Time sequence dynamic characteristic evaluation method of corn nitrogen-reducing combined application biochar effect

Abstract

The invention discloses a time sequence dynamic characteristic evaluation method for a corn nitrogen-reducing combined application biochar effect, which comprises the following steps of: firstly, setting different combined application combinations, and collecting indexes such as chlorophyll fluorescence parameters, canopy spectral reflectivity and the like and yield data in key growth periods such as a jointing period and a large bell mouth period of corn; secondly, preprocessing spectral data, calculating a single-temporal red edge feature and constructing a time sequence dynamic feature, screening a biochar sensitive wave band through a reflectivity difference value percentage in a growth period, designing and adjusting a red edge index, and constructing a fluorescence weight spectral index in combination with a broadband vegetation index; then integrating data to construct a multi-temporal spectral feature data set, and constructing a time sequence dynamic chlorophyll inversion model containing dynamic inversion, yield prediction and attribution branches, the former outputting chlorophyll and photosynthetic coding vectors in a filling period, and the latter predicting yield and contribution degrees of related parameters; and finally, training the model, constructing a combined application decision map based on the trained model, obtaining an optimal ratio from the map, and providing support for corn planting fertilization decision.

Description

Technical Field

[0001] This invention relates to a method for evaluating the time-series dynamic characteristics of the effect of nitrogen reduction combined with biochar application in maize, belonging to the field of crop yield assessment technology. Background Technology

[0002] Nitrogen is an essential nutrient for crop growth and development. Rational application of nitrogen fertilizer can effectively promote efficient photosynthesis, thereby significantly increasing crop yield. However, in actual agricultural production, excessive application of nitrogen fertilizer is a common problem, leading to a significant reduction in nitrogen fertilizer utilization and resulting in resource waste.

[0003] Biochar, as a porous carbon-rich material, possesses excellent nitrogen adsorption properties, effectively reducing nitrogen loss and thus improving nitrogen fertilizer utilization efficiency. In recent years, numerous studies have confirmed that the combined application of biochar and nitrogen fertilizer can simultaneously increase crop yield and nitrogen use efficiency, providing a new technical approach to addressing many problems caused by excessive nitrogen fertilizer application and demonstrating promising application prospects in agricultural production.

[0004] Maize is a vital food crop globally, making research on fertilization efficiency enhancement related to maize of significant practical importance. Chlorophyll, as the core pigment in crop photosynthesis, has content and fluorescence parameters that are key indicators reflecting the photosynthetic physiological state of plants, directly related to the crop's light energy conversion efficiency and growth status. Hyperspectral remote sensing technology, with its unique advantages of high resolution and continuous sampling, can achieve non-destructive monitoring of crop chlorophyll content through characteristic spectral information such as "red edges," thereby accurately assessing crop growth status and becoming an important technical means in the field of crop physiological state monitoring.

[0005] Currently, existing research on the combined application of biochar and nitrogen fertilizer mainly focuses on their impact on crop nitrogen use efficiency and yield, while the underlying mechanisms by which they synergistically affect crop photosynthetic physiological processes remain unclear. In particular, how biochar influences crop light energy conversion efficiency by regulating maize chlorophyll fluorescence kinetics and canopy spectral characteristics under different nitrogen application levels still lacks systematic technical means and scientific evaluation methods, specifically exhibiting the following shortcomings: 1. Static monitoring methods and lack of dynamic process research: Most existing studies rely on monitoring data from a single or a few crop growth stages, which can only obtain static snapshot information of crop photosynthetic physiological status. They cannot achieve quantitative analysis of the dynamic changes in key photosynthetic physiological processes (such as chlorophyll accumulation and decline) of maize under nitrogen reduction treatment or biochar application treatment, resulting in an incomplete and inaccurate assessment of the effects of biochar and nitrogen fertilizer application.

[0006] 2. Generalized spectral characteristics, insufficient specificity: While widely used spectral indices (such as the Normalized Difference Vegetation Index, NDVI) show some sensitivity to nitrogen response, their design was not specifically tailored to the "biochar-soil-crop" system. These generalized spectral indices are slow to respond to the synergistic effects of biochar through indirect pathways such as improving soil moisture and promoting crop root growth, making them difficult to effectively identify and limiting their monitoring accuracy and identification capabilities in this specific system.

[0007] 3. Predictive models are often black boxes with weak mechanistic analysis capabilities: Existing studies often directly apply general machine learning models such as random forests and support vector machines for relevant predictive analysis. These models are typical data black boxes. Although they can predict some indicators, they cannot reveal the intrinsic mechanism by which biochar affects crop spectral characteristics and yield by regulating photosynthetic physiological characteristics, thus hindering the optimization and promotion of the technology.

[0008] 4. Data silos and superficial fusion: In existing studies, chlorophyll fluorescence data and hyperspectral data are mostly only subjected to simple correlation analysis in the later stages, lacking early and in-depth fusion at the feature level. This data utilization method fails to fully leverage the unique advantages and potential of multi-source data in revealing the synergistic relationship between crop "function and structure," limiting the in-depth exploration of the regulatory mechanisms of biochar and nitrogen fertilizer application.

[0009] 5. Disconnect between technology application and decision-making, resulting in weak practical guidance: Most existing assessment methods stop at comparing and analyzing core indicators such as crop yield and nitrogen use efficiency, failing to transform complex model predictions into intuitive and easy-to-use agronomic decision-making tools, such as precision fertilization recommendation maps. This leads to a disconnect between related technological research and actual agricultural production decisions, making it difficult to provide effective guidance for agricultural production practices.

[0010] In summary, given the numerous shortcomings in existing research on the combined application of biochar and nitrogen fertilizer in monitoring and elucidating the regulatory mechanisms of maize photosynthetic physiology, there is an urgent need to establish a technical solution that can achieve dynamic monitoring, has a specific system, clarifies the mechanism of action, deeply integrates multi-source data, and effectively supports agronomic decision-making. Summary of the Invention

[0011] The technical problem to be solved by this invention is to provide a time-series dynamic characteristic evaluation method for the effect of nitrogen reduction combined with biochar application in maize, so as to achieve non-destructive, dynamic and interpretable evaluation of the effect of nitrogen reduction combined with biochar application, and provide technical support for precision fertilization.

[0012] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A time-series dynamic characteristic evaluation method for the effect of nitrogen reduction combined with biochar application in maize includes the following steps: Step 1: For the maize planting area to be studied, different combinations of nitrogen fertilizer and biochar are set up to obtain observation index data and maize yield data of maize at each key growth stage under each combination; the observation index includes chlorophyll fluorescence parameters, relative chlorophyll content and canopy spectral reflectance, and the key growth stages include jointing stage, large trumpet stage, tasseling stage and grain filling stage. Step 2: Preprocess the canopy spectral reflectance data, calculate the single-temporal basic features based on the preprocessed canopy spectral reflectance data, including red edge position, red edge amplitude, and red edge area; construct temporal dynamic features based on the single-temporal basic features; Step 3: Calculate the percentage difference in reflectance between two adjacent critical growth stages based on the preprocessed canopy spectral reflectance data. Identify candidate bands sensitive to biochar response based on the percentage difference in reflectance. Superimpose the three candidate bands obtained from the four critical growth stages to obtain biochar-responsive bands. Design the biochar regulation red-edge index for each critical growth stage based on the biochar-responsive bands. Step 4: Select a broadband vegetation index that is sensitive to and stable in terms of canopy structure or chlorophyll as the basic index, and construct fluorescence weighted spectral indices for each key growth stage. Step 5: Based on Steps 1-4, construct a multi-temporal spectral feature dataset and simultaneously construct a time-series dynamic chlorophyll inversion model for yield attribution. The time-series dynamic chlorophyll inversion model includes a dynamic inversion branch and a yield prediction and attribution branch. The dynamic inversion branch takes multi-temporal spectral features as input and the relative chlorophyll content during the grain-filling stage and the photosynthetic dynamic coding vector as output. The yield prediction and attribution branch takes chlorophyll fluorescence parameters during the grain-filling stage, the relative chlorophyll content during the grain-filling stage, and the photosynthetic dynamic coding vector as input, and predicts maize yield and the contribution ratio of relative chlorophyll content to chlorophyll fluorescence parameters as output. Step 6: Train the time-series dynamic chlorophyll inversion model using the multi-temporal spectral feature dataset to obtain a trained model. Use the trained model to construct a nitrogen fertilizer-biochar application decision map and obtain the optimal nitrogen fertilizer-biochar ratio from the nitrogen fertilizer-biochar application decision map.

[0013] Compared with the prior art, the present invention, employing the above technical solution, has the following technical effects: 1. This invention, by extracting time-series dynamic spectral features, has for the first time achieved continuous and quantitative tracking of photosynthetic physiological processes in maize under nitrogen reduction combined with biochar application. This changes the current situation where existing technologies can only obtain static snapshot information, making the evaluation results more comprehensive and the response more sensitive. It provides reliable support for accurately grasping the dynamic influence of biochar and nitrogen fertilizer application on maize photosynthetic physiology.

[0014] 2. This invention abandons the general spectral index and constructs and screens the special spectral index BCREI that is sensitive to the biochar effect. This effectively solves the problem that the general index is not very targeted to the special system of "biochar-improved soil-crop". It significantly improves the ability of spectral features to identify the target treatment combination of "nitrogen reduction + biochar" and lays a high-quality feature foundation for subsequent accurate inversion and prediction.

[0015] 3. This invention innovatively proposes a fluorescence weighted spectral index, which embeds the maximum photochemical efficiency into the structural spectral features in the form of weights. This breaks the isolation of existing technologies where chlorophyll fluorescence and hyperspectral data are only subjected to simple correlation analysis in the later stages. It realizes the deep integration of physiological mechanisms at the feature level of multi-source data and fully leverages the potential of multi-source data to reveal the synergistic relationship between crop "function and structure".

[0016] 4. This invention features a unique dual-branch attribution neural network, which can simultaneously complete the two core tasks of chlorophyll content inversion and yield prediction. Furthermore, through a built-in attention mechanism, it automatically quantifies the contribution ratio of chlorophyll content and photosystem II efficiency to yield, realizing the visualization and analysis of the intrinsic mechanism of "biochar-photosynthetic physiology-yield". This solves the problem that existing general machine learning models cannot reveal the mechanism and provides a clear mechanistic basis for technology optimization.

[0017] 5. This invention generates a decision support map for nitrogen fertilizer-biochar application based on the model output results, transforming complex algorithm analysis results into a clear field operation guide, realizing a closed loop of the technology chain from photosynthetic physiological assessment to agronomic decision-making. It is highly practical and easy to promote and apply in the field.

[0018] 6. This invention provides a precise and reliable technical tool for achieving the goals of reducing carbon and nitrogen and increasing fertilizer efficiency. It can reduce the total amount of nitrogen fertilizer applied and improve nitrogen fertilizer utilization efficiency while ensuring corn yield and quality. It has significant ecological and economic benefits and provides strong support for promoting sustainable agricultural development. Attached Figure Description

[0019] Figure 1 This is the overall framework diagram of the time-series dynamic characteristic evaluation method for the effect of nitrogen reduction and biochar application in maize according to the present invention; Figure 2 This is a flowchart of a time-series dynamic chlorophyll inversion model oriented towards yield attribution; Figure 3 This is a surface graph showing the response of maize yield to the combined application of nitrogen fertilizer and biochar. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0021] like Figure 1 As shown, this invention provides a time-series dynamic characteristic evaluation method for the effect of nitrogen reduction combined with biochar application in maize based on chlorophyll fluorescence and canopy hyperspectral characteristics. The specific steps are as follows: Step 1: Field trial design. A completely randomized design was adopted, setting up different combinations of nitrogen application rates and biochar application to conduct a controlled field trial. Observation indicators included: leaf rapid chlorophyll fluorescence parameters, relative chlorophyll content, canopy hyperspectral data, and maize yield.

[0022] 1. Chlorophyll fluorescence parameter measurement: During key growth stages of maize, such as the jointing stage, the large trumpet stage, the tasseling stage, and the grain-filling stage, measurements were taken on sunny mornings from 10:00 to 14:00. Dark adaptation treatment: The experimental field was divided into several small areas. Five representative plants were selected from each small area. The middle part of the third fully unfolded leaf at the top of the canopy was clamped with a dark adaptation clip and allowed to dark adapt for 20 minutes. Fluorescence measurement: Measurements were performed using a chlorophyll fluorometer (model Yaxin-1161G). The probe was placed close to the leaf, and the saturated pulse light intensity was set to [value missing]. The pulse duration is 1 second. Parameter acquisition: Directly read or calculate the following fluorescence parameters: initial fluorescence Maximum fluorescence Maximum photochemical efficiency ,in, .

[0023] 2. Measurement of relative chlorophyll content (SPAD): A handheld chlorophyll meter (SPAD-502) was used. Five plants were selected from each small area. For each plant, measurements were taken from the upper, middle, and lower parts of the third fully expanded leaf at the top of the canopy. The average value was taken as the SPAD value of that leaf, and then the average value of the small area was calculated.

[0024] 3. Canopy hyperspectral measurement: Performed simultaneously with fluorescence measurement, and completed between 10:00 and 14:00 in clear, windless or light-wind weather; Instrument preparation: A portable ground object spectrometer (ASD Field Spec4) with a spectral range of 350-2500 nm was used. White plate calibration was performed before measurement. Measurement method: The probe is held vertically downward, about 20 cm from the top of the canopy. Each small area is measured 10 times consecutively, and the average value is taken as the canopy spectral reflectance of that small area. Band selection: Since the absorption of water is significantly affected after 1350nm, this invention selects the 350-1350nm band for analysis.

[0025] Step 2, data preprocessing and feature extraction, aims to extract dynamic information sensitive to the "nitrogen reduction combined with biochar application" treatment from multi-temporal data and construct targeted time-series dynamic features. For each experimental sub-region, the four key growth stages—jointing stage, large trumpet stage, tasseling stage, and grain-filling stage—are considered as a time series, and features reflecting dynamic changes in physiological state are extracted: 1. Canopy spectral reflectance data preprocessing: Spectral smoothing and stitching correction were performed using ViewSpec 6.0 software; 2. Calculate basic characteristics for a single time phase: Calculate the first derivative of the spectral data for each growth stage and extract the traditional red-edge parameters: (1) Red border position The wavelength corresponding to the maximum value of the first-order differential spectrum in the range of 680-760nm; (2) Red edge amplitude : The first-order differential value at; (3) Area of ​​the red edge The integral area of ​​the first-order differential curve in the range of 680-760nm.

[0026] 3. Constructing time-series dynamic features: Based on the aforementioned red-edge parameters, calculate their dynamic change indicators: (1) Change rate characteristics: Calculate the difference in red edge amplitude between adjacent fertility periods to reflect the rate of change: , in, For the first A critical reproductive period.

[0027] (2) Cumulative effect characteristics: From the jointing stage to the target growth stage, the weighted cumulative sum of the red edge area is calculated to reflect the persistence of the effect: , in, For the first Weighting coefficients for key reproductive periods.

[0028] The weights were determined using the following method: based on the local standard maize growth period days and typical leaf area index dynamics, combined with historical average sunshine hours, a simplified light capture model was used. , in, The extinction coefficient of the maize canopy is taken as an empirical value of 0.65; Typical leaf area index for each growth stage is used, and typical values ​​for each growth stage are taken.

[0029] Step 3: Construct and screen specific spectral indices for "nitrogen reduction combined with biochar application". To enhance the sensitivity and discrimination of target treatment combinations (especially the biochar effect), this invention designs and screens specific spectral indices: 1. Sensitive band identification: Compare and analyze the average canopy spectral reflectance curves of two adjacent key growth stages during the yield formation period, and calculate the percentage difference in reflectance: , in, For the first Each key reproductive period is at the same wavelength The pre-processed canopy spectral reflectance. Identification. The band intervals exhibiting significant peak values ​​were selected as candidate bands sensitive to biochar response. Three candidate bands obtained from four key growth stages were superimposed to obtain the biochar-responsive bands. to .

[0030] 2. Design of Dedicated Spectral Index: Based on the identified biochar response-sensitive bands and combined with the principles of vegetation index construction, a novel index is designed, proposing a biochar-modified red-edge index: , in, This is an empirical adjustment factor (initial recommended value is 0.2-0.5). This index is used in the enhanced red-edge region ( While being sensitive to chlorophyll, the background spectral variation caused by biochar itself or the soil moisture conditions it improves is suppressed by introducing a near-infrared band ratio (950 / 870).

[0031] Step 4, Construction of fluorescence weighted spectral index To achieve a deep fusion at the characteristic level between the photosynthetic functional state characterized by chlorophyll fluorescence and the canopy structure / pigment information reflected by hyperspectral imaging, a fluorescence-weighted spectral index is constructed: 1. Selection of base index: Select a classic broadband vegetation index that is sensitive to and stable in terms of canopy structure or chlorophyll as the base, such as the Normalized Difference Vegetation Index (NDVI).

[0032] 2. Fluorescence weighting calculation: based on the average grouting period of all sub-regions. Value as the benchmark Calculate the relative fluorescence efficiency offset for each small region: , Through an empirical adjustment coefficient (The recommended range is 0.3-0.8) Convert it into a weighting factor. : .

[0033] 3. Index Generation: Multiply the base index by the fluorescence weighting factor to generate the fluorescence weighted spectral index. : , This index allows canopies with the same NDVI value but higher photosynthetic efficiency to obtain higher characteristic values, thus providing a more refined characterization of their health status.

[0034] Step 5: Combine the temporal dynamic features extracted in Step 2, the dedicated spectral indices selected in Step 3, and the fluorescence weighted spectral indices constructed in Step 4 with the single-phase basic features, canopy spectral reflectance, and measured chlorophyll fluorescence parameters from Step 2. The features were then merged. Subsequently, the Z-score normalization method was used to normalize all features, ultimately forming a multi-dimensional, temporal, and deeply integrated multi-temporal spectral feature dataset of physiological function information, which served as the input for subsequent model construction.

[0035] A time-series dynamic chlorophyll inversion model oriented towards yield attribution was constructed. This model is a two-branch hybrid neural network model, including a dynamic inversion branch and a yield prediction and attribution branch. It can not only accurately invert chlorophyll content during key growth periods, but also simultaneously analyze the quantitative relationship between chlorophyll dynamics and final yield. Figure 2 This is the model flowchart.

[0036] Branch 1: The dynamic inversion branch is a time series model based on long short-term memory network (LSTM). The input is a multi-temporal spectral feature sequence arranged in the order of the reproductive period. This branch learns the dynamic pattern of the evolution of spectral features with the reproductive period and outputs two key results: (1) the continuous inversion value of chlorophyll content (SPAD) during the grain filling period, and (2) a comprehensive "photosynthetic dynamic encoding vector", which condenses the photosynthetic physiological dynamic information of the entire reproductive period.

[0037] To enhance the physiological interpretability of the model, a "virtual photosynthetic accumulation layer" was designed after the LSTM. This layer performs a weighted summation of intermediate features from each growth stage in the LSTM output to simulate the accumulation process of photosynthetic products. , in, For the first The hidden state vectors of the LSTM for each reproductive stage. The multi-temporal spectral feature sequence is input into the LSTM. The LSTM processes the features of each reproductive stage sequentially and outputs the hidden state vector corresponding to each reproductive stage. This vector contains accumulated information from the current period and previous periods; weights Instead of being fixed, it is dynamically generated by a small neural network based on the input features themselves. This simulates the differences in the contribution of each growth stage to the final biomass under different treatments (such as water shortage and nitrogen deficiency). This design allows the model to adaptively learn how "nitrogen reduction" or "biochar" changes the distribution and accumulation rhythm of photosynthetic products.

[0038] Branch Two: Yield Prediction and Attribution. This branch is a multilayer perceptron (MLP) module combined with an attention mechanism. The input comes from two outputs of Branch One: the inverted SPAD value during the grain-filling stage and the "photosynthetic dynamics encoding vector," combined with measured chlorophyll fluorescence parameters during the grain-filling stage. The goal of this branch is to predict the final yield and use the built-in attention weight vector. The system automatically learns and quantifies the relative contribution of chlorophyll relative content (SPAD) and chlorophyll fluorescence parameters to yield formation. .

[0039] Step 6: The model training is guided by a joint loss function: , Among them, the equilibrium hyperparameters and The method is determined by combining grid search with k-fold cross-validation, with the goal of minimizing the combined error between chlorophyll inversion and yield prediction on the independent validation set. To account for chlorophyll inversion loss, the mean square error between the SPAD inversion values ​​of each growth stage output by branch one and the measured SPAD values ​​is calculated, with particular emphasis on the weight of the inversion accuracy during the grain-filling stage. To account for the yield prediction loss, the mean squared error between the predicted yield value output by branch two and the measured yield is calculated. The training process is as follows: First, branch one is pre-trained using a subset of data (samples with SPAD measurements for all fertile periods) to optimize the yield prediction. Then, by fixing a portion of the weights in branch one, the entire network is jointly trained and optimized. This step-by-step training strategy ensures the reliability of the model in the fundamental task of chlorophyll inversion.

[0040] For any field (with known treatment), by inputting its multi-temporal spectral characteristics, the model can directly output the predicted SPAD value and predicted yield during the grain-filling stage. The attention weights of the model (branch two) directly provide the "inverted SPAD" and "measured yield". "Importance score in yield forecasting (e.g., SPAD contributes 65%)" (Contributing 35%). This quantitatively reveals whether, under specific treatments, yield increases are primarily achieved through increased chlorophyll or optimized light energy conversion efficiency. The SPAD inversion sequences for each growth stage provided in Branch 1 can be plotted as chlorophyll dynamic curves, visually demonstrating how different treatments affect the establishment and decline of chlorophyll.

[0041] Based on the model, a system simulation was conducted to construct a decision support graph for nitrogen fertilizer-biochar application. The steps are as follows: (1) Establish a grid with nitrogen application rate and biochar application rate as coordinate axes; (2) At each grid point, the SPAD value and relative yield during the grouting period are predicted using the model; (3) Draw equal SPAD lines and equal yield lines on the graph, and mark the high-yield and high-efficiency areas. Figure 3 To predict the results, the nitrogen fertilizer-biochar ratio corresponding to the highest yield is the optimal ratio, which is found to be 224 kg / ha of nitrogen fertilizer and 0.2 t / ha of biochar.

[0042] Based on the same inventive concept, this application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the aforementioned method for evaluating the time-series dynamic characteristics of the effect of nitrogen reduction and biochar application on corn.

[0043] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned method for evaluating the time-series dynamic characteristics of the effect of nitrogen reduction and biochar application on corn.

[0044] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0045] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0046] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0047] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0048] The above embodiments are merely illustrative of the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solutions based on the technical concept proposed in this invention shall fall within the scope of protection of this invention.

Claims

1. A method for evaluating the time-series dynamic characteristics of the effect of nitrogen reduction combined with biochar application in maize, characterized in that, Includes the following steps: Step 1: For the maize planting area to be studied, different combinations of nitrogen fertilizer and biochar are set up to obtain observation index data and maize yield data of maize at each key growth stage under each combination; the observation index includes chlorophyll fluorescence parameters, relative chlorophyll content and canopy spectral reflectance, and the key growth stages include jointing stage, large trumpet stage, tasseling stage and grain filling stage. Step 2: Preprocess the canopy spectral reflectance data, calculate the single-temporal basic features based on the preprocessed canopy spectral reflectance data, including red edge position, red edge amplitude, and red edge area; construct temporal dynamic features based on the single-temporal basic features; Step 3: Calculate the percentage difference in reflectance between two adjacent critical growth stages based on the preprocessed canopy spectral reflectance data. Identify candidate bands sensitive to biochar response based on the percentage difference in reflectance. Superimpose the three candidate bands obtained from the four critical growth stages to obtain biochar response-sensitive bands. Design of biochar-regulated red-edge index for each key growth stage based on biochar response-sensitive bands; Step 4: Select a broadband vegetation index that is sensitive to and stable in terms of canopy structure or chlorophyll as the basic index, and construct fluorescence weighted spectral indices for each key growth stage. Step 5: Based on Steps 1-4, construct a multi-temporal spectral feature dataset and simultaneously construct a time-series dynamic chlorophyll inversion model for yield attribution. The time-series dynamic chlorophyll inversion model includes a dynamic inversion branch and a yield prediction and attribution branch. The dynamic inversion branch takes multi-temporal spectral features as input and the relative chlorophyll content during the grain-filling stage and the photosynthetic dynamic coding vector as output. The yield prediction and attribution branch takes chlorophyll fluorescence parameters during the grain-filling stage, the relative chlorophyll content during the grain-filling stage, and the photosynthetic dynamic coding vector as input, and predicts maize yield and the contribution ratio of relative chlorophyll content to chlorophyll fluorescence parameters as output. Step 6: Train the time-series dynamic chlorophyll inversion model using the multi-temporal spectral feature dataset to obtain a trained model. Use the trained model to construct a nitrogen fertilizer-biochar application decision map and obtain the optimal nitrogen fertilizer-biochar ratio from the nitrogen fertilizer-biochar application decision map.

2. The method for evaluating the time-series dynamic characteristics of the nitrogen reduction effect of maize combined with biochar application according to claim 1, characterized in that, In step 1, the corn planting area to be studied is divided into several small areas of the same size. For each application combination, the observation index data of corn in each small area at each critical growth stage are obtained. For chlorophyll fluorescence parameters, five uniformly growing maize plants were randomly selected from each small area. The middle portion of the third fully expanded leaf from the top of the canopy of each plant was clamped using a dark adaptation clip. After 20 minutes of dark adaptation, the initial and maximum fluorescence were measured using a chlorophyll fluorometer, and the maximum photochemical efficiency of each maize plant was calculated. The average of the maximum photochemical efficiencies of the five plants was taken as the maximum photochemical efficiency of each small area. The formula for calculating the maximum photochemical efficiency of each maize plant is as follows: ， in, To achieve maximum photochemical efficiency, For maximum fluorescence, Initial fluorescence; For the relative chlorophyll content, five uniformly growing maize plants were randomly selected from each small area. The relative chlorophyll content of the upper, middle and lower parts of the third fully expanded leaf at the top of the canopy of each maize plant was measured using a handheld chlorophyll meter. The average of the relative chlorophyll content of the three parts was taken as the relative chlorophyll content of each maize plant. The average of the relative chlorophyll content of the five maize plants was taken as the relative chlorophyll content of each small area. For canopy spectral reflectance, five uniformly growing maize plants were randomly selected from each small area. The canopy spectral reflectance of each maize plant at a distance of 20 cm from the top of the canopy was measured using a portable ground spectrometer. The average value of the canopy spectral reflectance in the 350-1350 nm band was taken as the canopy spectral reflectance of each maize plant. The average value of the canopy spectral reflectance of the five maize plants was taken as the canopy spectral reflectance of each small area.

3. The method for evaluating the time-series dynamic characteristics of the nitrogen reduction and biochar application effect in maize according to claim 1, characterized in that, In step 2, the first-order differential spectrum is calculated for the canopy spectral reflectance after pretreatment at each key growth period in each small region. The red edge position is the wavelength corresponding to the maximum value of the first-order differential spectrum in the range of 680-760nm. The red edge amplitude is the first-order differential value at the red edge position. The red edge area is the integral area of ​​the first-order differential curve in the range of 680-760nm. The time-series dynamic characteristics include rate of change characteristics and cumulative effect characteristics. Among them, the rate of change characteristic is the difference between the red-edge positions of adjacent critical growth periods, and the formula is as follows: ， in, Characteristic of rate of change The first A key reproductive period The first The red-bordered position of a key reproductive period; The cumulative effect is characterized by the weighted cumulative sum of the red-edge area from the jointing stage to the current critical growth stage, as shown in the following formula: ， ， in, Characterized by cumulative effect. For the first The red-edge area of ​​a key reproductive period For the first Weighting coefficients for key reproductive periods This refers to the number of children from the jointing stage to the current critical reproductive stage. The extinction coefficient of the maize canopy. For the first Typical leaf area index for each critical growth stage.

4. The method for evaluating the time-series dynamic characteristics of the nitrogen reduction and biochar application effect in maize according to claim 1, characterized in that, The specific process of step 3 is as follows: Step 3.1: Calculate the percentage difference in reflectance between two adjacent critical growth stages based on the preprocessed canopy spectral reflectance, using the following formula: ， in, The percentage difference in reflectance. The first Each key reproductive period is at the same wavelength Canopy spectral reflectance after pretreatment; Identification The band intervals exhibiting significant peak values ​​were selected as candidate bands sensitive to biochar response. The three candidate bands obtained from the four key growth stages were superimposed to obtain the biochar-responsive bands. to ,in, These are the minimum and maximum wavelengths of the biochar response sensitive band, respectively; Step 3.2: Design the biochar regulation red-edge index for each key growth stage based on the biochar response-sensitive band. The formula is: ， in, To adjust the red edge index of biochar, This is an empirical adjustment coefficient. Wavelengths for each critical reproductive period The corresponding pre-processed canopy spectral reflectance, The pretreated canopy spectral reflectance corresponds to a wavelength of 870 nm for each key growth stage. The pre-treated canopy spectral reflectance corresponds to a wavelength of 950 nm for each key reproductive period.

5. The method for evaluating the time-series dynamic characteristics of the nitrogen reduction effect of maize combined with biochar application according to claim 2, characterized in that, The specific process of step 4 is as follows: Step 4.1: Select a broadband vegetation index that is sensitive to and stable for canopy structure or chlorophyll, i.e., select the normalized difference vegetation index. As a basic index; Step 4.2: For each critical growth stage, the average of the maximum photochemical efficiencies during the grain-filling period of all small areas is used as the benchmark. Calculate the relative fluorescence efficiency offset for each small region. : ， Through empirical adjustment coefficients Will Convert to fluorescence weighting factor : ， Step 4.3: Multiply the base index by the fluorescence weighting factor to generate the fluorescence weighted spectral index. : 。 6. The method for evaluating the time-series dynamic characteristics of the nitrogen reduction effect of maize combined with biochar application according to claim 2, characterized in that, The specific process of step 5 is as follows: One sample in the multi-temporal spectral feature dataset corresponds to a small region. Each sample includes data on the jointing stage, the large trumpet stage, the tasseling stage, and the grain-filling stage, as well as maize yield data. Data for each key growth stage includes: chlorophyll fluorescence parameters, canopy spectral reflectance, single-temporal basic features, temporal dynamic features, biochar-regulated red edge index, and fluorescence weighted spectral index. The dynamic inversion branch includes a long short-term memory network and a virtual photosynthetic accumulation layer connected to the long short-term memory network. The virtual photosynthetic accumulation layer is used to perform weighted summation of the intermediate features of each key reproductive stage output by the long short-term memory network to obtain the photosynthetic dynamic encoding vector. : ， in, For the first The weight of key reproductive periods For the first Intermediate characteristics of a key reproductive period; The yield prediction and attribution branch is a multilayer perceptron incorporating an attention mechanism, which outputs the predicted maize yield and the contribution ratio of relative chlorophyll content and chlorophyll fluorescence parameters to maize yield.

7. The method for evaluating the time-series dynamic characteristics of the nitrogen reduction effect of maize combined with biochar application according to claim 2, characterized in that, The specific process of step 6 is as follows: The time-series dynamic chlorophyll inversion model is trained using a joint loss function. as follows: ， in, For chlorophyll inversion loss, Forecast losses for production output. All are balanced hyperparameters; A grid was established with nitrogen fertilizer and biochar application rates as coordinate axes. At each grid point, the trained model was used to predict the relative chlorophyll content and maize yield during the grain-filling period. Isochlorophyll content lines and isoyield lines were drawn, and the nitrogen fertilizer-biochar ratio corresponding to the highest yield point was taken as the optimal ratio.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the time-series dynamic characteristic evaluation method for the effect of nitrogen reduction and biochar application on maize as described in any one of claims 1 to 7.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the time-series dynamic characteristic evaluation method for the effect of nitrogen reduction and biochar application on maize as described in any one of claims 1 to 7.

Images