An analysis method for effect of straw returning to field combining soil nutrient and crop yield

Abstract

The present application relates to the field of agricultural data analysis, and more particularly to a method for analyzing the effect of straw returning to field in combination with soil nutrients and crop yield, which comprises: collecting a multi-source feature data matrix for risk assessment; obtaining a static nitrogen deficiency risk amplification factor by jointly analyzing crop yield data and soil attribute data; obtaining a dynamic growth response asymmetric adjustment factor by performing difference analysis on vegetation index data, yield data and cluster centroid; obtaining a risk perception asymmetric distance measure based on yield data by fusing and evaluating the static nitrogen deficiency risk amplification factor and the dynamic growth response asymmetric adjustment factor; and obtaining optimal management zoning results for variable fertilization strategy formulation by partition optimization of the clustering process of the risk perception asymmetric distance measure based on yield data, so as to reduce the nitrogen deficiency risk of crops in the seedling stage, improve the nutrient utilization efficiency and yield stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural data analysis, and particularly relates to a method for analyzing the effect of straw returning to field by combining soil nutrients and crop yield. BACKGROUND

[0002] In the field of precision agriculture, in order to achieve fine management of farmland, it is usually necessary to divide the farmland into several management zones with similar production characteristics. A commonly used technical method to achieve this purpose is to use clustering algorithms to process historical crop yield data. The algorithm can automatically divide the entire farmland into different regions, such as high yield area, medium yield area and low yield area, according to the numerical value of crop yield. These divided management zones provide a data basis for subsequent development of differentiated field operation strategies such as variable fertilization and variable seeding. In conventional applications, areas with higher historical yield values are usually considered to have better soil conditions and higher production potential, so they will be allocated more agricultural resources such as fertilizer and irrigation water in subsequent production management. The clustering algorithm assigns each data point to the cluster that is closest to it in numerical value through an iterative process using Euclidean distance. In agricultural production, straw returning to field is a common technical measure to improve soil fertility. However, when the clustering algorithm based on standard Euclidean distance is directly applied to historical yield data of farmland implementing straw returning to field, the division results of the management zones generated will deviate significantly from the actual production potential of the farmland plots in the next production cycle, thereby affecting the effectiveness of subsequent field management decisions. Specifically, when processing data of farmland implementing straw returning to field, the limitation of the standard Euclidean distance is that it only relies on the numerical value of the data points when calculating the difference between two yield data points, and cannot take into account factors associated with the yield value, i.e. the high nitrogen competition effect caused by the high amount of straw returned to field at the beginning of the next crop growing season. The higher the yield value of a plot in the farmland implementing straw returning to field, the more straw it returns to the field. Because the carbon-nitrogen ratio of straw is much higher than that of soil microorganisms, when microorganisms decompose these high-carbon input straws, they need to absorb a large amount of available nitrogen from the soil environment to meet their own reproductive needs, which will compete with crop seedlings for soil available nitrogen, resulting in nitrogen deficiency of crops during the growing period. The standard Euclidean distance metric function cannot reflect this relationship, as it will consider a data point with high yield and high nitrogen competition risk to be the same distance from the centroid of a high yield cluster as another data point with similar yield value but lower risk. For example, a plot with a high historical yield value (Y1) corresponds to a large amount of straw returned to the field, and has the highest nitrogen competition risk at the beginning of the next crop growing season; while another plot with a similar high historical yield value (Y2) has a slightly lower nitrogen competition risk. The clustering algorithm will assign both plots to the same high yield cluster, and the management zone division results generated by the algorithm will not reflect the difference in nitrogen competition risk between the two plots. When the clustering algorithm is directly applied to historical yield data of farmland implementing straw returning to field, the division results of the management zones generated will deviate significantly from the actual production potential of the farmland plots in the next production cycle, thereby affecting the effectiveness of subsequent field management decisions. Specifically, when processing data of farmland implementing straw returning to field, the limitation of the standard Euclidean distance is that it only relies on the numerical value of the data points when calculating the difference between two yield data points, and cannot take into account factors associated with the yield value, i.e. the high nitrogen competition effect caused by the high amount of straw returned to field at the beginning of the next crop growing season. The higher the yield value of a plot in the farmland implementing straw returning to field, the more straw it returns to the field. Because the carbon-nitrogen ratio of straw is much higher than that of soil microorganisms, when microorganisms decompose these high-carbon input straws, they need to absorb a large amount of available nitrogen from the soil environment to meet their own reproductive needs, which will compete with crop seedlings for soil available nitrogen, resulting in nitrogen deficiency of crops during the growing period. The standard Euclidean distance metric function cannot reflect this relationship, as it will consider a data point with high yield and high nitrogen competition risk to be the same distance from the centroid of a high yield cluster as another data point with similar yield value but lower risk. For example, a plot with a high historical yield value (Y1) corresponds to a large amount of straw returned to the field, and has the highest nitrogen competition risk at the beginning of the next crop growing season; while another plot with a similar high historical yield value (Y2) has a slightly lower nitrogen competition risk. The clustering algorithm will assign both plots to the same high yield cluster, and the management zone division results generated by the algorithm will not reflect the difference in nitrogen competition risk between the two plots. ​​) have relatively low nitrogen competition risk. The standard Euclidean distance would consider that the two plots have the same distance to the same high yield cluster centroid (e.g. ), the algorithm would divide a plot with the highest nitrogen competition risk at crop seedling stage in the next season due to extremely high yield in the last season into the same high yield management zone as a plot with relatively low nitrogen competition risk due to similar yield value. Thus, the fertilization decision made according to the division result cannot target the plot with the highest nitrogen competition risk, resulting in the growth of crops on the plot being hindered and yield being lost. SUMMARY

[0003] Therefore, the present application aims to provide a straw returning effect analysis method combining soil nutrients and crop yield, so as to solve the problem that the existing clustering method cannot identify the potential nitrogen competition risk of high yield plots, resulting in mismatch of fertilization strategy.

[0004] To achieve the above-mentioned purpose, the technical scheme of the present application is as follows:

[0005] A straw returning effect analysis method combining soil nutrients and crop yield, comprising the following steps:

[0006] Step S1: acquiring a multi-source feature data matrix for risk assessment by collecting and gridding crop yield, soil available nitrogen, organic matter and vegetation index data;

[0007] Step S2: acquiring a static nitrogen deficiency risk amplification factor by jointly analyzing crop yield data and soil attribute data;

[0008] Step S3: acquiring a dynamic growth response asymmetry adjustment factor by difference analysis of vegetation index data, yield data and cluster centroid;

[0009] Step S4: acquiring a risk perception asymmetry distance measure based on yield data by fusion evaluation of the static nitrogen deficiency risk amplification factor and the dynamic growth response asymmetry adjustment factor;

[0010] Step S5: acquiring an optimal management zone division result for variable fertilization strategy formulation by partition optimization of the clustering process of the risk perception asymmetry distance measure based on yield data.

[0011] Further, the step of acquiring a multi-source feature data matrix for risk assessment by collecting and gridding crop yield, soil available nitrogen, organic matter and vegetation index data comprises:

[0012] Crop yield data corresponding to each grid unit in the farmland is collected during the crop harvesting period by a combine harvester equipped with crop yield monitoring sensors and a positioning system, and a crop yield matrix is formed, wherein each element in the crop yield matrix is the yield of the corresponding grid unit;

[0013] Soil available nitrogen content data and soil organic matter content data are obtained by grid sampling of the soil in the farmland and analyzing the soil samples;

[0014] Soil available nitrogen content data and soil organic matter content data are obtained by grid sampling of the soil in the farmland and analyzing the soil samples;

[0015] Crop seedling stage normalized vegetation index data is obtained by remote sensing image collection of the farmland by a multi-spectral sensor-equipped unmanned aerial vehicle during the crop seedling stage, and based on image calculation.

[0016] Further, the static nitrogen deficit risk amplification factor is obtained by joint analysis of the crop yield data and the soil attribute data, including:

[0017] The effective nitrogen deficit ratio of the grid unit is obtained by collaborative analysis of the crop yield data and the soil attribute data, and the static nitrogen deficit risk amplification factor is obtained by mapping transformation of the effective nitrogen deficit ratio.

[0018] Further, the effective nitrogen deficit ratio of the grid unit is obtained by collaborative analysis of the crop yield data and the soil attribute data, including:

[0019] The median of the soil available nitrogen content data of all grid units in the farmland is taken as the half-saturation constant of soil available nitrogen effect; for any target grid unit in the farmland, the soil available nitrogen content of the target grid unit is taken as the numerator, the calculation result of adding the soil available nitrogen content of the target grid unit and the half-saturation constant of soil available nitrogen effect is taken as the denominator, and the corresponding fraction is taken as the first adjustment factor of the target grid unit;

[0020] The calculation result of multiplying the soil organic matter content of the target grid unit and the first adjustment factor of the target grid unit is taken as the effective contribution evaluation of the organic matter of the target grid unit;

[0021] A biological conversion coefficient is set; the calculation result of multiplying the biological conversion coefficient and the last period crop yield data of the target grid unit is taken as the numerator, the calculation result of adding the soil available nitrogen content of the target grid unit and the effective contribution evaluation of the organic matter of the target grid unit is taken as the denominator, and the corresponding fraction is taken as the effective nitrogen deficit ratio of the target grid unit.

[0022] Further, the static nitrogen deficiency risk amplification factor is obtained by mapping the effective nitrogen deficiency ratio, comprising:

[0023] The maximum effective nitrogen deficiency ratio in all grid units of the farmland is obtained; for any target grid unit in the farmland, the effective nitrogen deficiency ratio of the target grid unit is divided by the maximum effective nitrogen deficiency ratio, and the calculation result is added by constant 1 as the static nitrogen deficiency risk amplification factor of the target grid unit.

[0024] Further, the dynamic growth response asymmetric adjustment factor is obtained by difference analysis on the vegetation index data, yield data and cluster centroid, comprising:

[0025] The crop performance divergence index is obtained by quantile difference evaluation on the normalized vegetation index data and yield centroid data; the dynamic growth response asymmetric adjustment factor is obtained by normalization processing on the crop performance divergence.

[0026] Further, the crop performance divergence index is obtained by quantile difference evaluation on the normalized vegetation index data and yield centroid data, comprising:

[0027] All cluster centroid is obtained in the clustering process, for any target cluster centroid, the yield quantile rank of the target cluster centroid is obtained by all cluster centroid; for any target grid unit in all grid units of the farmland, the vegetation index quantile rank of the target grid unit is obtained by the normalized vegetation index of all grid units;

[0028] For any target cluster centroid and any target grid unit, the calculation result of subtracting the yield quantile rank of the target cluster centroid from the vegetation index quantile rank of the target grid unit is taken as the first vigor difference evaluation; the calculation result of subtracting the yield data of the target cluster centroid from the yield data of the target grid unit is taken as the numerator, the absolute value of the calculation result of subtracting the yield data of the target cluster centroid from the yield data of the target grid unit is taken as the denominator, and the corresponding fraction is taken as the first vigor difference direction evaluation;

[0029] The calculation result of multiplying the first vigor difference evaluation and the first vigor difference direction evaluation is taken as the crop performance divergence index.

[0030] Further, the dynamic growth response asymmetric adjustment factor is obtained by normalization processing on the crop performance divergence, comprising:

[0031] The crop performance divergence index is normalized by the maximum value and the minimum value, and the calculation result of adding constant 1 to the corresponding normalized processing result is taken as the dynamic growth response asymmetric adjustment factor.

[0032] Further, the risk perception asymmetric distance metric based on yield data is obtained by fusing the static nitrogen deficiency risk amplification factor and the dynamic growth response asymmetric adjustment factor, including:

[0033] In the clustering process, for any target grid cell and any target cluster centroid, the calculation result of the product of the static nitrogen deficiency risk amplification factor and the dynamic growth response asymmetric adjustment factor of the target grid cell is taken as a first distance optimization weight; and the calculation result of the product of the first distance optimization weight and the Euclidean distance between the target grid cell and the target cluster centroid is taken as the risk perception asymmetric distance metric based on yield data between the target grid cell and the target cluster centroid.

[0034] Further, the optimal management zone division result for variable fertilization strategy formulation is obtained by partition optimization of the clustering process based on the risk perception asymmetric distance metric based on yield data, including:

[0035] The number of clusters in the clustering process is set, the clustering process is completed by taking the risk perception asymmetric distance metric based on yield data as the distance between data points and cluster centroids, and the optimal management zone division result for variable fertilization strategy formulation is obtained.

[0036] Compared with the prior art, the present application has the following advantages:

[0037] The straw returning effect analysis method combining soil nutrients and crop yield can effectively identify the soil nitrogen competition risk caused by high yield by constructing a static nitrogen deficiency risk amplification factor and a dynamic growth response asymmetric adjustment factor, and dynamically corrects the real growth performance of the crop at the seedling stage, breaks the symmetry assumption of risk judgment in the traditional yield clustering method, and thus more accurately divides the farmland management zones. Compared with the existing clustering method based on Euclidean distance, the risk perception asymmetric distance metric mechanism introduced in the present application can significantly improve the identification ability of high-risk plots in the clustering process, and avoid misclassifying risk plots into resource priority allocation zones due to similar yield values.

[0038] In actual agricultural application scenarios, the method can identify those plots with high yield records but insufficient soil buffering capacity and outstanding nitrogen competition risk in farmland environments implementing straw returning, and compensate for differential fertilization, thereby reducing the nitrogen deficiency risk of crops at the seedling stage and improving nutrient utilization efficiency and yield stability. The present application establishes a multi-source data fusion evaluation path from historical yield-soil properties-crop growth performance, provides more robust decision support for agricultural intelligent management, and is conducive to promoting the landing of green and efficient fertilization modes. BRIEF DESCRIPTION OF DRAWINGS

[0039] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The embodiments of the application illustrated in the drawings are presented by way of example or for purposes of illustration and not limitation.

[0040] Figure 1 A method flowchart of a straw returning effect analysis method combining soil nutrients and crop yield according to an embodiment of the present application. DETAILED DESCRIPTION

[0041] The present application will be described in detail below with reference to the accompanying drawings and embodiments.

[0042] Reference Figure 1 A method flowchart of a straw returning effect analysis method combining soil nutrients and crop yield according to an embodiment of the present application is shown in FIG. 1, which can include: Figure 1

[0043] Step S1, through collecting and gridding crop yield, soil available nitrogen, organic matter and vegetation index data, a multi-source feature data matrix for risk assessment is obtained.

[0044] In an embodiment, first, data collection of four grid unit data of farmland is completed, including:

[0045] The farmland is divided into grid units, and the crop yield data corresponding to the farmland grid units is collected by the combine harvester equipped with crop yield monitoring sensors and positioning systems during crop harvesting, corresponding to form a crop yield matrix, each element in the crop yield matrix being the yield of the corresponding grid unit.

[0046] Through gridding soil sampling in the farmland, soil available nitrogen content data and soil organic matter content data are obtained by analyzing the soil samples; through the soil available nitrogen content data and soil organic matter content data, a soil available nitrogen content data matrix and a soil organic matter content data matrix are obtained.

[0047] Through the unmanned aerial vehicle equipped with multispectral sensors, farmland remote sensing image collection is carried out in the crop seedling stage, and normalized vegetation index data is obtained based on image calculation, corresponding to form a crop seedling normalized vegetation index matrix.

[0048] At this point, through collecting and gridding crop yield, soil available nitrogen, organic matter and vegetation index data, a multi-source feature data matrix for risk assessment is obtained.

[0049] Step S2, through joint analysis of crop yield data and soil attribute data, a static nitrogen deficiency risk amplification factor is obtained. ​

[0050] In the farmland where the straw is returned to the field, the crop yield data of the grid unit in the previous season has a direct positive correlation with the nitrogen competition risk faced by the grid unit at the beginning of the next season crop growth, because high yield corresponds to a large amount of straw returned to the soil, and the straw as a high carbon-nitrogen ratio organic material requires soil microorganisms to consume a large amount of soil available nitrogen during the decomposition process, which causes the competition between microorganisms and crop seedlings for nitrogen. In the traditional clustering process, the yield difference between grid units is evaluated by the Euclidean distance, but the traditional Euclidean distance only processes the yield data itself between the target grid units when calculating, and cannot take into account the nitrogen competition risk caused by high yield. Therefore, in order to make the distance measurement in the clustering process reflect the nitrogen competition risk, it is necessary to evaluate the nitrogen competition risk of the grid unit by the carbon source input represented by the returned straw and the nitrogen supply and buffer capacity of the soil itself, so as to adjust the Euclidean distance in the clustering process by the nitrogen competition risk evaluation of the grid unit.

[0051] In summary, the present application obtains the effective nitrogen deficit ratio of the grid unit by synergistic analysis of crop yield data and soil attribute data; then continues to obtain the static nitrogen deficit risk amplification factor by mapping transformation of the effective nitrogen deficit ratio. In order to make the distance measurement reflect this risk, the present application constructs the static nitrogen deficit risk amplification factor and uses it to amplify the calculation result of the standard Euclidean distance. When the carbon source input of a plot far exceeds its soil nitrogen supply and buffer capacity, its nitrogen competition risk is high, and the value of the static nitrogen deficit risk amplification factor is also large, so that a large amplification penalty effect is produced in the process of distance measurement; on the contrary, when the soil nitrogen supply and buffer capacity can cope with the carbon source input, the risk is low, and the value of the static nitrogen deficit risk amplification factor should tend to 1, that is, no or very small amplification penalty effect is produced. By adding this static nitrogen deficit risk amplification factor as a multiplication item to the distance measurement process, the optimized distance measurement can quantify and take into account the potential growth risk caused by straw returning to the field.

[0052] To realize the above logic, first, the effective nitrogen deficit ratio of the grid unit is obtained by synergistic analysis of crop yield data and soil attribute data, specifically,

[0053] The median of the available nitrogen content in all grid cells of the farmland is used as the half-saturation constant of available nitrogen action. For any target grid cell in the farmland, the available nitrogen content in the target grid cell is used as the numerator, and the result of adding the available nitrogen content in the target grid cell to the half-saturation constant of available nitrogen action is used as the denominator. The resulting fraction is used as the first adjustment factor of the target grid cell. The result of multiplying the soil organic matter content in the target grid cell by the first adjustment factor is used as the assessment of the effective contribution of organic matter in the target grid cell.

[0054] A bioconversion coefficient is set. The bioconversion coefficient is a coefficient used for dimensional calibration. When calculating relative risk, its specific value does not affect the final normalization result. In this embodiment of the invention, the bioconversion coefficient is set to 1. This coefficient can be adjusted according to the specific scenario and is not required. The result of multiplying the bioconversion coefficient by the crop yield data of the previous cycle of the target grid unit is used as the numerator. The result of adding the available nitrogen content of the soil of the target grid unit to the effective contribution assessment of the organic matter of the target grid unit is used as the denominator. The corresponding fraction is used as the effective nitrogen deficit ratio of the target grid unit.

[0055] In one embodiment, it is assumed that the bioconversion coefficient is ;No. The crop yield data for each grid cell is as follows: ;No. The available nitrogen content in the soil of each grid cell is ;No. The soil organic matter content of each grid cell is The half-saturation constant of available nitrogen in soil is: Then the first The formula for calculating the effective nitrogen deficit ratio of a grid cell is:

[0056]

[0057] in, Indicates the first The effective nitrogen deficit ratio of each grid cell; Indicates the bioconversion coefficient; Indicates the first Crop yield data for each grid cell; Indicates the first Soil available nitrogen content in each grid cell; Indicates the first Soil organic matter content of each grid cell; The half-saturation constant represents the availability of nitrogen in the soil.

[0058] After the effective nitrogen deficit ratio of the target grid cell is obtained, the static nitrogen deficit risk amplification factor is obtained by mapping and transforming the effective nitrogen deficit ratio. Specifically, the maximum effective nitrogen deficit ratio in all grid cells of the farmland is obtained; for any target grid cell in the farmland, the calculation result of dividing the effective nitrogen deficit ratio of the target grid cell by the maximum effective nitrogen deficit ratio and then adding a constant 1 is taken as the static nitrogen deficit risk amplification factor of the target grid cell.

[0059] In an embodiment, assuming that the maximum effective nitrogen deficit ratio in all grid cells of the farmland is , then the static nitrogen deficit risk amplification factor of the i-th grid cell is calculated as follows:

[0060]

[0061] wherein, represents the static nitrogen deficit risk amplification factor of the i-th grid cell; represents the effective nitrogen deficit ratio of the i-th grid cell; represents the maximum effective nitrogen deficit ratio in all grid cells of the farmland.

[0062] ​​​It should be noted that the construction of the effective nitrogen deficit ratio is directly aimed at the two basic elements of the risk of nitrogen competition, namely the characteristic evaluation of nitrogen demand and effective nitrogen supply, and the imbalance of the two elements is directly reflected through the form of fraction. For the numerator part of the effective nitrogen deficit ratio calculation formula, the biological conversion coefficient is multiplied by the crop yield data of the target grid unit in the last period to quantify the nitrogen demand. In the scenario of straw returning to field, the nitrogen demand factor that triggers the nitrogen competition cannot be directly measured. According to the basic principles of agronomy, the total nitrogen demand produced by microbial decomposition of straw is proportional to the total amount of straw returned to the field (i.e. the total amount of carbon source), and for a specific crop type and growth environment, the total amount of straw replaced has a stable positive correlation with the crop yield of the last season. Therefore, the crop yield, which can be directly measured, is an effective basis for indicating the relative strength of the unmeasurable nitrogen demand. Based on this, the present application uses the biological conversion coefficient multiplied by the crop yield data of the grid unit in the last period to identify the nitrogen demand. The biological conversion coefficient here theoretically represents all proportional relationships between crop yield and microbial nitrogen demand, including complex factors such as crop harvest index, carbon effectiveness in straw, carbon assimilation efficiency of microorganisms, etc. However, the ultimate purpose of this step is not to calculate the absolute value of nitrogen demand of each plot, but to determine the relative height of nitrogen demand intensity among different plots. In the calculation process of the static nitrogen deficit risk amplification factor, the biological conversion coefficient applied to all data points is removed during the normalization process of dividing the effective nitrogen deficit ratio by the maximum effective nitrogen deficit ratio. Therefore, the specific value of the biological conversion coefficient will not affect the final calculated static nitrogen deficit risk amplification factor. The denominator part of the effective nitrogen deficit ratio is used to quantify the effective nitrogen supply. The total amount of nitrogen that can be provided by the soil to resist microbial nitrogen demand is not equal to its available nitrogen content. Soil organic matter will also slowly mineralize and release nitrogen under the action of microorganisms to form another part of the nitrogen supply. However, the contribution of organic matter is significantly affected by the level of available nitrogen in the soil. When available nitrogen is sufficient, microbial activity is high, organic matter decomposition is fast, and the contribution is high. Conversely, the contribution is small. In order to consider this synergistic and restrictive relationship, an effective total amount of nitrogen supply that is more consistent with soil ecology is obtained. The part in the expression is used to quantify the effective contribution of organic matter, and the adjustment factor The value varies between 0 and 1, and the adjustment factor dynamically couples the soil organic matter content with the level of soil available nitrogen. When the soil available nitrogen content of the target grid cell is low, the adjustment factor tends to 0, and the effective contribution of soil organic matter is correspondingly reduced. When the soil available nitrogen content is high, the adjustment factor approaches 1, and the effective contribution of soil organic matter approaches its base amount. The calculation result of the addition of the soil available nitrogen content of the grid cell and the effective contribution of organic matter of the target grid cell is taken as the evaluation of the effective nitrogen supply. In the calculation expression of the static nitrogen deficit risk amplification factor of the subsequent grid cell, the risk index is normalized by dividing the effective nitrogen deficit ratio of each grid cell by the highest effective nitrogen deficit ratio among all grid cells of the farmland, mapping all risk index values to the interval of 0 to 1. Then, by adding 1, the value range of the final static nitrogen deficit risk amplification factor is set to between 1 and 2. This design ensures that the static nitrogen deficit risk amplification factor is an amplification coefficient greater than or equal to 1, the lowest risk point does not produce amplification effect, the highest risk point produces the maximum amplification effect, and the amplification degree is determined by the relative size of the risk index reflected by the effective nitrogen deficit ratio.

[0063] At this point, the static nitrogen deficit risk amplification factor is obtained by jointly analyzing the crop yield data and soil attribute data.

[0064] Step S3, a dynamic growth response asymmetry adjustment factor is obtained by difference analysis of the vegetation index data, yield data, and cluster centroid.

[0065] The static nitrogen deficit risk amplification factor constructed in step S2 solves the problem of quantifying the nitrogen competition risk of each plot. However, in the process of calculating the static nitrogen deficit risk amplification factor, the influence of the nitrogen competition risk on the yield of the crop is not considered. The standard Euclidean distance in the distance calculation of clustering has symmetry, which is reflected in that it only measures the size of the numerical difference between the data points and the cluster centroid, and cannot distinguish the relative position relationship of the difference. For example, for a high yield cluster, a point with a yield value slightly higher than its centroid and a point with a yield value slightly lower than its centroid, as long as the absolute value of the difference is equal, the standard Euclidean distance will consider that their distances from the cluster are the same. In the straw return to field agricultural production scene, this symmetrical judgment is insufficient, a grid cell with high yield and high risk, when it is compared with a high yield potential cluster, its high risk negative attribute needs to be taken into account; and when it is compared with a low yield potential cluster, it is consistent with the risk expectation to be classified into the cluster. In addition, the risk assessment in step S2 based on pre-sowing static data does not include the dynamic influence of actual climate conditions and field management on crop growth in the year, and the actual growth of crop seedlings is a true reflection of the combined effect of static risk and dynamic environment. Therefore, in order to solve the symmetry problem of the standard Euclidean distance and introduce the actual performance of the early growth of crops as a dynamic feedback, a dynamic growth response asymmetric adjustment factor related to both the grid cell and the centroid of the cluster compared with it needs to be constructed. The construction logic of the factor is that its numerical size should reflect the mismatch between the actual growth performance of the grid cell and the expected performance of the centroid of the cluster compared with it. The greater the mismatch, the greater the value of the dynamic growth response asymmetric adjustment factor, so as to secondarily adjust the amplification effect of the static nitrogen deficiency risk amplification factor, and realize a non-symmetrical distance correction based on the actual growth feedback of crops.

[0066] In summary, the present application obtains the crop performance dispersion index by performing quantile difference evaluation on the normalized vegetation index data and the yield centroid data; and then continues to obtain the dynamic growth response asymmetric adjustment factor by normalizing the crop performance dispersion.

[0067] First, crop performance divergence index is obtained by evaluating the quantile difference between normalized vegetation index (NZVI) data and yield centroid data. Specifically, during the clustering process, all cluster centroids are obtained. For any target cluster centroid, the yield quantile rank of the target cluster centroid is obtained from all cluster centroids. For any target grid cell in all grid cells in the farmland, the vegetation index quantile rank of the target grid cell is obtained from the normalized vegetation index of all grid cells. For any target cluster centroid and any target grid cell, the result of subtracting the yield quantile rank of the target cluster centroid from the vegetation index quantile rank of the target grid cell is used as the first growth vigor difference assessment. The result of subtracting the yield data of the target grid cell from the yield data of the target cluster centroid is used as the numerator, and the absolute value of the result of subtracting the yield data of the target grid cell from the yield data of the target cluster centroid is used as the denominator. The resulting fraction is used as the first growth vigor difference direction assessment. The result of multiplying the first growth vigor difference assessment and the first growth vigor difference direction assessment is used as the crop performance divergence index.

[0068] In one implementation, it is assumed that during the clustering process, the first... The centroid of each cluster is ;No. The normalized vegetation index of each grid cell is ; The quantile function is Then the first The grid cell relative to the first The formula for calculating the divergence index of crop performance for each cluster is as follows:

[0069]

[0070] in, Indicates the first The grid cell relative to the first Crop performance divergence index for each cluster; Indicates the first step in the clustering process The centroid of each cluster class; Indicates the first Normalized vegetation index of each grid cell; Indicates the first Crop yield data for each grid cell.

[0071] After obtaining the crop performance divergence index, the crop performance divergence is further normalized to obtain the dynamic growth response asymmetric adjustment factor. Specifically, the crop performance divergence index is normalized by the maximum and minimum values, and the result of adding the constant 1 to the corresponding normalization result is used as the dynamic growth response asymmetric adjustment factor.

[0072] In an embodiment, let the set of crop performance divergence indicators of all crops be , then the calculation expression of the dynamic growth response asymmetric adjustment factor of the th grid cell relative to the th cluster class is:

[0073]

[0074] , wherein, represents the dynamic growth response asymmetric adjustment factor of the th grid cell relative to the th cluster class; represents the crop performance divergence indicator of the th grid cell relative to the th cluster class; represents the set of crop performance divergence indicators of all crops; represents the maximum function; represents the minimum function.

[0075] It should be noted that the present application aims to solve the symmetry problem inherent in the existing distance measurement method in step S3, and dynamically corrects the static risk assessment by introducing the actual performance data of the early growth of crops. In order to solve this problem, the present application obtains the crop performance divergence , and maps the expected mismatch into a numerical value reflecting the mismatch direction and degree.

[0076] The expression of the dynamic growth response asymmetric adjustment factor is composed of two parts, the first part , which acts to quantify the gap between the expected growth and the actual growth, thereby introducing dynamic feedback. In a farmland, the production potential of a grid cell (represented by the centroid of its belonging cluster ) and the growth performance of the crop in the seedling stage (embodied by ) should be related. In order to compare these two different types of data on a unified scale, the present application uses quantile function for transformation. represents the expected growth level corresponding to the th cluster class, and represents the actual growth level of the grid cell. The difference between the two quantifies the gap, when the difference is positive, it means that the actual growth of the th grid cell does not reach the expected level corresponding to its potential; when the difference is negative, it means that the actual growth of the th grid cell exceeds the expected level. This design introduces the dynamic information of the early growth of crops into the model, and corrects the static risk assessment obtained in step S2. The second part Its function is to determine the direction of the difference, thereby introducing asymmetry to assess the source of the gap between growth and expectations. For example, a grid cell with a yield higher than the centroid of a high-yielding cluster needs to be closely monitored for poor growth; while a point with a yield lower than the centroid of a high-yielding cluster may be considered normal for poor growth. The result of this expression is approximately +1 (when...). (when) or -1 (when) (Time), it is used to determine the first The yield of each grid cell is at its centroid of comparison. Is it the high-yield side or the low-yield side? This is determined by multiplying the first part (the magnitude of the difference) by the second part (the direction). Ultimately, this allows us to characterize an asymmetric mismatch information that is relevant to the comparison object. For example, when a point with a yield higher than the centroid of a high-yielding cluster (+1 in the second part) has poor actual growth (positive in the first part), A large positive number indicates the most critical mismatch. After constructing the crop performance divergence index, the final step is to evaluate the moderating factors. This step is used to convert divergence values ​​that have positive and negative values ​​and different ranges. This is converted into a multiplier factor that can be directly used to adjust the static nitrogen deficit risk amplification factor. By employing a maximum-minimum normalization method, all divergence values ​​are mapped to... The range is then determined by adding 1, setting the final range of the dynamic growth response asymmetric adjustment factor to be within a certain interval. Between these values, the most severe mismatch corresponds to the largest adjustment factor of 2, while the best match corresponds to the smallest adjustment factor of 1 (i.e., no additional adjustment is made).

[0077] Thus, the asymmetric regulatory factors of dynamic growth response were obtained by differential analysis of vegetation index data, yield data, and cluster centroids.

[0078] Step S4: By fusing and evaluating the static nitrogen deficit risk amplification factor and the dynamic growth response asymmetric adjustment factor, a risk perception asymmetric distance metric based on yield data is obtained.

[0079] After obtaining the static nitrogen deficiency risk amplification factor and the dynamic growth response asymmetric adjustment factor, the risk perception asymmetric distance metric based on yield data is obtained by fusion evaluation of the static nitrogen deficiency risk amplification factor and the dynamic growth response asymmetric adjustment factor. Specifically, in the clustering process, for any target grid cell and any target cluster centroid, the calculation result of the product of the static nitrogen deficiency risk amplification factor and the dynamic growth response asymmetric adjustment factor of the target grid cell is taken as the first distance optimization weight; and the calculation result of the product of the first distance optimization weight and the Euclidean distance between the target grid cell and the target cluster centroid is taken as the risk perception asymmetric distance metric between the target grid cell and the target cluster centroid based on yield data.

[0080] In an embodiment, the calculation expression of the risk perception asymmetric distance between the first grid cell and the second cluster centroid is:

[0081]

[0082]

[0083] At this point, the risk perception asymmetric distance metric based on yield data is obtained by fusion evaluation of the static nitrogen deficiency risk amplification factor and the dynamic growth response asymmetric adjustment factor.

[0084] Step S5: The optimal management zoning result for variable fertilization strategy formulation is obtained by partition optimization of the clustering process of the risk perception asymmetric distance metric based on yield data.

[0085] The number of clusters in the clustering process is set. In the embodiment of the present application, the number of clusters is set to a range of 1 to 10 by the method of a range of candidate cluster numbers. For each cluster number value, the risk perception asymmetric distance metric between the grid cell and the cluster centroid based on yield data is optimized by the method of the range of candidate cluster numbers. ​​​​​​​​​​​​​​​The clustering process is completed, and the average silhouette coefficient of each cluster is evaluated by the risk-aware asymmetric distance metric based on yield data. The number of clusters with the highest average silhouette coefficient is selected as the optimal number of clusters, and the optimal management zone division result for variable fertilization strategy is obtained.

[0086] After determining the number of clusters in the clustering process, the clustering process is completed by evaluating the distance between data points and cluster centroids based on the risk-aware asymmetric distance metric based on yield data, and the optimal management zone division result for variable fertilization strategy is obtained.

[0087] After the clustering algorithm is executed, the optimal clustering result is obtained, and the variable fertilization strategy is formulated based on the optimal clustering result. First, the characteristics of each cluster are analyzed to determine its intrinsic properties. For any management zone in the optimal management zone division result, the average value of each key indicator of all grid cells in the management zone is calculated to determine the agronomic attribute characteristics of the cluster. The specific characteristic indicators include: the average yield level of all grid cells in the management zone, and the arithmetic mean of the static nitrogen deficiency risk amplification factors of all grid cells constituting the management zone. Through the above calculation, the comprehensive characteristics of the grid cell type represented by each management zone are determined.

[0088] Then, a specific fertilization scheme is matched for each characteristic management zone, and the strategy is formulated based on logical judgment of the characteristics of the management zone. In an embodiment, the strategy is formulated as follows:

[0089] For a management zone whose characteristics are analyzed as high average yield level and low average static risk level, it indicates that the management zone represents a true high-yield potential area. The historical yield of these grid cells is high, and the soil fertility is good and the buffering capacity is strong. Therefore, for all grid cells in the management zone, a standard high-yield fertilization quota is uniformly adopted.

[0090] For a management zone whose characteristics are analyzed as high average yield level and high average static risk level, it indicates that the management zone represents a high-risk high-yield potential area. The clustering algorithm of the present application has successfully identified and separated these grid cells with high nitrogen competition risk due to high yield. The historical yield of these grid cells is high, but the soil fertility buffering capacity is poor, and the nitrogen competition is extremely sensitive. Therefore, for all grid cells in the management zone, a risk compensation fertilization quota is adopted, which increases the nitrogen fertilizer amount by a certain amount based on the standard high-yield fertilization quota to compensate for the nitrogen deficiency caused by microbial fixation.

[0091] For a management zone which is analyzed to have a medium average yield level, it is indicated that the management zone represents an average level area in the farmland. Therefore, a standard medium yield fertilizer rate is applied to all grid cells in the management zone.

[0092] For a management zone which is analyzed to have a low average yield level, it is indicated that the management zone represents a low yield potential area. Its upper limit of production potential is low regardless of its risk. Therefore, a conservative low yield fertilizer rate is applied to all grid cells in the management zone to avoid ineffective investment in low potential areas.

[0093] So far, the zoning optimization by clustering process of risk perception asymmetric distance measurement based on yield data is completed, and the optimal management zoning result for variable fertilization strategy formulation is obtained.

[0094] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for analyzing the effect of straw returning to the field on the combination of soil nutrients and crop yield, characterized in that, The method includes the following steps: Step S1: By collecting and gridding data on crop yield, available nitrogen in soil, organic matter and vegetation index, a multi-source feature data matrix for risk assessment is obtained; Step S2: Obtain the static nitrogen deficit risk amplification factor by jointly analyzing crop yield data and soil property data; Step S3: By performing differential analysis on vegetation index data, yield data, and cluster centroids, the asymmetric regulation factor of dynamic growth response is obtained; Step S4: By fusing and evaluating the static nitrogen deficit risk amplification factor and the dynamic growth response asymmetric adjustment factor, a risk perception asymmetric distance metric based on yield data is obtained. Step S5: Optimize the clustering process based on the risk perception asymmetric distance metric of yield data to obtain the optimal management area division result for variable fertilization strategy formulation; The method of obtaining a static nitrogen deficit risk amplification factor by jointly analyzing crop yield data and soil attribute data includes: obtaining the effective nitrogen deficit ratio of grid cells by conducting collaborative analysis of crop yield data and soil attribute data; and obtaining the static nitrogen deficit risk amplification factor by mapping and transforming the effective nitrogen deficit ratio. The method of obtaining asymmetric regulatory factors for dynamic growth response by performing differential analysis on vegetation index data, yield data, and cluster centroids includes: obtaining crop performance divergence index by evaluating the quantile difference between normalized vegetation index data and yield centroid data; and obtaining asymmetric regulatory factors for dynamic growth response by normalizing the crop performance divergence.

2. The method for analyzing the effect of straw returning to field according to claim 1, wherein, The process involves collecting and gridding data on crop yield, available soil nitrogen, organic matter, and vegetation index to obtain a multi-source feature data matrix for risk assessment, including: During the crop harvesting period, a combine harvester equipped with crop yield monitoring sensors and a positioning system collects crop yield data corresponding to farmland grid units, forming a crop yield matrix. Each element in the crop yield matrix represents the yield of the corresponding grid unit. By conducting gridded soil sampling in farmland and analyzing the soil samples, data on available nitrogen content and soil organic matter content were obtained. The soil available nitrogen content data matrix and the soil organic matter content data matrix are obtained by corresponding the soil available nitrogen content data and the soil organic matter content data. By using drones equipped with multispectral sensors to collect remote sensing images of farmland during the crop seedling stage, and calculating normalized vegetation index data based on the images, a normalized vegetation index matrix for the crop seedling stage is formed.

3. The method for analyzing the effect of straw returning to field according to claim 1, wherein, The method of obtaining the effective nitrogen deficit ratio of grid cells through collaborative analysis of crop yield data and soil property data includes: The median of the available nitrogen content in the soil of all grid cells in the farmland is used as the half-saturation constant of available nitrogen action. For any target grid cell in the farmland, the available nitrogen content in the target grid cell is used as the numerator, and the result of adding the available nitrogen content in the target grid cell to the half-saturation constant of available nitrogen action is used as the denominator. The corresponding fraction is used as the first adjustment factor of the target grid cell. multiplying the soil organic matter content of the target grid cell by the first adjustment factor of the target grid cell to obtain an organic matter effective contribution assessment of the target grid cell; setting a biological conversion coefficient, multiplying the biological conversion coefficient by the crop yield data of the target grid cell in the last period to obtain a numerator, adding the soil available nitrogen content of the target grid cell to the organic matter effective contribution assessment of the target grid cell to obtain a denominator, and obtaining a corresponding fraction as the effective nitrogen deficit ratio of the target grid cell.

4. The method for analyzing the effect of straw returning to field according to claim 1, wherein, The static nitrogen deficit risk amplification factor is obtained by mapping and transforming the effective nitrogen deficit ratio, including: obtaining the maximum effective nitrogen deficit ratio in all grid cells of the farmland; for any target grid cell in the farmland, dividing the effective nitrogen deficit ratio of the target grid cell by the maximum effective nitrogen deficit ratio and adding the constant 1 to obtain the static nitrogen deficit risk amplification factor of the target grid cell.

5. The method for analyzing the effect of straw returning to field according to claim 1, wherein, The crop performance divergence index is obtained by performing quantile difference assessment on the normalized vegetation index data and the yield centroid data, including: obtaining all cluster centroids in the clustering process, for any target cluster centroid, obtaining the yield quantile rank of the target cluster centroid through all cluster centroids; for any target grid cell in all grid cells of the farmland, obtaining the vegetation index quantile rank of the target grid cell through the normalized vegetation index of all grid cells; for any target cluster centroid and any target grid cell, subtracting the yield quantile rank of the target cluster centroid from the vegetation index quantile rank of the target grid cell to obtain a first vigor difference assessment; subtracting the yield data of the target cluster centroid from the yield data of the target grid cell to obtain a numerator, and taking the absolute value of the yield data of the target grid cell minus the yield data of the target cluster centroid as a denominator, and obtaining a corresponding fraction as a first vigor difference direction assessment; multiplying the first vigor difference assessment and the first vigor difference direction assessment to obtain the crop performance divergence index.

6. The method for analyzing the effect of straw returning to field according to claim 1, wherein, The dynamic growth response asymmetric adjustment factor is obtained by normalizing the crop performance divergence, including: normalizing the crop performance divergence index by the maximum value and the minimum value, and adding the constant 1 to the corresponding normalized processing result to obtain the dynamic growth response asymmetric adjustment factor.

7. The method for analyzing the effect of straw returning to field according to claim 1, wherein, The risk perception asymmetric distance measure based on yield data is obtained by fusing the static nitrogen deficit risk amplification factor and the dynamic growth response asymmetric adjustment factor, including: in the clustering process, for any target grid cell and any target cluster centroid, multiplying the static nitrogen deficit risk amplification factor of the target grid cell by the dynamic growth response asymmetric adjustment factor to obtain a first distance optimization weight; multiplying the first distance optimization weight by the Euclidean distance between the target grid cell and the target cluster centroid to obtain the risk perception asymmetric distance measure based on yield data between the target grid cell and the target cluster centroid.

8. The method for analyzing the effect of straw returning to field according to claim 1, wherein, The optimal management zone division result for variable fertilization strategy formulation is obtained by partition optimization of the clustering process based on the risk perception asymmetric distance metric of yield data, comprising: The number of cluster classes of the clustering process is set, the clustering process is evaluated and completed by taking the risk perception asymmetric distance metric based on yield data as the distance between data points and cluster class centroids, and the optimal management zone division result for variable fertilization strategy formulation is obtained.

Images