A farmland drought evaluation method and system based on nematode food web monitoring

Abstract

The present application provides a kind of farmland drought assessment method and system based on nematode food web monitoring, it is related to drought assessment technical field, the present application obtains the sample of different irrigation strategies under nematode community characteristics and actual drought grade in test field, obtains actual control set;Threshold value is set based on each nematode community characteristics drought grade judgment rule, combine the nematode community characteristics of sample constitute drought judgment set, the threshold value of each nematode community characteristics is used as optimization parameter to generate individual vector, with initial population;Based on actual control set and drought judgment set generates optimization index, based on the difference of the risk distribution of the edge of each nematode community characteristics in actual control set relative to its corresponding threshold value, improve genetic algorithm, based on the improved genetic algorithm to obtain optimal individual vector, as final farmland drought assessment rule, consider the optimization parameter of the difference of the risk distribution of the edge, the judgment accuracy of different drought grades and the stability of boundary sample can be considered.

Description

Technical Field

[0001] This invention relates to the field of drought assessment technology, specifically to a method and system for assessing farmland drought based on nematode food web monitoring. Background Technology

[0002] In farmland water-saving irrigation management and drought early warning, it is necessary to classify and assess the degree of water stress suffered by crops during the growing season under different fields and irrigation strategies, so as to adjust the irrigation plan in a timely manner and minimize crop yield loss at maturity. In addition to monitoring based on crop phenotype and soil physicochemical environment, soil biological communities have the characteristics of rapid response to water changes and can characterize changes in soil ecological processes. Therefore, in agricultural ecological monitoring, nematode functional groups are often used to characterize the state of soil food web.

[0003] In the implementation process, it is not necessarily required to completely depict all trophic levels. If a complete food web is depicted, it will make nematode identification and index calculation complicated, making it difficult to meet the efficiency and feasibility of farmland dynamic monitoring. Therefore, the key functional groups of decomposition channels can be used as food web characteristics. For example, changes in bacteriophages and fungiophages, as well as channel indices reflecting the relative dominance of bacterial and fungal channels, can be used to indirectly reflect changes in soil decomposition channels and energy flow under drought stress, thereby serving the classification and assessment of farmland drought.

[0004] Existing assessment methods often rely on a single monitoring indicator or a preset judgment threshold, making it difficult to simultaneously reflect the correspondence between crop yield changes at different irrigation strategies and soil biological community responses. This can lead to discrepancies between the obtained drought level and the actual crop drought situation. In particular, when using nematode food web-related characteristics as drought response indicators, if only fixed thresholds or manual empirical rules are used for judgment, it is easy to misjudge the drought level due to differences in the distribution of nematode community characteristics in different fields and under different irrigation conditions.

[0005] Meanwhile, existing threshold determination methods based on optimization algorithms often only focus on overall judgment accuracy, failing to fully consider misjudgments at individual drought levels or address the risk of nematode community characteristics approaching threshold boundaries. For samples near the judgment boundary, even small data fluctuations can cause changes in drought level, thus reducing the stability and applicability of drought assessment rules. Therefore, a farmland drought assessment method is needed that can construct a correlation between nematode community characteristics under different irrigation strategies and actual drought levels, and optimize the thresholds in the drought level judgment rules to improve the accuracy and robustness of drought level judgment.

[0006] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for assessing farmland drought based on nematode food web monitoring, in order to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A method for assessing farmland drought based on nematode food web monitoring, comprising the following steps: Step 1: Obtain yield data and nematode community characteristics at crop maturity under different irrigation strategies in the experimental field, and determine the actual drought level based on the yield data. Construct samples of nematode community characteristics and actual drought levels under all irrigation strategies, and all samples together constitute the actual control set. Step 2: Set drought level judgment rules based on the threshold of each nematode community characteristics, and obtain the analysis drought level of all samples in the actual control set based on the drought level judgment rules, and form a drought judgment set. Use the threshold of each nematode community characteristics as optimization parameters to generate individual vectors, and generate an initial population based on multiple individual vectors. Step 3: Generate optimized indices based on the actual control set and drought judgment set. The optimized indices are the misjudgment rates of each drought level. Based on the differences in the edge risk distribution of each nematode community feature relative to its corresponding threshold in the actual control set, improve the selection operation of the genetic algorithm. Based on the distance of each nematode community feature to its corresponding threshold, improve the mutation operation of the genetic algorithm. Step 4: With minimizing the optimization index as the optimization objective, optimize the initial population based on the improved genetic algorithm to obtain the optimal individual vector, and use the drought level judgment rule corresponding to the optimal individual vector as the final farmland drought assessment rule.

[0009] Furthermore, the yield data refers to crop yield per unit area, and the nematode community characteristics include: deviation rate of bacteriophageal nematodes, deviation rate of fungiophageal nematodes, and channel index offset. The rule for determining the actual drought level based on yield data is: a first yield threshold and a second yield threshold are preset, and the second yield threshold is less than the first yield threshold; If the yield data is greater than the preset first yield threshold, the drought level is determined to be level 0; if the crop yield is not greater than the preset first yield threshold but not less than the preset second yield threshold, the drought level is determined to be level 1; if the yield is less than the preset second yield threshold, the drought level is determined to be level 2.

[0010] Furthermore, the logic for setting drought level judgment rules based on thresholds of various nematode community characteristics is as follows: Preset a first feature threshold and a second feature threshold for the deviation rate of bacteriophages, the deviation rate of fungiophages, and the channel index offset, respectively, and the second feature threshold is less than the first feature threshold. The real-time deviation rates of bacteriophageal nematodes, fungiophageal nematodes, and channel index offsets are compared one by one with their corresponding two-level thresholds: When the characteristics of the first-line insect community are greater than the corresponding first characteristic threshold, the preliminary drought level corresponding to this characteristic is level 2; When the characteristics of the first-line insect community are not greater than the corresponding first characteristic threshold and not less than the corresponding second characteristic threshold, the preliminary drought level corresponding to this characteristic is level 1; When the first-line insect community characteristic is less than the corresponding second characteristic threshold, the preliminary drought level corresponding to this characteristic is level 0; The majority voting method was used to count the frequency of the three preliminary drought levels, and the level with the highest frequency was taken as the final drought level. If the frequency of each preliminary drought level is the same, the level with the highest value will be selected as the final drought level.

[0011] Furthermore, based on the drought level judgment rules, the drought level of all samples in the actual control set is obtained, and a drought judgment set is constructed. The specific logic is as follows: The deviation rates of bacteriophages, fungiophages, and channel index for each sample in the actual control set were extracted sequentially. The drought level was determined for each sample according to the drought level judgment rules, and the drought level of each sample was obtained. The judgment results of all samples were summarized to form a drought judgment set.

[0012] Furthermore, an initial population is generated based on multiple individual vectors, with the specific logic as follows: A single individual vector is constructed using the thresholds corresponding to the deviation rates of bacteriophages, fungiophages, and channel index offsets as parameters. Each individual vector represents an independent combination of feature thresholds. Generate a number of distinct individual vectors according to a preset quantity, and integrate all individual vectors together to form the initial population of the genetic algorithm.

[0013] Furthermore, the logic for obtaining the misjudgment rate for each drought level is as follows: Establish a one-to-one correspondence between samples within the drought judgment set and the actual control set, and compare the drought judgment level of the same sample with the actual drought level; The total number of samples and the number of misclassified samples were counted separately for the three drought levels: Level 0, Level 1, and Level 2. The misclassification rate for a single drought level is calculated by dividing the number of misclassified samples for that level by the total number of samples for that level.

[0014] Furthermore, based on the differences in edge-fitting risk distribution of each nematode community characteristic relative to its corresponding threshold in the actual control set, the selection operation of the genetic algorithm is improved. The specific logic is as follows: With minimizing the misjudgment rate of each drought level as the optimization objective, the vectors of all individuals within the initial population are non-dominated and sorted to form a non-dominated sorting layer. Two individual vectors are randomly selected from the population for selection; if the two individual vectors belong to different non-dominated sorting levels, the individual vector with the best level is selected first to enter the offspring population. If two individual vectors are in the same non-dominated ordination layer, the risk of each threshold relative to the actual control set sample is quantified. The safety margin of each threshold is judged in combination with the distribution of nematode community characteristic samples. The individual vector with the lowest risk of edge contact and the highest safety margin is selected first to complete this round of selection and screening. This process is repeated until the selection and update of the entire generation population is completed. The specific process of quantifying the edge risk of each group's threshold relative to the actual control set samples and assessing the degree of safety margin is as follows: Calculate the distance between the nematode community feature and the corresponding nematode community feature threshold for each sample in the actual control set. Normalize the distance for each nematode community feature. For the threshold of each nematode community feature, form a boundary distance set by combining the normalized boundary distances of all samples in the actual control set, and extract the distance corresponding to the preset low quantile from the boundary distance set as the safety interval distance for the threshold of that nematode community feature. Calculate the difference between the normalized result of the threshold of the nematode community feature and the safety interval feature as the normalized safety interval feature for the threshold of that nematode community feature. For any individual vector, sum the normalized safety interval features of the thresholds of all its nematode community features to obtain the boundary safety interval difference, and select the individual vector with the largest boundary safety interval difference to enter the offspring population.

[0015] Furthermore, the logic for improving the genetic algorithm mutation based on the distance between the nematode community features and the corresponding threshold in the actual control set is as follows: preset the mutation amplitude of each nematode community feature; for the threshold of any nematode community feature in any individual vector to be mutated, construct a mutation interval with the threshold as the center and the length of the preset mutation amplitude; and construct a high edge risk interval with the threshold as the center and the length of twice the safety interval distance. Candidate mutation values ​​are randomly generated within the mutation interval, and it is determined whether the candidate mutation value is located within the high edge risk interval; if the candidate mutation value is not located within the high edge risk interval, the candidate mutation value is determined as the mutation result of the threshold. If the candidate mutation value is located within the high edge risk interval, then a new candidate mutation value is randomly generated within the mutation interval and used as the mutation result of the threshold.

[0016] Furthermore, the initial population is optimized using the improved genetic algorithm to obtain the optimal individual vector. The specific logic is as follows: With minimizing the misjudgment rate of each drought level as the optimization objective, the improved selection, crossover and mutation operations are sequentially performed on the initial population to complete a single population iteration update. The process is repeated iteratively, and the optimization index corresponding to the vectors of all individuals in each generation of the population is recorded in real time. Set the maximum number of iterations as the algorithm termination condition, and terminate the optimization when the number of iterations reaches the preset threshold. After the iteration terminates, the individual vectors in the population are statistically analyzed and sorted non-dominatedly according to the optimization objective. For the individual vectors in the best non-dominated sorting level, the corresponding comprehensive misclassification rate is calculated, and the individual vector with the smallest comprehensive misclassification rate is selected as the optimal individual vector. The overall misjudgment rate is the linear sum of the misjudgment rates for each drought level.

[0017] This invention also provides a farmland drought assessment system based on nematode food web monitoring. The system is used to implement the aforementioned farmland drought assessment method based on nematode food web monitoring, specifically including: The sample construction module is used to obtain yield data and nematode community characteristics of crops at maturity under different irrigation strategies in the experimental field, and to determine the actual drought level based on the yield data. It constructs samples based on the nematode community characteristics and actual drought levels under all irrigation strategies, and all samples together constitute the actual control set. The population construction module is used to set drought level judgment rules based on the threshold of each nematode community characteristics, and to obtain the analysis drought level of all samples in the actual control set based on the drought level judgment rules, and to form a drought judgment set. The threshold of each nematode community characteristics is used as an optimization parameter to generate individual vectors, and an initial population is generated based on multiple individual vectors. The genetic improvement module is used to generate optimization indicators based on the actual control set and the drought judgment set. The optimization indicators are the misjudgment rate of each drought level. Based on the difference in edge risk distribution of each nematode community feature relative to its corresponding threshold in the actual control set, the selection operation of the genetic algorithm is improved. Based on the distance of each nematode community feature to the corresponding threshold, the mutation operation of the genetic algorithm is improved. The genetic optimization module is used to optimize the initial population based on an improved genetic algorithm with the goal of minimizing the optimization index, to obtain the optimal individual vector, and the drought level judgment rule corresponding to the optimal individual vector is used as the final farmland drought assessment rule.

[0018] Compared with the prior art, the beneficial effects of the present invention are: This invention obtains crop yield and nematode community characteristics at different irrigation strategies and determines the actual drought level based on crop growth. It establishes a correspondence between nematode food web monitoring results and actual crop drought performance. Compared to methods that rely solely on empirical thresholds or single monitoring indicators, this approach uses actual control sets and drought judgment sets to correct the drought level judgment rules, making the final judgment rules more consistent with the actual drought performance of farmland under different irrigation conditions. This reduces the judgment bias caused by the mismatch between nematode community characteristics and actual drought levels.

[0019] Furthermore, this scheme uses the misclassification rate of each drought level as the optimization index, and incorporates the difference in edge risk distribution of nematode community features relative to the corresponding threshold into the selection process of the genetic algorithm, and incorporates the distance between nematode community features and the threshold into the mutation process of the genetic algorithm. This can take into account the judgment accuracy of different drought levels and the stability of boundary samples when optimizing the threshold. The drought level judgment rule corresponding to the optimal individual vector obtained in this way can be used to judge the current drought level, so that the drought assessment results have better accuracy, resistance to fluctuations and applicability. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a diagram illustrating the iterative process of the genetic algorithm of this invention; Figure 3 This is a schematic diagram of the overall system structure of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0022] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0023] Example: Please see Figures 1-3 The present invention provides a technical solution: A method for assessing farmland drought based on nematode food web monitoring, comprising the following steps: Step 1: Obtain yield data and nematode community characteristics at crop maturity under different irrigation strategies in the experimental field, and determine the actual drought level based on the yield data. Construct samples of nematode community characteristics and actual drought levels under all irrigation strategies, and all samples together constitute the actual control set. To simulate drought conditions, the irrigation strategy with the highest yield in the historical irrigation strategies is selected. Based on this irrigation strategy, the reduction in irrigation time is set as a certain percentage of the initial irrigation strategy. The percentage is selected based on the actual conditions, usually between 1% and 8%, to ensure a sufficient number of samples and to ensure sample diversity.

[0024] It should be noted that in the experimental field, the environment and the amount of crops planted were exactly the same, only the irrigation strategy was different, in order to ensure that the drought was caused only by insufficient irrigation, that is, water shortage and drought.

[0025] The irrigation strategy includes irrigating at a preset time every day, and then reducing the duration of each irrigation session by a set reduction rate based on the existing irrigation strategy to generate a new irrigation strategy. Subsequently, based on this new strategy, the duration of each irrigation session is further reduced by the same or set reduction rate, thereby gradually generating multiple different irrigation strategies.

[0026] Furthermore, the yield data refers to crop yield per unit area, and the nematode community characteristics include: deviation rate of bacteriophageal nematodes, deviation rate of fungiophageal nematodes, and channel index offset. The acquisition is achieved through existing methods, specifically by: acquiring the abundance of bacteriophageal nematodes and fungiophageal nematodes per unit volume of farmland soil sample; acquiring the standard bacteriophageal nematode abundance and fungiophageal nematode abundance for the region; and dividing the difference between the standard bacteriophageal nematode abundance and the standard bacteriophageal nematode abundance by the standard bacteriophageal nematode abundance to obtain the bacteriophageal nematode deviation rate. It should be noted that if the bacteriophageal nematode deviation rate is less than 0, it is set to 0, because drought usually inhibits the survival of bacteriophageal nematodes; if the bacteriophageal nematode deviation rate is less than 0, it indicates that the number of nematodes has increased relative to the standard, reflecting that the environment can provide a better survival environment for bacteriophageal nematodes.

[0027] The deviation rate of fungi-eating nematodes is obtained by dividing the difference between the abundance of standard fungi-eating nematodes and the abundance of standard fungi-eating nematodes by the abundance of standard fungi-eating nematodes. It should be noted that if the deviation rate of bacteriophages is less than 0, the deviation rate of bacteriophages is set to 0. The same principle applies to bacteriophages, which will not be elaborated here.

[0028] The calculation of the channel index is an existing technique. Specifically, the numerator is the abundance of bacteriophages, and the denominator is the sum of the abundance of bacteriophages and fungiophages. The resulting fraction is called the channel index. The channel index offset is the difference between the channel index and the standard channel index.

[0029] The specific methods for obtaining the channel index, bacteriophage abundance, and fungiophage abundance can be found in "The Influence of Mowing on the Soil Nematode Community Structure in Semi-arid Grasslands" in the Journal of Grassland Science, Vol. 34, No. 2, 2026, and will not be elaborated further.

[0030] The standard fungal nematode abundance, standard bacteriophage nematode abundance, and standard channel index are obtained based on historical data. Specifically, multiple sets of fungal nematode abundance, bacteriophage nematode abundance, and channel index are obtained at the crop maturity period in local historical drought-free years, and their average value is taken as the standard fungal nematode abundance, standard bacteriophage nematode abundance, and standard channel index to characterize the standard fungal nematode abundance, standard bacteriophage nematode abundance, and standard channel index under drought-free conditions.

[0031] The rule for determining the actual drought level based on yield data is: a first yield threshold and a second yield threshold are preset, and the second yield threshold is less than the first yield threshold; If the yield data is greater than the preset first yield threshold, the drought level is determined to be level 0; if the crop yield is not greater than the preset first yield threshold but not less than the preset second yield threshold, the drought level is determined to be level 1; if the yield is less than the preset second yield threshold, the drought level is determined to be level 2.

[0032] The first and second yield thresholds are determined based on standard yields, which are the average yields per unit area of ​​the same crop in the same region under fully irrigated conditions or in historically drought-free years at maturity. 90% of this average is used as the first yield threshold, and 70% as the second yield threshold. Using 90% of the standard yield as the first yield threshold can distinguish between natural fluctuations in the field, measurement errors, or slight yield reductions and the actual impact of drought. Using 70% of the standard yield as the second yield threshold can distinguish between general yield reductions and significant yield reductions. Thus, a yield higher than the first yield threshold corresponds to no significant drought, a yield between the two thresholds corresponds to mild to moderate drought, and a yield lower than the second yield threshold corresponds to severe drought, ensuring that the actual drought level determined based on yield corresponds to the actual drought-induced yield reduction of the crop.

[0033] Step 2: Set drought level judgment rules based on the threshold of each nematode community characteristics, and obtain the analysis drought level of all samples in the actual control set based on the drought level judgment rules, and form a drought judgment set. Use the threshold of each nematode community characteristics as optimization parameters to generate individual vectors, and generate an initial population based on multiple individual vectors. Furthermore, the logic for setting drought level judgment rules based on thresholds of various nematode community characteristics is as follows: Preset a first feature threshold and a second feature threshold for the deviation rate of bacteriophages, the deviation rate of fungiophages, and the channel index offset, respectively, and the second feature threshold is less than the first feature threshold. The real-time deviation rates of bacteriophageal nematodes, fungiophageal nematodes, and channel index offsets are compared one by one with their corresponding two-level thresholds: When the characteristics of the first-line insect community are greater than the corresponding first characteristic threshold, the preliminary drought level corresponding to this characteristic is level 2; When the characteristics of the first-line insect community are not greater than the corresponding first characteristic threshold and not less than the corresponding second characteristic threshold, the preliminary drought level corresponding to this characteristic is level 1; When the first-line insect community characteristic is less than the corresponding second characteristic threshold, the preliminary drought level corresponding to this characteristic is level 0; The majority voting method was used to count the frequency of the three preliminary drought levels, and the level with the highest frequency was taken as the final drought level. If the frequency of each preliminary drought level is the same, the level with the highest value will be selected as the final drought level.

[0034] The deviation rates of bacteriophageal nematodes, fungiophageal nematodes, and channel index offsets all reflect the degree of deviation of soil nematode communities relative to normal moisture conditions. When farmland is affected by drought, changes in soil moisture content affect the soil microbial environment and nematode food sources, thereby causing changes in nematode community structure. Therefore, the greater the deviation of nematode community characteristics, the more significant the drought impact on farmland. Based on this, this embodiment sets a first feature threshold and a second feature threshold for each nematode community characteristic, dividing the continuously changing nematode community characteristics into three level intervals to correspond to different degrees of drought impact. When a nematode community characteristic is greater than the corresponding first feature threshold, it indicates that the characteristic deviates significantly from the normal state, and therefore the drought level corresponding to the characteristic is determined to be level 2; when a nematode community characteristic is between the first and second feature thresholds, it indicates that the characteristic has a certain deviation, and therefore the drought level corresponding to the characteristic is determined to be level 1; when a nematode community characteristic is less than the corresponding second feature threshold, it indicates that the characteristic deviates slightly, and therefore the drought level corresponding to the characteristic is determined to be level 0. Because the numerical ranges, variability, and drought response sensitivities of different nematode community characteristics vary, separate thresholds were set for the deviation rates of bacteriophageal nematodes, fungiophageal nematodes, and channel index offsets, instead of using a uniform threshold. After obtaining the preliminary drought levels corresponding to the three nematode community characteristics, a majority voting method was used to determine the final drought level. This method integrates the judgment results of multiple nematode community characteristics, reducing the impact of abnormal fluctuations in a single characteristic or sampling errors on the final judgment result. When the three preliminary drought levels occur with the same frequency, the level with the highest numerical value is selected as the final drought level. This is to prioritize the higher drought risk when judgment results are inconsistent, avoid underestimating more severe drought conditions, and make the drought assessment results more robust.

[0035] Furthermore, based on the drought level judgment rules, the drought level of all samples in the actual control set is obtained, and a drought judgment set is constructed. The specific logic is as follows: The deviation rates of bacteriophages, fungiophages, and channel index for each sample in the actual control set were extracted sequentially. The drought level was determined for each sample according to the drought level judgment rules, and the drought level of each sample was obtained. The judgment results of all samples were summarized to form a drought judgment set.

[0036] While yield data at maturity can serve as a reliable basis for assessing actual drought levels, its acquisition is time-delayed, typically only available after crop maturity or harvest, failing to meet the needs for process-based monitoring and early assessment of farmland drought. In contrast, the deviation rates of bacteriophageal nematodes, fungiophageal nematodes, and channel index offsets can be obtained through soil sample testing during crop growth, offering better real-time or phased monitoring value. Therefore, this embodiment utilizes yield at maturity to determine the actual drought level and uses this actual drought level as a control label. The judgment threshold corresponding to nematode community characteristics is optimized, enabling nematode community characteristics to replace the delayed yield results in practical applications, allowing for early assessment of farmland drought levels.

[0037] Furthermore, an initial population is generated based on multiple individual vectors, with the specific logic as follows: A single individual vector is constructed using the thresholds corresponding to the deviation rates of bacteriophages, fungiophages, and channel index offsets as parameters. Each individual vector represents an independent combination of feature thresholds. Generate a number of distinct individual vectors according to a preset quantity, and integrate all individual vectors together to form the initial population of the genetic algorithm.

[0038] Furthermore, the logic for obtaining the misjudgment rate for each drought level is as follows: Establish a one-to-one correspondence between samples within the drought judgment set and the actual control set, and compare the drought judgment level of the same sample with the actual drought level; The total number of samples and the number of misclassified samples were counted separately for the three drought levels: Level 0, Level 1, and Level 2. The misclassification rate for a single drought level is calculated by dividing the number of misclassified samples for that level by the total number of samples for that level.

[0039] Genetic algorithms include selection, crossover, and mutation operations. Specifically, individuals are selected from the initial population to form a crossover offspring population. After the crossover, the offspring population is mutated. Usually, a certain proportion of individuals are retained in the initial population and together with the mutated offspring population to form the next generation offspring population. Since selection usually only selects half of the individuals, the proportion retained is lower than the proportion selected.

[0040] Step 3: Generate optimized indices based on the actual control set and drought judgment set. The optimized indices are the misjudgment rates of each drought level. Based on the differences in the edge risk distribution of each nematode community feature relative to its corresponding threshold in the actual control set, improve the selection operation of the genetic algorithm. Based on the distance of each nematode community feature to its corresponding threshold, improve the mutation operation of the genetic algorithm. The misclassification rate for a single drought level is calculated by dividing the number of misclassified samples for that level by the total number of samples for that level.

[0041] Furthermore, based on the differences in edge-fitting risk distribution of each nematode community characteristic relative to its corresponding threshold in the actual control set, the selection operation of the genetic algorithm is improved. The specific logic is as follows: With minimizing the misjudgment rate of each drought level as the optimization objective, the vectors of all individuals within the initial population are non-dominated and sorted to form a non-dominated sorting layer. It should be noted that non-dominated sorting is an existing technology. Non-dominated sorting is a sorting method used in multi-objective optimization to compare the merits of different candidate schemes. In this embodiment, each individual vector corresponds to a set of nematode community feature thresholds, and its optimization objective includes minimizing the misclassification rate of drought levels 0, 1, and 2. If an individual vector is not higher than another individual vector in the misclassification rate of all drought levels, and is lower than another individual vector in the misclassification rate of at least one drought level, then the individual vector is considered to dominate the other individual vector. Non-dominated sorting is to stratify all individual vectors in the population according to the above dominance relationship: individuals not dominated by any other individual vector are assigned to the first layer; after removing individuals from the first layer, the remaining individuals not dominated by other remaining individuals are assigned to the second layer, and so on. The earlier the layer, the better the overall performance of the individual vector under multiple misclassification rate objectives, and the more suitable it is as a candidate threshold combination for subsequent genetic iterations.

[0042] Therefore, the following preferential selection of the optimal level specifically refers to selecting the individual vectors in the lower non-dominated sorting level.

[0043] Two individual vectors are randomly selected from the population for selection; if the two individual vectors belong to different non-dominated sorting levels, the individual vector with the best level is selected first to enter the offspring population. If two individual vectors are in the same non-dominated ordination layer, the risk of each threshold relative to the actual control set sample is quantified. The safety margin of each threshold is judged in combination with the distribution of nematode community characteristic samples. The individual vector with the lowest risk of edge contact and the highest safety margin is selected first to complete this round of selection and screening. This process is repeated until the selection and update of the entire generation population is completed. If two individual vectors are in the same non-dominated ordination layer, it indicates that it is difficult to directly distinguish their superiority or inferiority in terms of misclassification rates at various drought levels. In this case, edge risk is further introduced as an auxiliary selection criterion. Edge risk refers to the situation where the sample feature values ​​in the actual control set are too close to a certain judgment threshold. When the sample feature values ​​are close to the threshold, sampling errors, detection errors, local differences in the field, or short-term environmental fluctuations may cause the sample to change from one side of the threshold to the other side, thereby causing a change in the drought level. Therefore, even if the current misclassification rates of two threshold combinations are similar, if one of the threshold combinations causes a large number of samples to concentrate near the threshold, then that threshold combination is more likely to have unstable judgments in practical applications.

[0044] For the reasons mentioned above, this embodiment prioritizes individual vectors with lower edge-grabbing risk and higher safety margin within the same non-dominated sorting layer. Specifically, for any candidate individual vector, the distance between the deviation rate of bacteriophageal nematodes, the deviation rate of fungiophageal nematodes, and the channel index offset of each sample in the actual control set and their corresponding thresholds is calculated, and the distances of different nematode community features are normalized. The reason for normalization is that the dimensions, numerical ranges, and fluctuation amplitudes of different nematode community features may be different. If the original distances are directly compared, features with larger numerical ranges will have a greater impact on the results, leading to unfair edge-grabbing risk assessment. Through normalization, the boundary distances of different features can be kept within a comparable range.

[0045] Subsequently, for each threshold, the normalized distances of all samples in the actual control set relative to that threshold are used to form a boundary distance set, and the distance corresponding to the preset low quantile is extracted from this set as the safety interval distance. The reason for using the low quantile instead of the average distance is that the risk of being too close to the threshold is mainly determined by a subset of samples. Even if most samples are far from the threshold, the actual judgment may still be unstable as long as a large number of samples are close to the threshold. Therefore, the low quantile can more sensitively reflect the distribution of samples close to the threshold and is more suitable for characterizing boundary risk than the average.

[0046] Low quantiles are typically taken from 5% to 20%. When there are no special data requirements, the 10% quantile is preferred. The 10% quantile reflects the situation of samples close to the threshold without being overly affected by individual outliers, detection errors, or random sampling points, unlike the 1% or 5% quantiles. If the quantile value is too low, the safety interval may be controlled by a single outlier, causing the algorithm to over-avoid it. If the quantile value is too high, such as 25% or 30%, the evaluation result is closer to the overall distribution and cannot sensitively reflect the risk of samples truly close to the threshold. Therefore, the 10% quantile is usually a compromise between risk sensitivity and resistance to anomaly interference.

[0047] For any individual vector, the safety interval evaluation results of the thresholds corresponding to each nematode community feature are summed to obtain the boundary safety interval difference of that individual vector. The larger the boundary safety interval difference, the further away the threshold group is from the dense sample boundary region, the lower the probability of the sample falling near the threshold, and the less likely the judgment result is to change due to small fluctuations. Therefore, in the same non-dominated ordination layer, individual vectors with larger boundary safety interval differences are selected to enter the offspring population. In this way, the boundary stability of the threshold combination can be improved while maintaining the direction of false positive rate optimization.

[0048] The specific process of quantifying the edge risk of each group's threshold relative to the actual control set samples and assessing the degree of safety margin is as follows: Calculate the distance between the nematode community feature and the corresponding nematode community feature threshold for each sample in the actual control set. Normalize the distance for each nematode community feature. For the threshold of each nematode community feature, form a boundary distance set by combining the normalized boundary distances of all samples in the actual control set, and extract the distance corresponding to the preset low quantile from the boundary distance set as the safety interval distance for the threshold of that nematode community feature. Calculate the difference between the normalized result of the threshold of the nematode community feature and the safety interval feature as the normalized safety interval feature for the threshold of that nematode community feature. For any individual vector, sum the normalized safety interval features of the thresholds of all its nematode community features to obtain the boundary safety interval difference, and select the individual vector with the largest boundary safety interval difference to enter the offspring population.

[0049] Furthermore, the logic for improving the genetic algorithm mutation based on the distance between the nematode community features and the corresponding threshold in the actual control set is as follows: preset the mutation amplitude of each nematode community feature; for the threshold of any nematode community feature in any individual vector to be mutated, construct a mutation interval with the threshold as the center and the length of the preset mutation amplitude; and construct a high edge risk interval with the threshold as the center and the length of twice the safety interval distance. Candidate mutation values ​​are randomly generated within the mutation interval, and it is determined whether the candidate mutation value is located within the high edge risk interval; if the candidate mutation value is not located within the high edge risk interval, the candidate mutation value is determined as the mutation result of the threshold. If the candidate mutation value is located within the high edge risk interval, then a new candidate mutation value is randomly generated within the mutation interval and used as the mutation result of the threshold.

[0050] When the first generated candidate variant value does not fall into the high edge risk range, it means that the candidate variant value is relatively far away from the sample edge area, and it is directly used as the mutation result of the threshold. When the first generated candidate variant value falls into the high edge risk range, it means that the candidate variant value may make the mutated threshold still near the sample dense boundary. At this time, candidate variant values ​​are randomly generated again in the mutation range, and the regenerated candidate variant values ​​are used as the mutation result of the threshold.

[0051] It should be noted that the above method of regenerating candidate mutant values ​​does not completely eliminate high-risk intervals, but rather reduces the probability that the mutation result falls into a high-risk interval; let the proportion of high-risk intervals in the mutation interval be... If a normal random mutation method is used, the probability that a candidate mutation value falls into the high-risk edge range is: In this embodiment, the final mutation result will only fall into the high-risk-adjacent-edge interval if both the first candidate mutation value and the second regenerated candidate mutation value fall into the high-risk-adjacent-edge interval. Therefore, the probability is reduced to [missing value]. .because Therefore This reduces the likelihood of the threshold mutating into the high-risk edge range.

[0052] Completely prohibiting mutated values ​​from entering high-risk border regions might excessively limit the search range of the genetic algorithm, making it difficult for the algorithm to discover threshold combinations that, while close to the boundary, have a low overall false positive rate. Conversely, without any restrictions, random mutations could easily cause thresholds to fall into densely populated boundary regions, making subsequent drought level judgments overly sensitive to detection errors or sample fluctuations. Therefore, this embodiment regenerates the threshold when a candidate mutated value falls into a high-risk border region. This retains the random search capability of the genetic algorithm while reducing the probability of high-risk border regions being selected, making the mutated threshold more likely to avoid sample border regions.

[0053] This mutation method reduces the likelihood of thresholds falling near densely populated sample boundaries, lowering the probability of drought level changes due to slight fluctuations in the threshold, thus improving the stability and robustness of the final drought level determination rule. Simultaneously, since this method does not completely eliminate high-risk edge-adjacency regions, it still retains a certain degree of search diversity, preventing premature convergence of the genetic algorithm.

[0054] Step 4: With minimizing the optimization index as the optimization objective, optimize the initial population based on the improved genetic algorithm to obtain the optimal individual vector, and use the drought level judgment rule corresponding to the optimal individual vector as the final farmland drought assessment rule.

[0055] Please see Figure 2 , Figure 2 This is a diagram illustrating the iterative process of the genetic algorithm of this invention; where the convergence region is... Furthermore, based on the differences in edge-fitting risk distribution of each nematode community characteristic relative to its corresponding threshold in the actual control set, the selection operation of the genetic algorithm is improved. The specific logic is as follows: Furthermore, the initial population is optimized using the improved genetic algorithm to obtain the optimal individual vector. The specific logic is as follows: With minimizing the misjudgment rate of each drought level as the optimization objective, the improved selection, crossover and mutation operations are sequentially performed on the initial population to complete a single population iteration update. The process is repeated iteratively, and the optimization index corresponding to the vectors of all individuals in each generation of the population is recorded in real time. Set the maximum number of iterations as the algorithm termination condition, and terminate the optimization when the number of iterations reaches the preset threshold. After the iteration terminates, the individual vectors in the population are statistically analyzed and sorted non-dominatedly according to the optimization objective. For the individual vectors in the best non-dominated sorting level, the corresponding comprehensive misclassification rate is calculated, and the individual vector with the smallest comprehensive misclassification rate is selected as the optimal individual vector. The overall misjudgment rate is the linear sum of the misjudgment rates for each drought level.

[0056] Please see Figure 3 The present invention further provides a farmland drought assessment system based on nematode food web monitoring. This system is used to implement the aforementioned farmland drought assessment method based on nematode food web monitoring, specifically including: The sample construction module is used to obtain yield data and nematode community characteristics of crops at maturity under different irrigation strategies in the experimental field, and to determine the actual drought level based on the yield data. It constructs samples based on the nematode community characteristics and actual drought levels under all irrigation strategies, and all samples together constitute the actual control set. The population construction module is used to set drought level judgment rules based on the threshold of each nematode community characteristics, and to obtain the analysis drought level of all samples in the actual control set based on the drought level judgment rules, and to form a drought judgment set. The threshold of each nematode community characteristics is used as an optimization parameter to generate individual vectors, and an initial population is generated based on multiple individual vectors. The genetic improvement module is used to generate optimization indicators based on the actual control set and the drought judgment set. The optimization indicators are the misjudgment rate of each drought level. Based on the difference in edge risk distribution of each nematode community feature relative to its corresponding threshold in the actual control set, the selection operation of the genetic algorithm is improved. Based on the distance of each nematode community feature to the corresponding threshold, the mutation operation of the genetic algorithm is improved. The genetic optimization module is used to optimize the initial population based on an improved genetic algorithm with the goal of minimizing the optimization index, to obtain the optimal individual vector, and the drought level judgment rule corresponding to the optimal individual vector is used as the final farmland drought assessment rule.

[0057] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0058] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0059] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0060] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that cannot be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for assessing farmland drought based on nematode food web monitoring, characterized in that, The specific steps include: Step 1: Obtain yield data and nematode community characteristics at crop maturity under different irrigation strategies in the experimental field, and determine the actual drought level based on the yield data. Construct samples of nematode community characteristics and actual drought levels under all irrigation strategies, and all samples together constitute the actual control set. Step 2: Set drought level judgment rules based on the threshold of each nematode community characteristics, and obtain the analysis drought level of all samples in the actual control set based on the drought level judgment rules, and form a drought judgment set. Use the threshold of each nematode community characteristics as optimization parameters to generate individual vectors, and generate an initial population based on multiple individual vectors. Step 3: Generate optimized indices based on the actual control set and drought judgment set. The optimized indices are the misjudgment rates of each drought level. Based on the differences in the edge risk distribution of each nematode community feature relative to its corresponding threshold in the actual control set, improve the selection operation of the genetic algorithm. Based on the distance of each nematode community feature to its corresponding threshold, improve the mutation operation of the genetic algorithm. Step 4: With minimizing the optimization index as the optimization objective, optimize the initial population based on the improved genetic algorithm to obtain the optimal individual vector, and use the drought level judgment rule corresponding to the optimal individual vector as the final farmland drought assessment rule.

2. The method for assessing farmland drought based on nematode food web monitoring according to claim 1, characterized in that: The yield data refers to crop yield per unit area, and the nematode community characteristics include: deviation rate of bacteriophage nematodes, deviation rate of fungiophage nematodes, and channel index offset. The rule for determining the actual drought level based on yield data is: a first yield threshold and a second yield threshold are preset, and the second yield threshold is less than the first yield threshold; If the yield data is greater than the preset first yield threshold, the drought level is determined to be level 0; if the crop yield is not greater than the preset first yield threshold but not less than the preset second yield threshold, the drought level is determined to be level 1; if the yield is less than the preset second yield threshold, the drought level is determined to be level 2.

3. The method for assessing farmland drought based on nematode food web monitoring according to claim 2, characterized in that: The logic for setting drought level judgment rules based on thresholds of various nematode community characteristics is as follows: Preset a first feature threshold and a second feature threshold for the deviation rate of bacteriophages, the deviation rate of fungiophages, and the channel index offset, respectively, and the second feature threshold is less than the first feature threshold. The real-time deviation rates of bacteriophageal nematodes, fungiophageal nematodes, and channel index offsets are compared one by one with their corresponding two-level thresholds: When the characteristics of the first-line insect community are greater than the corresponding first characteristic threshold, the preliminary drought level corresponding to this characteristic is level 2; When the characteristics of the first-line insect community are not greater than the corresponding first characteristic threshold and not less than the corresponding second characteristic threshold, the preliminary drought level corresponding to this characteristic is level 1; When the first-line insect community characteristic is less than the corresponding second characteristic threshold, the preliminary drought level corresponding to this characteristic is level 0; The majority voting method was used to count the frequency of the three preliminary drought levels, and the level with the highest frequency was taken as the final drought level. If the frequency of each preliminary drought level is the same, the level with the highest value will be selected as the final drought level.

4. The method for assessing farmland drought based on nematode food web monitoring according to claim 3, characterized in that: Based on the drought level judgment rules, the drought level of all samples in the actual control set is obtained and a drought judgment set is constructed. The specific logic is as follows: The deviation rates of bacteriophages, fungiophages, and channel index for each sample in the actual control set were extracted sequentially. The drought level was determined for each sample according to the drought level judgment rules, and the drought level of each sample was obtained. The judgment results of all samples were summarized to form a drought judgment set.

5. The method for assessing farmland drought based on nematode food web monitoring according to claim 4, characterized in that: The initial population is generated based on multiple individual vectors, and the specific logic is as follows: A single individual vector is constructed using the thresholds corresponding to the deviation rates of bacteriophages, fungiophages, and channel index offsets as parameters. Each individual vector represents an independent combination of feature thresholds. Generate a number of distinct individual vectors according to a preset quantity, and integrate all individual vectors together to form the initial population of the genetic algorithm.

6. The method for assessing farmland drought based on nematode food web monitoring according to claim 4, characterized in that: The logic for obtaining the misjudgment rate for each drought level is as follows: Establish a one-to-one correspondence between samples within the drought judgment set and the actual control set, and compare the drought judgment level of the same sample with the actual drought level; The total number of samples and the number of misclassified samples were counted separately for the three drought levels: Level 0, Level 1, and Level 2. The misclassification rate for a single drought level is calculated by dividing the number of misclassified samples for that level by the total number of samples for that level.

7. The method for assessing farmland drought based on nematode food web monitoring according to claim 6, characterized in that: Based on the differences in edge-fitting risk distribution of each nematode community characteristic relative to its corresponding threshold in the actual control set, the selection operation of the genetic algorithm is improved. The specific logic is as follows: With minimizing the misjudgment rate of each drought level as the optimization objective, the vectors of all individuals within the initial population are non-dominated and sorted to form a non-dominated sorting layer. Two individual vectors are randomly selected from the population for selection; if the two individual vectors belong to different non-dominated sorting levels, the individual vector with the best level is selected first to enter the offspring population. If two individual vectors are in the same non-dominated ordination layer, the risk of each threshold relative to the actual control set sample is quantified. The safety margin of each threshold is judged in combination with the distribution of nematode community characteristic samples. The individual vector with the lowest risk of edge contact and the highest safety margin is selected first to complete this round of selection and screening. This process is repeated until the selection and update of the entire generation population is completed. The specific process of quantifying the edge risk of each group's threshold relative to the actual control set samples and assessing the degree of safety margin is as follows: Calculate the distance between the nematode community feature and the corresponding nematode community feature threshold for each sample in the actual control set. Normalize the distance for each nematode community feature. For the threshold of each nematode community feature, form a boundary distance set by combining the normalized boundary distances of all samples in the actual control set, and extract the distance corresponding to the preset low quantile from the boundary distance set as the safety interval distance for the threshold of that nematode community feature. Calculate the difference between the normalized result of the threshold of the nematode community feature and the safety interval feature as the normalized safety interval feature for the threshold of that nematode community feature. For any individual vector, sum the normalized safety interval features of the thresholds of all its nematode community features to obtain the boundary safety interval difference, and select the individual vector with the largest boundary safety interval difference to enter the offspring population.

8. The method for assessing farmland drought based on nematode food web monitoring according to claim 7, characterized in that: The logic for improving the genetic algorithm mutation based on the distance between the characteristics of each nematode community in the actual control set and the corresponding threshold is as follows: preset the mutation amplitude of each nematode community feature, for the threshold of any nematode community feature in any individual vector to be mutated, construct a mutation interval with the preset mutation amplitude as the center, and construct a high edge risk interval with twice the length of the safety interval distance as the center. Within the mutation interval, candidate mutation values ​​are randomly generated, and it is determined whether the candidate mutation values ​​are located within the high edge risk interval. If the candidate mutation value is not located within the high edge risk range, then the candidate mutation value is determined as the mutation result of the threshold. If the candidate mutation value is located within the high edge risk interval, then a new candidate mutation value is randomly generated within the mutation interval and used as the mutation result of the threshold.

9. The method for assessing farmland drought based on nematode food web monitoring according to claim 7, characterized in that: The initial population is optimized using an improved genetic algorithm to obtain the optimal individual vector. The specific logic is as follows: With minimizing the misjudgment rate of each drought level as the optimization objective, the improved selection, crossover and mutation operations are sequentially performed on the initial population to complete a single population iteration update. The process is repeated iteratively, and the optimization index corresponding to the vectors of all individuals in each generation of the population is recorded in real time. Set the maximum number of iterations as the algorithm termination condition, and terminate the optimization when the number of iterations reaches the preset threshold. After the iteration terminates, the individual vectors in the population are statistically analyzed and sorted non-dominatedly according to the optimization objective. For the individual vectors in the best non-dominated sorting level, the corresponding comprehensive misclassification rate is calculated, and the individual vector with the smallest comprehensive misclassification rate is selected as the optimal individual vector. The overall misjudgment rate is the linear sum of the misjudgment rates for each drought level.

10. A farmland drought assessment system based on nematode food web monitoring, characterized in that: The system is used to implement the farmland drought assessment method based on nematode food web monitoring as described in any one of claims 1-9, specifically including: The sample construction module is used to obtain yield data and nematode community characteristics of crops at maturity under different irrigation strategies in the experimental field, and to determine the actual drought level based on the yield data. It constructs samples based on the nematode community characteristics and actual drought levels under all irrigation strategies, and all samples together constitute the actual control set. The population construction module is used to set drought level judgment rules based on the threshold of each nematode community characteristics, and to obtain the analysis drought level of all samples in the actual control set based on the drought level judgment rules, and to form a drought judgment set. The threshold of each nematode community characteristics is used as an optimization parameter to generate individual vectors, and an initial population is generated based on multiple individual vectors. The genetic improvement module is used to generate optimization indicators based on the actual control set and the drought judgment set. The optimization indicators are the misjudgment rate of each drought level. Based on the difference in edge risk distribution of each nematode community feature relative to its corresponding threshold in the actual control set, the selection operation of the genetic algorithm is improved. Based on the distance of each nematode community feature to the corresponding threshold, the mutation operation of the genetic algorithm is improved. The genetic optimization module is used to optimize the initial population based on an improved genetic algorithm with the goal of minimizing the optimization index, to obtain the optimal individual vector, and the drought level judgment rule corresponding to the optimal individual vector is used as the final farmland drought assessment rule.

Images