nitrogen concentration corresponding to the time phase of the remote sensing data is generated.

[0035] Furthermore, the actual biomass spatial distribution map is spatially overlaid with the spatialized pCNDC parameter layer, and spatial cell matching is performed to obtain the actual biomass and parameters a and b for each spatial cell. The critical nitrogen concentration of the spatial cell is then calculated and expressed as:

[0036] N c =a×(W actual ) -b ;

[0037] In the formula, N c W represents the critical nitrogen concentration. actual Indicates actual biomass;

[0038] The critical nitrogen concentrations of all spatial units are collected and a spatial distribution map of the critical nitrogen concentrations of the corresponding time phases covering the target area is generated.

[0039] The actual nitrogen concentration spatial distribution map is spatially overlaid with the critical nitrogen concentration spatial distribution map, and spatial cell matching is performed to obtain the actual plant nitrogen concentration value and critical nitrogen concentration value of each spatial cell. The nitrogen nutrient index of the spatial cell is then calculated and expressed as:

[0040] NNI = N actual / N c ;

[0041] In the formula, NNI represents the nitrogen nutritional index, N actual This represents the actual nitrogen concentration value of the plant, N. c Indicates the critical nitrogen concentration;

[0042] The nitrogen nutrient index of all spatial units is collected to generate a spatial distribution map of the nitrogen nutrient index of the corresponding time phase covering the target area;

[0043] Based on the spatial distribution map of nitrogen nutrient index, the crop nitrogen nutrient status of each spatial unit is assessed by dynamically preset thresholds. When the nitrogen nutrient index is lower than the threshold, the nitrogen deficit is calculated, and a variable fertilization prescription map is generated by combining soil nitrogen supply potential and nitrogen fertilizer utilization efficiency.

[0044] Furthermore, the steps for generating a variable fertilization prescription map include: reading the critical nitrogen concentration, actual nitrogen concentration, and actual biomass of each spatial unit, calculating the nitrogen deficit, and expressing it as:

[0045] N deficit =(N c -N actual )×W actual ;

[0046] In the formula, N deficit Nitrogen deficit, N c The critical nitrogen concentration for the current biomass, N actual W represents the real-time actual nitrogen concentration of the crop. actual This refers to the actual biomass of the crop in real time.

[0047] Subtracting the soil's nitrogen supply potential from the nitrogen deficit yields the net nitrogen demand (N). net_demand , represented as:

[0048] N net_demand =N deficit -SoilN supply ;

[0049] Predicting nitrogen fertilizer use efficiency and calculating recommended nitrogen application rate ΔN recommend , represented as:

[0050] ΔN recommend =N net_demand / NUE;

[0051] The recommended nitrogen application rates for all spatial units are collected and stored as raster layers according to spatial location to obtain a variable fertilization prescription map.

[0052] As a second aspect of this application, some embodiments of this application provide a regional-scale crop nitrogen dynamic regulation device, including a memory storing a computer program; a processor executing the computer program to implement the steps of the above-described method; and a communication module for receiving input from an external meteorological data server and a remote sensing data platform.

[0053] Compared with the prior art, the advantages of this invention are:

[0054] (1) A spatialization method based on the potential critical nitrogen dilution curve (pCNDC) of crop growth model is proposed. The parameter is fitted for each spatial unit by simulating the nitrogen gradient without water stress, which accurately quantifies the impact of meteorological, soil and variety differences on crop nitrogen demand. This method replaces the traditional point-based repeated test measurement mode, significantly reduces data collection costs and improves regional application.

[0055] (2) Establish a dynamic diagnosis and real-time decision-making mechanism to improve management timeliness. By integrating the growth "potential" simulated by the model with the "actual" crop growth monitored by remote sensing, a spatial distribution map of the nitrogen nutrient index (NNI) is dynamically generated, enabling real-time diagnosis of crop nitrogen surplus or deficit. The decision-making process also incorporates weather forecasts to adjust fertilization plans, which is significantly better than the traditional static management model, ensuring the real-time and efficient nature of fertilization decisions.

[0056] (3) By introducing a closed-loop learning mechanism of yield-nitrogen absorption-economic benefits, the crop growth model, remote sensing algorithm, and fertilization decision logic can be automatically corrected and optimized using actual data such as the yield map at the end of the season. This self-learning closed-loop system enables the method to continuously adapt to changes in climate, varieties, and management measures, ensuring the robustness and advancement of the technical solution in long-term application. Attached Figure Description

[0057] Figure 1 This is a flowchart of the steps of a regional-scale crop nitrogen dynamic regulation method in one embodiment of the present invention;

[0058] Figure 2 This is the critical nitrogen dilution curve of a spatial unit in one embodiment of the present invention. Detailed Implementation

[0059] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0060] Combination Figures 1 to 2 The present invention provides a method for dynamic regulation of crop nitrogen at a regional scale, comprising the following steps:

[0061] S1. Constructing a spatiotemporal database

[0062] Multi-source data from the target region are acquired to construct a spatiotemporal database, which comprehensively characterizes the meteorological conditions, soil characteristics, crop growth dynamics, and crop management measures of the target region, thereby serving as the data foundation for crop growth model calibration, nitrogen dynamic diagnosis, and variable fertilization decisions.

[0063] Specifically, multi-source data includes meteorological data, soil data, crop growth dynamics data, crop management data, and remote sensing data, and the acquisition process is as follows:

[0064] (1) Meteorological data acquisition:

[0065] Acquire historical daily meteorological data (at least the past 10 years) and real-time daily meteorological data from meteorological stations in or around the target area. The meteorological data includes maximum temperature, minimum temperature, average temperature, precipitation, solar radiation, wind speed, and relative humidity.

[0066] Meteorological data sources include the national meteorological network, local meteorological stations, and reanalysis datasets. Spatial interpolation methods (such as inverse distance weighting or kriging interpolation) are used to process the meteorological data observed at the stations, ultimately generating a gridded meteorological data layer covering the target area, reflecting the spatiotemporal differentiation of climate conditions.

[0067] (2) Soil data acquisition:

[0068] Soil texture (percentage of sand, silt and clay), bulk density, pH value, organic matter content, total nitrogen, available nitrogen, phosphorus and potassium content, field water holding capacity and wilting coefficient are extracted from soil survey maps or digital soil maps. These properties are then interpolated using geostatistical methods to generate a gridded soil attribute layer, resulting in soil data used to quantify soil fertility and nitrogen supply potential.

[0069] (3) Acquisition of crop growth dynamic data:

[0070] Crop growth dynamics data aims to quantitatively characterize the morphology, phenology, and nutrient uptake processes of different crops throughout their entire growth period within a target region, for use in calibrating subsequent crop growth models. Crop growth dynamics data includes growth indicators at key growth stages, as well as soil moisture and soil mineral nitrogen dynamics.

[0071] Crop growth dynamics data were obtained by designing and implementing field trials covering the main cultivated varieties, major soil types, and nitrogen supply level gradients in the target area. During the field trials, crop growth indicators, soil moisture, and soil mineral nitrogen dynamics at key growth stages were monitored as crop growth dynamics data. The main cultivated varieties refer to the primary varieties planted on the cultivated land in the target area, such as those accounting for more than 10% of the cultivated land area. Major soil types refer to soil types whose cultivated land area accounts for more than 10% of the cultivated land area in the target area.

[0072] The critical growth period refers to a period that has a significant impact on a crop's nitrogen response, nutrient uptake, and final yield formation. For example, for rice, it typically includes the tillering stage, jointing stage, booting stage, heading stage, grain-filling stage, and maturity stage.

[0073] Growth indicators include aboveground biomass, biomass of different organs (such as stems, leaves, panicles, or grains in rice), plant nitrogen concentration or content, leaf area index (LAI), final yield and yield components, and quality indicators. Quality indicators may include protein content, amylose content, chalkiness, etc.

[0074] In one specific embodiment, when field trial conditions are lacking, historical data from field trials already conducted in the target area can be obtained by consulting literature, which can then be used as dynamic data on crop growth.

[0075] (4) Data acquisition of crop management measures:

[0076] Data on crop management measures are obtained through agricultural statistical yearbooks, records from local agricultural and rural departments, or farmer surveys.

[0077] The data on crop management practices obtained include the main varieties and their growth characteristics, sowing date, transplanting date, planting density, irrigation system, and historical fertilization data. Historical fertilization data includes the type of fertilizer, the timing of fertilization, and the amount of fertilizer used. These data together constitute the crop management practices data that depict the basic situation of agricultural production activities in the region.

[0078] (5) Remote sensing data acquisition:

[0079] Remote sensing data is obtained by acquiring RGB (red, green, and blue band visible light images), multispectral, or hyperspectral remote sensing images covering the entire crop growing season and preprocessing them.

[0080] Specifically, the data sources for remote sensing are medium-to-high resolution satellites (such as Sentinel-2 or Landsat 8 / 9), unmanned aerial vehicle (UAV) platforms, or airborne remote sensing. Images of the target area with no cloud cover or a cloud cover percentage of less than 10% are selected and preprocessed with radiometric calibration, atmospheric correction, geometric fine correction, and cloud masking to obtain the remote sensing data. This preprocessing ensures that the image quality meets the accuracy requirements for subsequent biomass and nitrogen concentration inversion.

[0081] The preprocessed remote sensing data undergoes spatiotemporal standardization and registration, and is then integrated into the spatiotemporal database as a key dynamic data layer. This process specifically includes:

[0082] Time series construction: Associate each remote sensing image with a precise imaging date and time, and organize and index them in chronological order to form an image time series that covers the entire growing season and reflects the dynamic process of crop growth.

[0083] Spatial alignment and fusion: Ensure that all remote sensing image layers use the same coordinate system and spatial resolution as the static data layers such as meteorological and soil data in the database. Through spatial alignment, spatial units at any location can be accurately matched and linked with soil properties and meteorological data at the corresponding location for analysis.

[0084] Through the above-described data entry process, discrete, static remote sensing images are transformed into dynamic monitoring layers in a spatiotemporal database that are queryable, analyzable, spatially explicit, and temporally continuous. This not only enables effective management of Earth observation data but also serves as a bridge connecting crop growth model simulations with actual field growth.

[0085] S2. Selection and Calibration of Crop Growth Models

[0086] Select a crop growth model and calibrate and validate the model based on a spatiotemporal database to obtain a localized model parameter set that reflects the dynamic response of crop nitrogen in the target area, specifically including:

[0087] (1) Model selection:

[0088] Choose crop growth models that can simulate the mechanisms of nitrogen absorption, distribution and utilization by crops and have a good application foundation, such as WOFOST, DSSAT series, APSIM, ORYZA and STICS.

[0089] First, candidate crop growth models with complete nitrogen response modules and applicable to the main crops in the target area (referring to the crop types that dominate the planting area, total yield, or economic value in the area) are selected. Quantitative evaluation is then used to select from the candidate crop growth models based on three evaluation dimensions: data fit, availability and scalability, and case and literature support.

[0090] Data fit is considered in relation to the degree of matching and availability of the input data required by the crop growth model with the spatiotemporal database. Availability and scalability are considered in relation to the cost of acquiring the crop growth model (open source preferred), the openness of its code, and the ease of parameter tuning and feature expansion. Case studies and literature support are considered in relation to the breadth and depth of published validation studies and application cases of the crop growth model in similar agroecological zones or for target crop types.

[0091] In one specific embodiment, when the target crop is rice, candidate crop growth models may include the ORYZA(v3) model and the CERES-Rice model in DSSAT v4.8. The ORYZA(v3) model is specifically designed for rice and has a high data fit; the CERES-Rice model, as a DSSAT module, is widely used but may require data format conversion. Finally, the most suitable crop growth model is selected based on a comprehensive evaluation of the above dimensions.

[0092] In one specific embodiment, candidate crop growth models can be quantitatively scored according to evaluation dimensions. For example, 1 point indicates poor performance or failure to meet requirements, 2 points indicates moderate performance or basic compliance with requirements, and 3 points indicates excellent performance or complete compliance with requirements.

[0093] In one specific embodiment, the target crop type is rice, and the candidate crop growth models are the ORYZA(v3) model with a complete nitrogen response module and the CERES-Rice model in DSSAT v4.8.

[0094] The ORYZA(v3) model is specifically designed for rice and has a high degree of data fit; the CERES-Rice model needs to be adapted to the physiological and ecological characteristics of rice, which means it requires data transformation or parameter estimation. Therefore, in this embodiment, the ORYZA(v3) model scores 3 points in data fit, and the CERES-Rice model scores 2 points in data fit.

[0095] The source code for some versions of the ORYZA(v3) model is available within the framework of scientific research collaboration; the CERES-Rice model has high accessibility but requires in-depth source code-level modification and secondary development. Therefore, in this embodiment, the ORYZA(v3) model scores 3 points in accessibility and scalability, while the CERES-Rice model scores 2 points in accessibility and scalability.

[0096] The ORYZA(v3) model, as one of the mainstream rice growth models, has been widely validated and applied, and its accuracy and reliability have been fully demonstrated; the CERES-Rice model also has a wide range of validations and applications. Therefore, in this embodiment, the ORYZA(v3) model is scored 3 points in terms of support from case studies and literature, and the CERES-Rice model is also scored 3 points in terms of support from case studies and literature.

[0097] As shown in Table 1, the ORYZA(v3) model scored 9 points across the three evaluation dimensions, while the CERES-Rice model in DSSAT v4.8 scored 7 points across the three evaluation dimensions. The ORYZA(v3) model was ultimately selected as the crop growth model for this embodiment.

[0098] Table 1 Comparison and evaluation of the ORYZA (v3) model and the CERES-Rice model

[0099] Evaluation Items ORYZA (v3) CERES-Rice Data fit 3 points (Designed specifically for rice) 2 points (data format conversion required) Accessibility and scalability 3 points (requires research collaboration / partially open source) 2 stars (commercial license required) Literature support 3 points (widely validated in global rice-growing regions) 3 points (longest historical application) score 9 points 7 points

[0100] (2) Model calibration:

[0101] Based on the genetic characteristics of crops in the target area, soil data and meteorological data in the spatiotemporal database, parameters are set for the selected crop growth model. The crop growth model parameters include light energy utilization rate, maximum photosynthetic rate, nitrogen absorption efficiency, root growth parameters, nitrogen distribution coefficient among organs, and factors affecting photosynthesis or growth under nitrogen stress.

[0102] Global sensitivity analysis was used to analyze the sensitivity of crop growth model parameters to output variables, and the key model parameters with the greatest impact on nitrogen response simulation were identified.

[0103] Using crop growth dynamics data from a spatiotemporal database, a multi-objective optimization algorithm (such as NSGA-II or MOPSO) was employed to calibrate the key model parameters determined by sensitivity analysis. The optimization objective was to simultaneously minimize the combined error between the simulated biomass, plant nitrogen concentration (or nitrogen content), key growth stages, and final yield of the crop growth model under all nitrogen treatment levels and the measured data (using a combination of root mean square error RMSE, normalized root mean square error nRMSE, and coefficient of determination R0). 2 (And indicators such as consistency index d). Ensure that the crop growth model achieves optimal overall performance across the entire nitrogen response range from nitrogen deficiency to nitrogen abundance, avoiding biases caused by calibration at a single nitrogen level.

[0104] The model calibration yields a localized crop growth model parameter set that reflects the dynamic response of crops to nitrogen in the target area.

[0105] (3) Model validation:

[0106] The calibrated crop growth model was evaluated using a dataset independent of the model calibration. Quantitative evaluation metrics such as root mean square error (RMSE), normalized root mean square error (nRMSE), and coefficient of determination (R²) were employed. 2 The consistency index (d-index) and other metrics are used to assess the consistency between crop growth model simulation results and measured data.

[0107] Specifically, the accuracy of the crop growth model's predictions is verified to meet pre-defined acceptance criteria. For example, the normalized root mean square error (nRMSE) of key growth indicators (such as biomass and plant nitrogen concentration) should be less than 15%, and the coefficient of determination (R²) should be... 2 The d-index is greater than 0.80 and the consistency index is greater than 0.85; meanwhile, the simulation error of the critical reproductive period is within ±7 days.

[0108] Specifically, if the prediction accuracy of the crop growth model fails to meet the aforementioned acceptance requirements, an iterative feedback adjustment mechanism is initiated, returning the model to the calibration phase. Based on the deviations revealed by the model validation results, key model parameters that significantly affect the crop growth model output and were identified during the sensitivity analysis phase are optimized and adjusted. This iterative validation and optimization process will continue until all indicators of the crop growth model during validation meet the preset acceptance requirements, thereby ensuring its reliable predictive ability and good generalization for different environmental conditions and management measures within the region.

[0109] This yields a locally calibrated and validated crop growth model capable of accurately simulating the dynamic growth response of crops to nitrogen within the target area. The crop growth model and its parameter set are the core tools in step S3 for generating spatialized potential critical nitrogen dilution curves (pCNDCs) and parameter layers at the regional scale. By simulating nitrogen gradient experiments under potential growth conditions, critical nitrogen concentration data points are extracted for each spatial unit. Furthermore, the biomass and nitrogen concentration predictions output by the crop growth model are combined with the actual data retrieved from remote sensing to form a closed-loop optimized data chain, ensuring continuous improvement of nitrogen management strategies.

[0110] This step, through model selection, calibration, and validation, constructs a crop growth model adapted to regional characteristics, addressing the issue of insufficient simulation accuracy of general crop growth model parameters under heterogeneous environments. Furthermore, the model system established by this technical solution possesses continuous evolutionary capabilities; by periodically incorporating new crop growth dynamic data, it can dynamically adapt to variety updates and climate change, overcoming the shortcomings of traditional empirical models in terms of regional adaptability.

[0111] S3. Construction of Spatialized Potential Critical Nitrogen Dilution Curve (pCNDC) Parameter Layer

[0112] The target region is divided into spatial units. Based on a crop growth model, the crop growth process under different nitrogen supply gradients is simulated under potential growth conditions without water stress. Potential critical nitrogen dilution curve (pCNDC) parameters are generated for each spatial unit within the target region, and a spatialized pCNDC parameter layer covering the target region is generated, specifically including:

[0113] (1) Regional grid division:

[0114] Based on the size of the target area, the resolution of the expected variable rate technology map (VRT map), and the spatial variability scale of factors affecting crop growth such as soil type and climate factors within the target area, the resolution of the spatial units is determined, and the target area is divided into square spatial units of uniform size.

[0115] Specifically, for small target areas such as farms, a high resolution (tens to hundreds of meters) is used, dividing the target area into spatial units such as 30 meters × 30 meters or 300 meters × 300 meters; for large target areas such as counties, provinces or larger areas, a lower resolution (several to tens of kilometers) is used, dividing the target area into spatial units such as 1 kilometer × 1 kilometer or 10 kilometers × 10 kilometers.

[0116] (2) Crop growth model input acquisition and simulation settings:

[0117] Soil and meteorological data corresponding to each spatial unit are extracted from the spatiotemporal database. The extracted unit soil and meteorological data, along with the localized model parameter set obtained in step S2, are input into the calibrated crop growth model. A nitrogen supply gradient is set in the crop growth model to simulate the crop growth process from sowing to maturity.

[0118] Specifically, a nitrogen supply gradient is set in the crop growth model, ranging from low to high, covering crops from nitrogen deficiency to nitrogen abundance. The nitrogen gradient is set to 0 N / ha - 400 kg N / ha. The gradient interval needs to be small enough to capture response changes, such as a gradient interval of 20 kg N / ha or 30 kg N / ha (e.g., 0 kg N / ha, 20 kg N / ha, 40 kg N / ha... 400 kg N / ha).

[0119] For each nitrogen gradient, the entire growth process of the crop from sowing to maturity is simulated. By setting nitrogen supply gradients from nitrogen deficiency to nitrogen abundance to simulate crop growth, it is possible to capture the biomass and plant nitrogen concentration response curves of the crop under different nitrogen supply conditions.

[0120] More specifically, crop growth models need to be operated under ideal growth conditions without water stress.

[0121] (3) Data extraction and critical point determination during the critical reproductive period:

[0122] When nitrogen supply is below the critical level, increasing nitrogen fertilizer will significantly increase biomass; when nitrogen supply reaches or exceeds the critical level, biomass tends to stabilize or increase slowly, while plant nitrogen concentration may continue to rise. By statistical analysis, the lowest nitrogen application rate treatment corresponding to the inflection point where biomass reaches a plateau is found, and the plant nitrogen concentration at that treatment is the critical nitrogen concentration at that biomass level.

[0123] Therefore, for the key growth stages in the simulation, the simulated aboveground biomass (W) and plant nitrogen concentration (N) under each nitrogen gradient were extracted. Analysis of variance was performed on the biomass under different nitrogen gradient treatments at the same growth stage, and all nitrogen gradient treatments were grouped based on the statistical significance of the differences (p<0.1).

[0124] Specifically, the key growth stages include tillering, jointing, booting, heading, grain-filling, and maturity. Based on ANOVA results and by comparing the significant differences in biomass among different nitrogen level treatments, the treatments were divided into several groups.

[0125] In one specific embodiment, different nitrogen gradient treatments were divided into low biomass, medium biomass, and high biomass or plateau phase groups. First, analysis of variance was used to divide the treatments of different nitrogen gradients into several groups with significant differences based on biomass values. The group with the smallest value was defined as the low biomass group, the group with the largest value was defined as the high biomass or plateau phase group, and the rest were defined as the medium biomass group.

[0126] Within the high biomass group or plateau stage group, select the treatment with the lowest nitrogen application rate. This treatment represents the minimum nitrogen supply required for the crop to reach its maximum growth rate during this growth stage. Record the aboveground biomass corresponding to this treatment as the optimal biomass (W). optimal The unit is t / ha, and the plant nitrogen concentration is used as the critical nitrogen concentration (N). c (The unit is %).

[0127] Repeat the above process during the key growth stages of the crop to obtain a series of (W) for this spatial unit. optimal N c Data point pairs.

[0128] (4) Spatialized pCNDC parameter layer generation:

[0129] Based on a series of (W) for each spatial unit optimal N c Data point pairs, with optimal biomass W optimal The critical nitrogen concentration N is the independent variable. c As the dependent variable, fit the power function equation of the critical nitrogen dilution curve: N c =a×W optimal -b The pCNDC parameters a and b of the unit are obtained by fitting.

[0130] like Figure 2 As shown, in a specific embodiment, with a spatial unit (W) optimal N c By fitting a power function equation to the data points, the critical nitrogen dilution curve N for this spatial cell was obtained. c =3.44×W optimal -0.437 .

[0131] The parameters 'a' obtained from fitting all spatial units are collected and stored as raster layers according to their spatial location, forming a spatialized pCNDC parameter layer 'a'. The parameters 'b' obtained from fitting all spatial units are collected and stored as raster layers according to their spatial location, forming a spatialized pCNDC parameter layer 'b'. These two layers together constitute a spatialized pCNDC parameter layer covering the entire target area.

[0132] This step extends the concept of critical nitrogen dilution curves at the point scale to the regional scale, generating unique pCNDC parameters a and b for each spatial unit within the region, based on potential growth conditions. The resulting spatialized pCNDC parameter layer accurately quantifies the impact of regional environmental heterogeneity on crop nitrogen physiological requirements, providing core dynamic data for subsequent remote sensing-based real-time spatial diagnosis of nitrogen nutrient indices and variable fertilization decisions.

[0133] S4. Dynamic monitoring of crop growth and nitrogen concentration based on remote sensing

[0134] Based on remote sensing data from a spatiotemporal database, an inversion algorithm is used to obtain the spatial distribution maps of actual crop biomass and actual nitrogen concentration within the target area, specifically including:

[0135] Specifically, remote sensing data covering key growth stages of crops are acquired from spatiotemporal databases. Vegetation indices sensitive to biomass are calculated using the remote sensing data. An actual biomass retrieval model is established based on the vegetation indices and measured ground biomass for retrieval, or a radiative transfer model (such as PROSAIL) is used for retrieval. This data is then applied to the remote sensing data to generate real-time actual biomass (W) covering the target area and corresponding to the time phase of the remote sensing data. actual Spatial distribution map.

[0136] Among them, vegetation indices include Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Soil Adjusted Vegetation Index (SAVI), and Wide Dynamic Range Vegetation Index (WDRVI).

[0137] Using spectral features sensitive to plant nitrogen in remote sensing data, such as red edge position (REP), chlorophyll index (SPAD), nitrogen reflectance index (NRI), or specific band combinations of spectral information, an inversion model for actual nitrogen concentration is established by combining ground-measured plant nitrogen concentration data, or an inversion based on a radiative transfer model is used. This model is applied to remote sensing data and generates real-time actual nitrogen concentrations (N) corresponding to the time phase of the remote sensing data. actual Spatial distribution map.

[0138] Specifically, ground-based measured data were used to validate the accuracy of the actual biomass retrieval model and the actual nitrogen concentration retrieval model. Quantitative indicators (such as root mean square error RMSE or coefficient of determination R) were employed. 2 (etc.) Evaluate the consistency between the inversion results and the measured values ​​to ensure that the inversion results have physiological rationality and spatiotemporal continuity.

[0139] The actual biomass (W) generated in this step actual Spatial distribution map and actual nitrogen concentration (N)actual The spatial distribution map is a necessary input for calculating the spatialized critical nitrogen concentration and nitrogen nutrient index. Furthermore, the real-time crop growth data obtained in this step can be fed back into the crop growth model in step S2 for dynamic optimization of model parameters. Compared to traditional sampling and detection methods, this step achieves dynamic regional monitoring, overcoming the drawbacks of traditional field sampling, such as time consumption, labor intensity, and limited spatial coverage. It enables high-frequency, large-scale, and non-destructive dynamic monitoring of crop growth status (biomass and nitrogen concentration) at the regional scale. The established standardized inversion process ensures the comparability, reliability, and traceability of monitoring results.

[0140] S5, Dynamic Assessment of Nitrogen Supply Profit and Loss

[0141] By combining spatialized pCNDC parameter layers, actual biomass spatial distribution maps, and actual nitrogen concentration spatial distribution maps, the critical nitrogen concentration and nitrogen nutrient index of each spatial unit within the target region are dynamically calculated and spatial distribution maps are generated respectively, thereby dynamically assessing the nitrogen surplus / deficit status of crops at the regional scale. Specifically, this includes:

[0142] The actual biomass (W) generated in step S4 actual The spatial distribution map is spatially overlaid and spatial cell matched with the spatialized pCNDC parameter layers (including parameter a layer and parameter b layer) generated in step S3. The specific method is as follows:

[0143] Ensure that all raster layers involved in the calculation have a unified coordinate reference system. If the resolutions of the layers are inconsistent, use a resampling algorithm based on the preset calculation resolution to resample all layers to the same grid system, thereby ensuring that each spatial unit can correspond one-to-one.

[0144] Under a unified grid system, each spatial unit is traversed, and at the same geographical location, the actual biomass W corresponding to that spatial unit is extracted from the actual biomass spatial distribution map, the pCNDC parameter a layer, and the parameter b layer. actual And parameters a and b. Based on the extracted values ​​of each spatial cell, the critical nitrogen concentration of that spatial cell is calculated using the critical nitrogen dilution curve equation:

[0145] N c =a×(W actual ) -b ;

[0146] The calculation results of all spatial units are aggregated to generate the critical nitrogen concentration (N) for the corresponding time phase covering the target region. c A spatial distribution map is used to represent the minimum plant nitrogen concentration required for a crop to reach its maximum growth potential or optimal economic benefits at the current actual biomass level.

[0147] The actual nitrogen concentration (N) generated in step S4 actual Spatial distribution map and the critical nitrogen concentration (N) generated above c Spatial distribution maps are overlaid and spatial units are matched.

[0148] From the actual nitrogen concentration (N actual The spatial distribution map obtains the actual plant nitrogen concentration (N) value for each spatial unit. actual From the critical nitrogen concentration (N) c The spatial distribution map is used to obtain the critical nitrogen concentration (N) value of this spatial unit. c Calculate the nitrogen nutrient index (NNI) for this space cell:

[0149] NNI = N actual / N c ;

[0150] The calculation results of all spatial units are aggregated to generate a spatial distribution map of the nitrogen nutrient index (NNI) for the corresponding time phase covering the target area.

[0151] Specifically, based on the spatial distribution map of the nitrogen nutrient index (NNI), the crop nitrogen nutrient status of each spatial unit is assessed by dynamically preset thresholds. These thresholds include nitrogen deficiency thresholds and nitrogen surplus thresholds. The thresholds are not fixed but are dynamically set according to crop type, growth stage, and management objectives. The assessment process is as follows:

[0152] When the NNI value is lower than the preset nitrogen deficiency threshold (e.g., 0.95), it indicates that the crop is under nitrogen stress and the actual nitrogen concentration is lower than the critical level. The smaller the value, the more severe the nitrogen deficiency.

[0153] When the NNI value is between the preset nitrogen deficiency threshold (e.g., 0.95) and the nitrogen surplus threshold (e.g., 1.05), it indicates that the crop's nitrogen nutrition is at or near the optimal level, and its recommended nitrogen application rate is set to zero.

[0154] When the NNI value is higher than the preset nitrogen surplus threshold (e.g., 1.05), it indicates that the crop has sufficient or excessive nitrogen supply, and there may be luxury absorption, so there is no need to supplement nitrogen fertilizer.

[0155] Luxury absorption refers to the phenomenon of excessive absorption when nitrogen supply exceeds crop demand.

[0156] The critical nitrogen concentration (N) generated in this step cSpatial distribution maps and nitrogen nutrient index (NNI) spatial distribution maps provide regional-scale benchmarks for crop nitrogen physiological requirements and quantitative indicators of crop nitrogen surplus / deficit status. Combined with remote sensing data and spatialized pCNDC parameters, this enables high-frequency, wide-area, and dynamic diagnosis of crop nitrogen nutrient status. Through spatial unit-level calculations, it can accurately identify specific locations within the target area where nitrogen deficit or excess exists, overcoming the limitations of traditional averaging or one-size-fits-all diagnostic methods based on limited points. The computational framework based on critical nitrogen dilution curve theory ensures that the diagnostic results conform to the basic physiological laws of crop nitrogen uptake and biomass accumulation.

[0157] S6. Generation of Variable Fertilizer Prescription Diagram

[0158] Based on the spatial distribution map of the nitrogen nutrient index (NNI), fertilization decisions are triggered when the NNI value of a spatial unit is lower than a preset NNI threshold. The nitrogen deficit is calculated, and after deducting the soil's nitrogen supply potential, a variable fertilization prescription map is generated, which specifically includes:

[0159] (1) Fertilization decision trigger:

[0160] The NNI threshold is set based on crop type and current growth stage. Real-time NNI values ​​for each spatial unit within the target area are read based on the NNI spatial distribution map. When the real-time NNI value of a spatial unit is lower than the NNI threshold, a fertilization decision is triggered for that unit. The NNI threshold during the tillering stage is lower than the NNI threshold during the jointing stage.

[0161] In one specific embodiment, the NNI threshold is set to 0.90 or 0.95.

[0162] (2) Calculation of nitrogen deficit:

[0163] For spatial units that trigger fertilization recommendations, their critical nitrogen concentration, actual nitrogen concentration, and actual biomass are read, and the nitrogen deficit is calculated. The nitrogen deficit is the amount of nitrogen required for the spatial unit to recover to a nitrogen-nutrition-adequate state (NNI≈1), and the calculation process is expressed as follows:

[0164] N deficit =(N c -N actual )×W actual ;

[0165] In the formula, N deficit Nitrogen deficit, N c The critical nitrogen concentration for the current biomass, N actual W represents the real-time actual nitrogen concentration of the crop. actual This represents the actual biomass of the crop in real time.

[0166] (3) Calculation of recommended nitrogen application rate:

[0167] For spatial units that require fertilization, their soil nitrogen supply potential (Soil N) supply This refers to the amount of available nitrogen that the soil itself can provide during the critical nitrogen uptake period, which can be estimated using a simplified empirical model. One specific estimation method is as follows:

[0168] SoilN supply =SOM×k min

[0169] In the formula, SOM represents the soil organic matter content of the spatial unit (which can be directly obtained from the spatiotemporal database constructed in step S1); k min The soil nitrogen mineralization coefficient is an empirical coefficient whose value can be determined based on the soil type and climate conditions of the target area by consulting localized agricultural research reports or by setting a regionally universal value (e.g., 2% for paddy soils).

[0170] The net nitrogen demand (N) is obtained by subtracting the estimated soil nitrogen supply potential from the nitrogen deficit. net_demand :

[0171] N net_demand =N deficit -SoilN supply ;

[0172] Specifically, when N net_demand A value ≤0 indicates that the soil supply is sufficient or exceeds the deficit, and the recommended nitrogen application rate is 0 kg N / ha.

[0173] Nitrogen fertilizer applied to the soil is not entirely absorbed by crops; therefore, the recommended application rate needs to be calculated based on the estimated nitrogen fertilizer use efficiency (NUE). In this invention, the estimated NUE can be determined by consulting a pre-defined "NUE reference lookup table." This "NUE reference lookup table" is pre-established based on extensive localized experimental data or publicly available literature. The lookup table provides an empirical baseline NUE value based on a combination of several key influencing factors (such as crop growth stage, fertilizer type, application method, soil texture, etc.). For example, the lookup table may specify an NUE value of 40%–50% under the conditions of "rice, jointing stage, urea, deep application."

[0174] Calculate the recommended nitrogen application rate ΔN based on the NUE value obtained from the table. recommend :

[0175] ΔN recommend =N net_demand / NUE;

[0176] Finally, the recommended nitrogen application rate should be adjusted based on short-term weather forecasts (e.g., for the next 3 days). If heavy rainfall is forecast, which may lead to nitrogen leaching, the nitrogen application rate should be reduced accordingly. recommend(e.g., reduce by 10-30%) or it is recommended to apply fertilizer in multiple applications or after rain. If the short-term weather forecast indicates continued high temperatures and drought without irrigation, adjust the timing of fertilization (e.g., in the evening) or fertilize after a small amount of irrigation to reduce volatilization loss and improve absorption efficiency.

[0177] (4) Generation of variable fertilization prescription diagram:

[0178] The recommended nitrogen application rates for all spatial units are collected and stored as raster layers according to spatial location to obtain the variable fertilization prescription map (VRT map).

[0179] Ensure that the VRT map uses a format compatible with mainstream precision agriculture equipment (such as Shapefile or ISOXML) so that it can be directly imported into the variable fertilizer applicator for field fertilization operations.

[0180] The variable fertilizer prescription map output in this step provides direct location-based nitrogen application instructions to the variable fertilizer application equipment. Based on real-time nitrogen deficit assessment, it recommends fertilization only in areas requiring it, precisely matching the degree of deficit, soil supply potential, and crop uptake capacity. By comprehensively considering soil nitrogen supply and expected nitrogen fertilizer utilization efficiency, it reduces ineffective inputs and waste. It avoids over-fertilization and reduces the risk of non-point source pollution (groundwater nitrate pollution, eutrophication, and greenhouse gas emissions) caused by nitrogen leaching and volatilization. Dynamic adjustments based on growth stages and weather forecasts ensure that fertilization strategies better match crop physiological needs and actual field conditions.

[0181] S7, Closed-Loop Learning and Iterative Optimization

[0182] Introducing a feedback mechanism involves optimizing crop growth model parameters, remote sensing inversion algorithms, or fertilization decision logic based on feedback data after applying variable fertilization prescription maps. Specifically, this includes:

[0183] Obtain actual feedback data at the end of the season or during the harvest period from the plots where variable fertilization is applied using the variable fertilization prescription map generated in step S6. This includes crop yield, crop nitrogen uptake data during the harvest period, fertilization cost data, and control data.

[0184] Specifically, actual crop yields are obtained through combine harvester yield monitoring equipment or on-site yield measurements, generating a spatial distribution map of yield reflecting the actual impact of variable fertilization on yield. Harvest-period crop nitrogen uptake data includes plant (or grain) nitrogen concentration and total nitrogen accumulation at harvest. This data is used to calculate actual nitrogen fertilizer use efficiency to assess the resource efficiency of the fertilization strategy. Fertilization cost data is calculated based on fertilizer and operational costs, as well as yield and quality output. Yield and nitrogen uptake data from conventional fertilization treatments or zero-nitrogen treatments within the plots are collected as control data to compare the synergistic effects of variable fertilization.

[0185] Compare the expected management objectives (such as target yield and expected NNI) with the actual feedback results. Use statistical analysis or model simulation diagnostic methods to pinpoint the causes of the discrepancies. Causes of discrepancies include biases in crop growth model parameters, biases in remote sensing inversion algorithms, and biases in fertilization decision-making logic.

[0186] Specifically, biases in crop growth model parameters can cause the generated pCNDC parameters or the biomass or nitrogen concentration simulated by the crop growth model to deviate significantly from the measured values ​​in some regions.

[0187] Bias in the remote sensing inversion algorithm will cause W to be inverted in step S4. actual or N actual There are systematic deviations from ground-measured data (such as underestimation of biomass due to vegetation index saturation).

[0188] Biased fertilization decision-making logic can lead to excessive recommended fertilization amounts in specific soil types (such as sandy soil) or a significant deviation between actual nitrogen fertilizer use efficiency and the estimated nitrogen fertilizer use efficiency.

[0189] When discrepancies are primarily caused by extreme environmental events not fully represented in the model (such as unpredictable prolonged droughts or floods) or large-scale pest and disease outbreaks, the system records the event and its impact area as a "perturbation event" and associates it with the corresponding yield loss data. During subsequent optimization, data from this region can be treated as outliers to prevent the system from incorrectly attributing yield reductions caused by non-fertilization factors to failed fertilization decisions. Furthermore, this accumulated perturbation event data can serve as a basis for future selection or improvement of crop model versions.

[0190] Based on the results of difference analysis and accumulated feedback data, targeted improvements were made to the regional-scale crop nitrogen dynamic regulation method, including optimizing crop growth model parameters, optimizing remote sensing inversion algorithms, and optimizing fertilization decision logic.

[0191] Specifically, optimizing crop growth model parameters involves supplementing the spatiotemporal database with feedback data, prioritizing the recalibration of key parameters that significantly affect nitrogen response identified in the sensitivity analysis of step S2, and improving the adaptability of the crop growth model to local environment and management practices.

[0192] Optimized remote sensing inversion involves comparing the feedback ground-based measured data with the inversion results to optimize the inversion model (such as by adjusting parameters, changing algorithms, and fusing multi-source remote sensing data) and improve inversion accuracy.

[0193] Optimizing fertilization decisions involves adjusting the NNI threshold based on feedback data and dynamically calibrating the NUE lookup table (e.g., if NUE is found to be generally high in sandy soil areas, the corresponding item in the table can be lowered from 40% to 35%) to make the recommended fertilization amount more in line with actual needs and optimal benefits.

[0194] This step dynamically adapts to climate change, the renewal of main cultivated varieties, and the evolution of field management practices by periodically incorporating new data. By promptly correcting for biases in crop growth model parameters, remote sensing inversion algorithms, and fertilization decision-making logic, it suppresses the accumulation of discrepancies, thereby enhancing the reliability and robustness of this method in long-term application. Continuous optimization based on real production feedback forms a self-improving closed-loop mechanism, ensuring continuous improvement in the accuracy and efficiency of nitrogen fertilizer management, and making fertilization recommendations closer to actual optimal levels.

[0195] Example 1

[0196] This embodiment takes the subtropical monsoon climate zone in the lower reaches of the Yangtze River as an example. The main rice varieties cultivated in this target area are japonica rice or hybrid indica rice. The specific process of applying the regional-scale crop nitrogen dynamic regulation method of this invention is as follows:

[0197] S1. Construct a spatiotemporal database:

[0198] Meteorological data acquisition:

[0199] This system integrates daily meteorological data from national and provincial meteorological stations from 1991 to the present, including maximum temperature, minimum temperature, average temperature, precipitation, sunshine duration, solar radiation, wind speed, and relative humidity. A 1km × 1km resolution daily meteorological grid is generated through inverse distance weighting interpolation, and the data is then updated annually by connecting to a real-time meteorological monitoring network.

[0200] Soil data acquisition:

[0201] Soil type, texture (percentage of sand, silt, and clay), bulk density, organic matter content, pH, total nitrogen, available nitrogen, available phosphorus, and available potassium were extracted from soil survey maps or digital soil maps for this region. Paddy soil characteristics, including plow pan depth and redox potential, were also added. A 250m × 250m resolution soil raster layer was generated using conventional kriging.

[0202] Crop growth dynamics data acquisition:

[0203] Field trials lasting 3 to 5 years were conducted in representative locations within the target area, including Nanjing, Jiaxing, and Hefei. The field trials targeted major rice varieties such as the Nanjing and Yongyou series, with nitrogen level gradients of 0 kg / ha, 90 kg / ha, 180 kg / ha, 270 kg / ha, and 360 kg / ha (total nitrogen used throughout the entire growth period).

[0204] We obtained aboveground biomass, biomass of stems, leaves, and spikelets, total nitrogen concentration, nitrogen index (LAI), and final yield and yield components at the tillering, jointing, booting, heading, grain-filling, and maturity stages. Additionally, we supplemented this data with crop growth dynamics data from historical literature for the same region.

[0205] Crop management data acquisition:

[0206] Data on the main rice varieties, sowing dates, planting densities, and irrigation methods in each county and city within the target area were collected as crop management measures.

[0207] Remote sensing data acquisition:

[0208] Sentinel-2L2A imagery data (10m resolution, 5-day revisit period) was acquired during the rice growing season (May to October), and imagery data with cloud cover percentage less than 10% in the target area were selected. After preprocessing including radiometric calibration, atmospheric correction, geometric fine correction, and cloud masking, remote sensing data was obtained.

[0209] Spatiotemporal database construction:

[0210] The above multi-source data is standardized, including unifying the coordinate system, resolution (250m), and time format. It is then integrated into the Geographic Information System (GIS) platform, and spatiotemporal indexes and association rules are established to obtain the constructed spatiotemporal database.

[0211] S2. Selection and Calibration of Crop Growth Models:

[0212] Model selection:

[0213] Based on the quantitative evaluation results in Table 1, this embodiment selects the ORYZA(v3) model as the crop growth model for rice cultivation in the lower reaches of the Yangtze River.

[0214] Model calibration:

[0215] Global sensitivity analysis was conducted using three years of crop growth dynamic data from representative locations to identify key parameters with the greatest impact on nitrogen response simulation. These included parameters related to rice development rate (DRDV, DRDF, DRDR), relative growth rate (RGRLF), light use efficiency (RUE), specific leaf area (SLA), and maximum leaf nitrogen concentration (NMAXLV), as well as parameters related to nitrogen uptake and distribution, such as maximum root nitrogen uptake rate (NUPRMX) and leaf / stem / panicle nitrogen allocation coefficients (FLN, FST, FSO). The PEST tool combined with the NSGA-II multi-objective optimization algorithm was used to calibrate the model for these key parameters. The goal of the model calibration was to minimize the simulated nitrogen response under multiple nitrogen levels.

[0216] Mean absolute error (MAE) between critical reproductive periods and measured values;

[0217] The root mean square error (RMSE), normalized root mean square error (nRMSE), and coefficient of determination R between aboveground biomass (dry weight), plant nitrogen concentration, leaf area index (LAI) and measured values ​​are calculated.2 And the consistency index (d-index);

[0218] RMSE, nRMSE, and R between final output and measured value 2 and d-index.

[0219] The model calibration results are as follows: the MAE for the simulated maturity period is less than 4 days; the nRMSE for the simulated aboveground biomass is 9.5%, and R... 2 The value was 0.91, and the d-index was 0.94; the simulated nRMSE for plant nitrogen concentration was 11.2%, and R... 2 The value was 0.88, and the d-index was 0.92; the nRMSE of the LAI simulation was 13.0%, and R... 2 The value is 0.85, and the d-index is 0.90; the final production simulation nRMSE is 7.8%, and R0.05 is 0.90. 2 The value is 0.90, and the d-index is 0.95.

[0220] Model validation:

[0221] The predictive accuracy of the crop growth model was evaluated using crop growth dynamic data acquired from Jiaxing from 2018 to 2020. Evaluation indicators included aboveground crop biomass, plant nitrogen concentration, final yield, and the simulation effect of key growth stages.

[0222] The pre-set acceptance requirements are as follows: the prediction error (MAE) of the critical growth period is within ±7 days; the nRMSE of the aboveground biomass and LAI predictions is less than 15%; the nRMSE of the plant nitrogen concentration prediction is less than 20%; the nRMSE of the final yield prediction is less than 15%; and the d-index of each indicator is greater than 0.85.

[0223] In this embodiment, the verification results showed that the predicted MAE for maturity was 5.2 days (within the ±7-day threshold); the predicted nRMSE for aboveground biomass was 12.8% (less than 15%), and R... 2 The value was 0.85, and the d-index was 0.91; the predicted nRMSE for plant nitrogen concentration was 15.5% (less than 20%), and R... 2 The value was 0.80, and the d-index was 0.88; the nRMSE predicted by LAI was 14.5% (less than 15%), and R... 2 The d-index is 0.89; the nRMSE for the final yield forecast is 10.5% (less than 15%), and R... 2 The value is 0.86, and the d-index is 0.92. The ORYZA(v3) model meets the preset acceptance requirements.

[0224] S3. Construction of spatialized potential critical nitrogen dilution curve (pCNDC) parameter layer:

[0225] Based on the resolution of meteorological and soil data, the target area was divided into spatial units of 250m×250m. Soil attribute data, meteorological data and localized parameter sets of each spatial unit were input into the ORYZA(v3) model to simulate the rice growth process with a nitrogen supply gradient of 0kg / ha to 400kg / ha (interval of 20kg / ha).

[0226] Simulated data from the tillering peak, jointing stage, booting stage, heading stage, and mid-grain filling stage were analyzed using ANOVA to determine the minimum nitrogen application treatment at the biomass plateau stage, and the biomass (W) of this treatment was recorded. optimal ) and plant nitrogen concentration (N c ).

[0227] Based on each spatial unit (W) optimal N c Data points (W) optimal N in t / ha c (in percentages), fitted N c =a×W optimal -b Equation. Generate a 1km resolution pCNDC parameter a layer and a pCNDC parameter b layer for the target region.

[0228] S4. Dynamic monitoring of crop growth and nitrogen concentration based on remote sensing:

[0229] Based on vegetation indices such as NDVI, EVI, and WDRVI, and measured ground biomass data, a piecewise regression model or machine learning model (such as random forest regression) based on the growth period is constructed. This generates WDRVI data for multiple key growth stages within the rice growing season. actual Spatial distribution map (unit: t / ha).

[0230] Using the red-edge band of Sentinel-2, indices such as NDRE and CI_rededge are calculated. Machine learning methods such as Partial Least Squares Regression (PLSR) or Gradient Boosting Decision Tree (GBDT) are employed, combined with ground-measured plant nitrogen concentration data, to construct an N... actual Estimate the model and generate N actual Spatial distribution map (unit: %).

[0231] S5. Dynamic assessment of nitrogen production profit and loss:

[0232] W obtained actual The spatial distribution map is spatially overlaid and spatially matched with the obtained pCNDC parameter a layer and parameter b layer. Here, we take the calculation at a resolution of 250m as an example.

[0233] Calculate the critical nitrogen concentration (N) unit by unit. c ): N c =a×W actual -b Generate N c Spatial distribution map. For example, a certain unit W. actual =5t / ha, a=5.0, b=0.4, then N c =5.0×5 -0.4 ≈2.63%.

[0234] The obtained N actual Spatial distribution map and N c Spatial distribution maps are spatially overlaid and spatial cell matched. The nitrogen nutrient index is calculated cell-by-cell: NNI = N actual / N c A spatial distribution map of NNI is generated. The map is then displayed in a tiered manner according to the NNI value range: NNI < 0.90 indicates nitrogen deficiency, NNI = 0.90 to 1.05 indicates adequate nitrogen, and NNI > 1.05 indicates excess nitrogen.

[0235] S6. Generation of variable fertilizer prescription diagram:

[0236] Based on the spatial distribution map of nitrogen nitrogen (NNI), a variable fertilizer prescription map is generated by integrating actual crop biomass, crop growth stage, soil nitrogen supply potential estimation, expected nitrogen fertilizer use efficiency, and short-term weather forecasts. Taking topdressing at the early jointing stage as an example, the NNI threshold is set at 0.90.

[0237] Fertilization decisions are triggered when the NNI of a spatial cell is less than 0.90, and the nitrogen deficit is calculated: N deficit =(N c -N actual )×W actual In this embodiment, a certain spatial unit Nc = 2.63%, N actual =2.2%, W actual =5t / ha, then N deficit =(0.0263-0.0220)×5000kg / ha=0.0043×5000kg / ha=21.5kg N / ha.

[0238] Based on soil data, the estimated soil nitrogen supply potential is 15 kg N / ha. Subtracting this potential, the net nitrogen requirement (N) is calculated. net_demand :N net_demand =N deficit -SoilN supply In this embodiment, a certain spatial unit N net_demand =21.5kg / ha-15kg / ha=6.5kg / ha.

[0239] Currently, rice has a high absorption efficiency in the early jointing stage. Using a combination of urea application and broadcasting, and considering the level of field management, the estimated nitrogen fertilizer use efficiency (NUE) is 40%, ΔN recommend =N net_demand / NUE, in this embodiment, a certain spatial unit ΔN recommend =6.5kg N / ha / 0.40=16.25kg N / ha.

[0240] Based on short-term weather forecasts, if heavy rainfall is predicted in the next 3 to 7 days, reduce the amount of fertilizer applied this time (by 10% to 30%) or it is recommended to apply it after the rain. If there is no extreme weather in the next 3 to 7 days, maintain the calculated value.

[0241] A raster map in Shapefile or ISOXML format, i.e., a variable fertilization prescription map (VRT map), is generated based on the recommended nitrogen application rate for each spatial unit.

[0242] The recommended nitrogen application rate is limited to 15 kg N / ha to 60 kg N / ha. If the recommended nitrogen application rate is 16.25 kg N / ha, it should be adjusted to an integer value according to agronomic needs, and upper and lower limits should be set, with a minimum of 15 kg N / ha and a maximum of 60 kg N / ha.

[0243] S7, Closed-Loop Learning and Iterative Optimization:

[0244] Feedback data collection:

[0245] In fields where VRT (Vacuum-Based Tillage) is applied, combine harvesters with yield monitoring capabilities are used to obtain spatial distribution maps of yield. Plant samples are collected from representative sites (based on NNI and VRT maps, selecting areas with different fertilization management levels) to determine plant nitrogen concentration and total nitrogen accumulation at harvest, calculating actual nitrogen fertilizer use efficiency (NUE). Simultaneously, fertilizer and operating costs are collected, and economic benefit data are calculated in conjunction with yield and quality (e.g., protein content).

[0246] Effect evaluation and difference analysis:

[0247] Compare the expected targets (such as target yield, NNI improvement) with the actual feedback results. Analyze the reasons for the discrepancies, which may include: model parameter bias (such as inaccurate ORYZA model parameters), remote sensing inversion errors (large deviations between Wactual or Nactual and actual measurements), flaws in decision-making logic (such as unreasonable NUE estimation leading to systematically high or low recommended fertilization rates, for example, recommended fertilization rates generally exceeding or falling below the actual optimal fertilization rate of 20 kg N / ha), or unforeseen external interference factors such as extreme weather and pests.

[0248] Iterative optimization:

[0249] Optimize crop growth model parameters: supplement the collected data on yield, biomass, nitrogen content, etc. into the spatiotemporal database of S1, and recalibrate or fine-tune the parameters in the ORYZA (v3) model that have a significant impact on nitrogen response.

[0250] Optimize remote sensing inversion algorithms: Improve the inversion models of Wactual and Nactual using ground-based measured data.

[0251] Optimize the fertilization recommendation algorithm: Adjust the NNI trigger threshold, SoilNsupply estimation model, NUE preset value (for example, if the NUE in a certain sandy soil area is found to be generally lower than 40%, it can be lowered to 35%) or coefficients in the recommendation formula in S6 based on the actual fertilization effect and economic benefits.

[0252] This embodiment fully demonstrates the technical process of precision nitrogen management in regional rice by constructing a spatiotemporal database, localized model calibration, spatialized pCNDC generation, remote sensing dynamic monitoring, variable fertilizer prescription generation, and closed-loop optimization. It accurately quantifies regional heterogeneity through multi-scale data processing ranging from 250m to 1km; achieves timely topdressing decisions based on real-time remote sensing data and model diagnostics; reduces nitrogen fertilizer application based on real-time NNI diagnostics; and enhances the robustness of the technical solution for long-term application by driving continuous optimization of crop growth model parameters through yield feedback.

[0253] Example 2

[0254] This embodiment targets the temperate monsoon climate zone of the North China Plain, where winter and spring droughts are frequent, and adopts a winter wheat-summer maize rotation. Steps S1-S5 are consistent with Example 1, with the following differences: soil data are supplemented with soil moisture characteristic parameters; the CERES-Wheat model is selected as the crop growth model (with enhanced water and nitrogen synergistic simulation); the inversion resolution is 20m, adapted to a variable-rate fertilizer applicator. The remaining steps are implemented as follows:

[0255] S6. Generation of variable fertilizer prescription diagram:

[0256] At the early jointing stage, fertilization decisions are triggered when the spatial unit NNI < 0.90. The nitrogen deficit Ndeficit_total is calculated, deducting the soil's current season nitrogen supply potential and the residual nitrogen from the previous season (maize) estimated based on soil data. Combining jointing irrigation and fertilization, good water conditions are beneficial to improving NUE, which, according to recent literature, is 42%. The recommended fertilization rate is adjusted to ΔN = (Ndeficit_total + Nitrogen demand increment from jointing to maturity - Soil supply) / NUE. The fertilization range is set at 30 N / ha to 90 kg N / ha.

[0257] Based on the recommended nitrogen application rates for all spatial units, a variable fertilization prescription map in ISOXML format with a resolution of 20m is generated.

[0258] S7, Closed-Loop Learning and Iterative Optimization:

[0259] Collect yield maps and plant / grain nitrogen content at harvest time, calculate actual nitrogen fertilizer use efficiency and grain protein content; analyze the impact of variable fertilization on yield, protein content and nitrogen fertilizer use efficiency, and optimize CERES-Wheat model parameters, remote sensing inversion model and fertilization recommendation algorithm (adjust NUE preset value and irrigation correlation coefficient based on economic benefit feedback).

[0260] This embodiment demonstrates that the present invention effectively addresses the contradiction between winter and spring drought and efficient nitrogen fertilizer application in wheat production in the North China Plain, as well as the heterogeneity of nitrogen supply caused by straw return to the field and the problem of coordinated regulation of quality and yield. Through water-nitrogen synergistic simulation, precise stress diagnosis, and dynamic decision optimization, this invention effectively solves the problems of nitrogen supply heterogeneity caused by straw return to the field and the problem of coordinated regulation of quality and yield. Compared with existing technologies, this method achieves significant benefits by reducing irrigation water and nitrogen fertilizer application by specific values ​​while ensuring food security.

[0261] The invention and its embodiments have been described above illustratively. This description is not restrictive, and the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. The accompanying drawings are only one embodiment of the invention, and the actual structure is not limited thereto. No reference numerals in the claims should limit the scope of the claims. Therefore, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the invention, such design should fall within the scope of protection of this patent. Furthermore, the word "comprising" does not exclude other elements or steps, and the word "a" preceding an element does not exclude the inclusion of "a plurality" of that element. Multiple elements stated in the product claims may also be implemented by a single element through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

Claims

1. A method for dynamic regulation of crop nitrogen at a regional scale, comprising the following steps: Multi-source data of the target area are acquired to construct a spatiotemporal database. The multi-source data includes meteorological data, soil data, crop growth dynamic data, crop management measures data, and remote sensing data. Select a crop growth model with a nitrogen response module, calibrate and validate the crop growth model based on a spatiotemporal database, obtain a crop growth model that reflects the dynamic response of crop nitrogen in the target area, and generate a localized model parameter set. The target area is divided into spatial units. Soil data, meteorological data and localized model parameter sets of each spatial unit are input into the crop growth model. The crop growth process under nitrogen gradient is simulated under no water stress conditions. The optimal biomass and critical nitrogen concentration data point pairs of each spatial unit are extracted and the spatialized pCNDC parameter layer of the target area is fitted. Based on remote sensing data, the actual biomass and nitrogen concentration spatial distribution map are inverted. The critical nitrogen concentration and nitrogen nutrient index of each spatial unit are dynamically calculated by combining the pCNDC parameter layer and spatial distribution maps are generated respectively. The nitrogen deficit is calculated based on the generated spatial map. The variable fertilizer prescription map is generated by combining the soil nitrogen supply potential and nitrogen fertilizer use efficiency to guide fertilization and realize the dynamic regulation of crop nitrogen at the regional scale.

2. The regional-scale crop nitrogen dynamic regulation method according to claim 1, characterized in that: After fertilization is carried out according to the variable fertilization prescription map, feedback data is collected and added to the spatiotemporal database. The feedback data includes crop yield, crop nitrogen uptake data at harvest, fertilization cost data, and control data. Based on the feedback data, the crop growth model is optimized, the remote sensing data inversion is optimized, and the threshold is adjusted to form a closed-loop control mechanism, thereby achieving continuous optimization of the regional-scale crop nitrogen dynamic control method.

3. The regional-scale crop nitrogen dynamic regulation method according to claim 1, characterized in that: The selection of crop growth models employs a quantitative evaluation system to screen candidate crop growth models. These models include a nitrogen response module and are applicable to the main cultivated varieties in the target region. The quantitative evaluation system includes: The data fit dimension is used to evaluate the degree of matching between the model's input requirements and the spatiotemporal database; Accessibility and scalability are dimensions used to evaluate the openness of the model code and the ease of parameter adjustment; The case studies and literature support dimension is used to count the number of application cases of the model in the target crop type.

4. The regional-scale crop nitrogen dynamic regulation method according to claim 1 or 3, characterized in that: The steps for calibrating the crop growth model include: setting parameters for the crop growth model; The sensitivity of crop growth model parameters to output variables was analyzed using a global sensitivity analysis method, and the key model parameters with the greatest impact on nitrogen response simulation were identified. By utilizing crop growth dynamic data from a spatiotemporal database and employing a multi-objective optimization algorithm to calibrate key model parameters, the overall performance of the crop growth model is optimized across the entire nitrogen response range, from nitrogen deficiency to nitrogen abundance.

5. The regional-scale crop nitrogen dynamic regulation method according to claim 4, characterized in that: In the simulated crop growth process with different nitrogen gradients, the treatments with different nitrogen gradients were divided into the following groups based on the value of biomass: the low biomass group with the smallest value, the high biomass group or plateau group with the largest value, and the rest were medium biomass groups. Within the high biomass group or plateau period group, the treatment with the lowest nitrogen application rate is selected, representing the minimum nitrogen supply required for the crop to reach its maximum growth rate during the current growth stage. The corresponding aboveground biomass is recorded as the optimal biomass, and the plant nitrogen concentration is recorded as the critical nitrogen concentration, thus obtaining the optimal biomass and critical nitrogen concentration data point pairs for the spatial unit.

6. The regional-scale crop nitrogen dynamic regulation method according to claim 5, characterized in that: Based on the data point pairs of optimal biomass and critical nitrogen concentration for each spatial cell, a power function equation for the critical nitrogen dilution curve is fitted to obtain the pCNDC parameters a and b for each spatial cell. The power function equation for the critical nitrogen dilution curve is expressed as: N c =a×W optimal -b ; In the formula, W optimal N represents biomass. c Indicates the critical nitrogen concentration; The parameters a and b of all spatial units are collected and stored as raster layers according to their spatial location, forming a spatialized pCNDC parameter a layer and a spatialized pCNDC parameter b layer, which together constitute a spatialized pCNDC parameter layer covering the target area.

7. The regional-scale crop nitrogen dynamic regulation method according to claim 1, characterized in that: The steps for retrieving the actual biomass and nitrogen concentration spatial distribution map based on remote sensing data include: acquiring remote sensing data covering the key growth stages of crops in a spatiotemporal database, and calculating vegetation indices sensitive to biomass. A biomass inversion model is established based on vegetation index and ground-measured biomass to generate a spatial distribution map of actual biomass covering the target area and corresponding to the time phase of remote sensing data. Based on the spectral features of plants sensitive to nitrogen in remote sensing data, including red edge parameters, chlorophyll index, nitrogen reflectance index or specific band combination spectral information, an actual nitrogen concentration inversion model is established by combining the measured nitrogen concentration of plants on the ground to generate a spatial distribution map of the actual nitrogen concentration corresponding to the time phase of the remote sensing data.

8. The regional-scale crop nitrogen dynamic regulation method according to claim 1 or 7, characterized in that: The actual biomass spatial distribution map is spatially overlaid with the spatialized pCNDC parameter layer, and spatial cell matching is performed to obtain the actual biomass and parameters a and b for each spatial cell. The critical nitrogen concentration of the spatial cell is then calculated and expressed as: N c =a×(W actual ) -b ; In the formula, N c W represents the critical nitrogen concentration. actual Indicates actual biomass; The critical nitrogen concentrations of all spatial units are collected and a spatial distribution map of the critical nitrogen concentrations of the corresponding time phases covering the target area is generated. The actual nitrogen concentration spatial distribution map is spatially overlaid with the critical nitrogen concentration spatial distribution map, and spatial cell matching is performed to obtain the actual plant nitrogen concentration value and critical nitrogen concentration value of each spatial cell. The nitrogen nutrient index of the spatial cell is then calculated and expressed as: NNI=N actual / N c ; In the formula, NNI represents the nitrogen nutritional index, N actual This represents the actual nitrogen concentration value of the plant, N. c Indicates the critical nitrogen concentration; The nitrogen nutrient index of all spatial units is collected to generate a spatial distribution map of the nitrogen nutrient index of the corresponding time phase covering the target area; Based on the spatial distribution map of nitrogen nutrient index, the crop nitrogen nutrient status of each spatial unit is assessed by dynamically preset thresholds. When the nitrogen nutrient index is lower than the threshold, the nitrogen deficit is calculated, and a variable fertilization prescription map is generated by combining soil nitrogen supply potential and nitrogen fertilizer utilization efficiency.

9. The regional-scale crop nitrogen dynamic regulation method according to claim 8, characterized in that: The steps for generating the variable fertilization prescription map include: reading the critical nitrogen concentration, actual nitrogen concentration, and actual biomass of each spatial unit, and calculating the nitrogen deficit, expressed as: N deficit =(N c -N actual )×W actual ; In the formula, N deficit Nitrogen deficit, N c The critical nitrogen concentration for the current biomass, N actual W represents the real-time actual nitrogen concentration of the crop. actual This refers to the actual biomass of the crop in real time. Subtracting the soil's nitrogen supply potential from the nitrogen deficit yields the net nitrogen demand (N). net_demand , is represented as: A net_demand =N deficit -SoilN supply ; Calculate the recommended nitrogen application rate ΔN to estimate nitrogen fertilizer use efficiency. recommend , is represented as: ΔN recommend = N net_demand / NUE; The recommended nitrogen application rates for all spatial units are collected and stored as raster layers according to spatial location to obtain a variable fertilization prescription map.

10. A regional-scale crop nitrogen dynamic regulation device, characterized in that, include: Memory, which stores computer programs; The processor, when executing the computer program, implements the steps of the method as described in any one of claims 1-9; The communication module is used to receive input from external meteorological data servers and remote sensing data platforms.

Images