A meteorological multi-scenario simulation method for plant pest and disease prediction and evaluation

Abstract

The application discloses a meteorological multi-scenario simulation method for plant disease and pest prediction and evaluation. The method collects historical meteorological data of a local area, extracts statistical quantities, and is divided into different situations such as low temperature, normal temperature, high temperature, little rain, normal rain and heavy rain according to national evaluation standards, and nine representative meteorological situations are obtained through permutation and combination. The meteorological data under the nine different situations is simulated and output respectively, and the extreme weather situation is covered. The temperature data is generated through a first-order linear autoregressive model, and the precipitation data is simulated through a first-order Markov chain and a skew normal model, which provides basic data support for realizing the response and robustness evaluation of the pest prediction model under the simulation of each meteorological situation. Experiments prove that the method has high reliability, reasonable parameter setting, good prediction effect and robustness.

Description

Technical Field

[0001] This invention belongs to the field of meteorological simulation technology, specifically relating to a multi-scenario meteorological simulation method for predicting and evaluating plant diseases and pests. Background Technology

[0002] Effective prediction of the occurrence and spread of plant diseases and pests can play a positive guiding role in prevention and control efforts. It can not only effectively reduce economic losses, but also reduce pesticide use through precise prevention and control, thus promoting green agricultural development.

[0003] To achieve effective prediction of plant diseases and pests, numerous prediction models have been proposed. However, at present, these models can only be routinely validated based on information within a limited spatiotemporal scope. It is difficult to grasp the model's response to different meteorological scenarios, making it difficult to comprehensively evaluate the model's predictive ability. Meteorological scenarios, especially some typical and extreme meteorological scenarios, are important conditions for the outbreak and disaster of diseases and pests, but the evaluation of the model's response and ability under these scenarios is usually difficult to conduct based on actual meteorological data.

[0004] Therefore, it is essential to propose a method for simulating different meteorological scenarios to achieve comprehensive verification and optimization of the performance of plant disease and pest prediction models under different meteorological scenarios, thereby further improving the prediction accuracy of the models. This also has important application significance in future plant disease and pest prediction research. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a meteorological multi-scenario simulation method for predicting and evaluating plant diseases and pests. Based on historical data, scenario analysis and modeling are performed, and the simulated outputs temperature and precipitation data under the set meteorological scenarios, providing a simulation basis for the prediction and evaluation of diseases and pests.

[0006] The specific steps of this invention are as follows:

[0007] Step 1: Acquisition of local long-term historical meteorological data

[0008] Collect local historical meteorological data, including historical precipitation and temperature data.

[0009] As a preferred option, historical meteorological data from the past thirty years were collected for the local area, with precipitation data in millimeters (mm) and temperature data in degrees Celsius (°C).

[0010] Step 2: Extraction of local meteorological data statistics

[0011] Statistical extraction was performed on the collected local historical meteorological data, including:

[0012] Average daily temperature on the same date over the years

[0013]

[0014] Where N is the number of years selected for calculation, and N is a multiple of 10. j is the year index, j = 1, 2, 3, N. i is the date, i = 1, 2, ..., 365; T ij These are the temperature values ​​for the same date in previous years.

[0015] Standard deviation of temperature on the same date over the years σ i :

[0016]

[0017] Average daily precipitation on the same date over the years

[0018]

[0019] Among them, P ij This represents the rainfall on the same date in previous years.

[0020] Percentage of anomalies on the same date over the years (PA) ij :

[0021]

[0022] Step 3: Setting up meteorological scenarios for plant disease and pest prediction and evaluation

[0023] 3-1. Setting Standards for Different Temperature Scenarios

[0024] Based on historical data and extracted temperature statistics, the following temperature scenario standards are set:

[0025] High temperature scenario standard: original input daily temperature value T ij Increase ΔT i Numerical value, 2.0σ i <ΔT i ≤3.0σ i .

[0026] Standard for normal temperature scenario: Maintain the original daily input temperature value T ij constant.

[0027] Low temperature scenario standard: original input daily temperature value T ij Increase ΔT i Numerical value, -3.0σ i <ΔT i ≤-2.0σ i .

[0028] 3-2. Standard settings for different precipitation scenarios

[0029] Based on historical data and extracted precipitation statistics, the following precipitation scenario standards are set:

[0030] Rainy Scenario Standard: Original Daily Input Precipitation Value P ij Increase in PA ij The percentage, 80% ≤ PA ij ≤95%.

[0031] Standard for constant rainfall scenario: Maintain the original daily input precipitation value P ij constant.

[0032] Low rainfall scenario standard: original daily input precipitation value P ij Increase in PA ij Percentage, -95% ≤ PA ij ≤-80%.

[0033] 3-3. Specific Classification of Meteorological Scenarios

[0034] Based on the established temperature and precipitation scenario standards, the following meteorological scenarios are classified:

[0035] ① High temperature and heavy rainfall scenario

[0036] Temperature: High temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, 2.0σ i <ΔT i ≤3.0σ i And randomly generate ΔT for three consecutive days each month. i =3.0σ i The high temperatures.

[0037] Precipitation: Using the rainy scenario standard, the original input daily precipitation value P ij Increase in PA ij The percentage, 80% ≤ PA ij ≤95%, and generates PA for three consecutive days each month. ij = 95% chance of heavy rainfall.

[0038] ②High temperature and frequent rain scenario

[0039] Temperature: High temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, 2.0σ i <ΔT i ≤3.0σ i And randomly generate ΔT for three consecutive days each month. i =3.0σ i The high temperatures.

[0040] Precipitation: The standard of constant rainfall scenario is adopted, and the original daily input precipitation value P is maintained. ij constant.

[0041] ③ High temperature and low rainfall scenario

[0042] Temperature: High temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, 2.0σ i <ΔT i ≤3.0σ i And randomly generate ΔT for three consecutive days each month. i =3.0σ i The high temperatures.

[0043] Precipitation: Using the low-rainfall scenario standard, the original input daily precipitation value P ij Increase in PA ij Percentage, -95% ≤ PA ij ≤-80%. PA is randomly generated for three consecutive days each month. ij = -95% chance of light precipitation.

[0044] ④ Normal temperature and normal rainfall scenario

[0045] Temperature: Adopting the normal temperature scenario standard, maintaining the original daily input temperature value T. ij constant.

[0046] Precipitation: The standard of constant rainfall scenario is adopted, and the original daily input precipitation value P is maintained. ij constant.

[0047] ⑤ Normal temperature and rainy weather scenario

[0048] Temperature: Adopting the normal temperature scenario standard, maintaining the original daily input temperature value T. ij constant.

[0049] Precipitation: Using the rainy scenario standard, the original input daily precipitation value P ij Increase in PA ij The percentage, 80% ≤ PA ij ≤95%, and generates PA for three consecutive days each month. ij = 95% chance of heavy rainfall.

[0050] ⑥ Normal temperature and low rainfall scenario

[0051] Temperature: Adopting the normal temperature scenario standard, maintaining the original daily input temperature value T. ij constant.

[0052] Precipitation: Using the low-rainfall scenario standard, the original input daily precipitation value P ij Increase in PA ij Percentage, -95% ≤ PAij ≤-80%. PA is randomly generated for three consecutive days each month. ij = -95% chance of light precipitation.

[0053] ⑦ Low temperature and low rainfall scenario

[0054] Temperature: Low temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, -3.0σ i <ΔT i ≤-2.0σ i And randomly generate ΔT = -3.0σ for three consecutive days each month. i The low temperature weather.

[0055] Precipitation: Using the low-rainfall scenario standard, the original input daily precipitation value P ij Increase in PA ij Percentage, -95% ≤ PA ij ≤-80%. PA is randomly generated for three consecutive days each month. ij = -95% chance of light precipitation.

[0056] ⑧ Low temperature and frequent rain scenario

[0057] Temperature: Low temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, -3.0σ i <ΔT i ≤-2.0σ i And randomly generate ΔT = -3.0σ for three consecutive days each month. i The low temperature weather.

[0058] Precipitation: The standard of constant rainfall scenario is adopted, and the original daily input precipitation value P is maintained. ij constant.

[0059] ⑨ Low temperature and rainy weather scenario

[0060] Temperature: Low temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, -3.0σ i <ΔT i ≤-2.0σ i And randomly generate ΔT = -3.0σ for three consecutive days each month. i The low temperature weather.

[0061] Precipitation: Using the rainy scenario standard, the original input daily precipitation value P ij Increase in PA ij The percentage, 80% ≤ PA ij≤95%, and generates PA for three consecutive days each month. ij = 95% chance of heavy rainfall.

[0062] 3-4. Setting the degree of meteorological scenario anomaly

[0063] In the above scenario settings, the duration of continuous high temperature and heavy precipitation is three days, which is considered a mild anomaly. In the moderate and severe anomaly scenarios, the duration of continuous heavy precipitation and high temperature is set to four days and five days, respectively.

[0064] Step 4: Specific simulation of each meteorological scenario

[0065] 4-1 Preprocessing of Input Data for Various Meteorological Scenario Simulations

[0066] Based on the different meteorological scenarios set in step three, select the temperature and precipitation values ​​to be simulated, and then set the degree of anomaly. Then, based on the set anomalous meteorological scenarios, update the temperature and precipitation data in the historical meteorological data, and recalculate the statistics for the updated data, including the temperature T on the same date in previous years. ij daily average and standard deviation σ i And precipitation P on the same date in previous years ij daily average and the percentage of anomalies PA ij .

[0067] 4-2. Temperature Simulation

[0068] Based on the updated meteorological data from April 1st, daily maximum and minimum temperatures for each year were selected, and the residual χ² of the temperature was generated using a first-order linear autoregressive model. j,i (h):

[0069] χ j,i (h)=Aχ j,i-1 (h)+Bε j,i (h)

[0070] Where, χ j,i-1 (h) and ε j,i (h) are all matrices of size 2*1. When h = 1 or 2, χ j,i-1 (h) represent the residuals of the highest temperature and the lowest temperature on the i-th day of the j-th year, respectively, and ε j,i (h) represents a random sequence with a mean of 0 and a variance of 1; A and B are both 2x2 matrices, calculated from measured data, used to ensure that the generated data has the same autocorrelation and cross-correlation coefficients as the measured data sequence.

[0071]

[0072]

[0073]

[0074]

[0075] Wherein, the superscripts -1 and T are the inverse matrix and transpose matrix, respectively; M0 and M1 are the covariance matrix of temperature and the covariance matrix of temperature data sequence lagged by one day, respectively.

[0076] The daily maximum temperature T is calculated based on the residuals generated by the first-order linear autoregressive model. imax The lowest daily temperature T imin :

[0077]

[0078]

[0079] Where, χ j,i The temperature residual is generated by the first-order linear autoregressive model.

[0080] 4-3 Precipitation Simulation

[0081] Using 0.1 mm of precipitation as the criterion for precipitation occurrence, a first-order Markov chain is used to determine whether precipitation has occurred based on the updated meteorological data. When precipitation occurs, a probability distribution model is used to simulate the specific amount of precipitation. The specific steps are as follows:

[0082] 4-3-1 Precipitation Simulation

[0083] Due to the lack of aftereffect in Markov chains, predicting whether precipitation will occur on a given day depends solely on whether precipitation occurred the previous day, and is independent of the precipitation status at any earlier point in time. A wet day (W) is defined as a day with precipitation ≥ 0.1 mm, and a dry day (D) is defined as a day with precipitation < 0.1 mm. Let P(WW) represent the probability of remaining a wet day given a wet day the previous day was, and P(WD) represent the probability of becoming a wet day given a dry day the previous day was. Then, the Markov chain can be uniquely determined by P(WW) and P(WD).

[0084] When the historical meteorological data spans more than thirty years, P(WW) and P(WD) are calculated directly using probability formulas. When the data spans less than thirty years, P(WW) and P(WD) are calculated using the following empirical formulas:

[0085] P(WD) = a * Pw

[0086] P(WW) = 1 - a + P(WD)

[0087] Where a is a constant, ranging from 0.6 to 0.9; Pw represents the probability of wet days occurring within a month.

[0088] As a preferred value, the value of a is 0.75.

[0089] 4-3-2 Random Simulation of Precipitation

[0090] Generate a random number between [0, 1]. Based on whether the previous day was a wet or dry day, compare this random number with the precipitation transition probability P(WW) or P(WD). If the random number is not greater than the precipitation transition probability, then the current day is a wet day. The precipitation deviation for that day is then generated using a skewed normal distribution, with the precipitation amount X. ij From daily precipitation P ij Add the standard normal deviation λ of precipitation iJ get:

[0091] X ij =P ij +λ ij

[0092]

[0093]

[0094]

[0095] Where, λ ij P represents the standard normal deviation. ij Daily precipitation; s i These are the mean and standard deviation of daily precipitation, respectively; g ij is the skewness coefficient; n is the data date length.

[0096] Step 5: Evaluation of Plant Disease and Pest Prediction Models

[0097] Based on the type of plant disease and pest prediction model, the prediction models were evaluated using both static and dynamic prediction model evaluation methods.

[0098] 5-1 Evaluation Methods for Static Prediction Models

[0099] Static prediction models forecast the overall occurrence of pests and diseases in a specific period. The robustness of the model output can be assessed based on changes in the indicators. The specific calculation expression for the static deviation index (Dev_S) is as follows:

[0100]

[0101]

[0102]

[0103] In the formula: X k Y is the output of the prediction model. k The data represents actual plant protection data; k is the prediction period number; num is the number of prediction periods; Mean is the mean; and Std is the standard deviation.

[0104] The static deviation index (Dev_S) is a real number. The value represents the degree of deviation between the model's prediction results and the actual results, that is, the degree of deviation in accuracy. The larger the value, the greater the model deviation and the lower the accuracy. Positive and negative values ​​indicate that the prediction results for pests and diseases are too high or too low, respectively. Positive values ​​indicate too high and negative values ​​indicate too low.

[0105] 5-2 Evaluation Methods for Dynamic Prediction Models

[0106] The dynamic prediction model predicts the occurrence of pests and diseases in a certain location at different times. The specific calculation expression for the dynamic deviation index (Dev_D) is as follows:

[0107]

[0108]

[0109]

[0110] In the formula: X kl Y is the output of the prediction model. kl The data represents actual plant protection data; k is the prediction period number; l is the specific date number within the prediction period; num is the number of prediction periods; num1 is the number of days within the prediction period; Mean is for calculating the mean; and Std is for calculating the standard deviation.

[0111] The present invention has the following beneficial effects:

[0112] Considering the frequent occurrence of extreme weather events, the predicted temperature scenarios are divided into three categories: high temperature, normal temperature, and low temperature. The predicted precipitation scenarios are divided into three categories: abundant rainfall, normal rainfall, and scarce rainfall. These are combined to form nine predicted meteorological scenarios: high temperature with abundant rainfall, high temperature with normal rainfall, high temperature with scarce rainfall, normal temperature with abundant rainfall, normal temperature with normal rainfall, normal temperature with scarce rainfall, low temperature with abundant rainfall, low temperature with normal rainfall, and low temperature with scarce rainfall. Finally, temperature and precipitation are predicted based on these scenarios. These nine scenarios not only include normal weather conditions with normal temperature and rainfall but also cover various extreme weather scenarios, facilitating the evaluation of the pest and disease prediction model's response and robustness under different meteorological conditions. Based on the evaluation indicators, the degree of deviation between the predicted results and actual data can be understood, enabling rapid model evaluation and subsequent parameter optimization to improve the model's predictive accuracy and robustness. Attached Figure Description

[0113] Figure 1Meteorological multi-scenario simulation flowchart;

[0114] Figure 2 A temperature and precipitation map for 2010 generated under a high-temperature and high-rainfall scenario in a certain area of ​​Zhejiang.

[0115] Figure 3 A comparison chart of daily ten-year average temperatures generated under a high-temperature and rainy scenario in a certain area of ​​Zhejiang.

[0116] Figure 4 A comparison chart of daily ten-year average precipitation generated under a high-temperature and high-rainfall scenario in a certain area of ​​Zhejiang.

[0117] Figure 5 Comparison of predicted disease severity under high temperature and high rainfall scenarios and normal temperature and normal rainfall scenarios using the tea green leafhopper pest prediction model;

[0118] Figure 6 Comparison of predicted disease pressure values ​​under high temperature and high rainfall scenarios and normal temperature and rainfall scenarios using the tea green leafhopper pest prediction model. Detailed Implementation

[0119] The present invention will be further explained below with reference to the accompanying drawings;

[0120] like Figure 1 As shown, a meteorological multi-scenario simulation method for predicting and evaluating plant diseases and pests includes the following steps:

[0121] Step 1: Acquisition of local long-term historical meteorological data

[0122] Temperature and precipitation data from 1990 to 2009 were collected for a certain area in Zhejiang Province as local historical meteorological data, and precipitation anomaly data with P=99999 mm were reassigned using P=10 mm.

[0123] Step 2: Extraction of local meteorological data statistics

[0124] Preliminary statistics were extracted from the historical meteorological data collected in step one, including the temperature T on the same date in previous years. ij daily average and standard deviation (σ) i Rainfall P on the same date in previous years ij daily average and the percentage of anomalies PA ij .

[0125] Step 3: Setting up meteorological scenarios for plant disease and pest prediction and evaluation

[0126] Referring to the Chinese temperature rating standard GB / T 35562-2017 (standard deviation evaluation), high temperature, normal temperature, and low temperature scenarios are set up:

[0127] Table 1: Standard Deviation Temperature Evaluation Indicators and Grades

[0128]

[0129] Referring to the national standard for drought levels—GB / T 20481-2017, scenarios of abundant rainfall, normal rainfall, and scarce rainfall are set up:

[0130] Table 2. Classification of Drought Levels by Precipitation Anomaly Percentage

[0131]

[0132] By arranging and combining different scenarios of temperature and rainfall, nine different meteorological scenarios were obtained.

[0133] Step 4: Specific simulation of each meteorological scenario

[0134] Based on the different meteorological scenarios set in step three, select the temperature and precipitation values ​​to be simulated, and then set the degree of anomaly. Then, based on the set abnormal meteorological scenarios, update the temperature and precipitation data in the historical meteorological data to simulate the daily maximum temperature. Daily minimum temperature Daily precipitation P ij .

[0135] Taking a high-temperature and high-rainfall scenario as an example, the 2010 meteorological data output by the simulator for this scenario is as follows: Figure 2 As shown, the output average data for the next ten years is compared with the actual average data. Figure 3 , Figure 4 As shown, under the high temperature and high rainfall scenario, both the overall temperature and precipitation are higher than the actual average, and the overall trend has not changed significantly, indicating that the simulation data is reasonable and effective. Using the same steps, data for two meteorological scenarios—normal temperature and rainfall, and low temperature and low rainfall—were simulated and generated.

[0136] Step 5: Evaluation of Plant Disease and Pest Prediction Models

[0137] In the evaluation process, based on the type of plant disease and pest prediction model, the evaluation methods for prediction models can be divided into static prediction model evaluation methods and dynamic prediction model evaluation methods.

[0138] 5-1 Evaluation Methods for Static Prediction Models

[0139] Static prediction models primarily predict the overall occurrence of pests and diseases during a specific period. Standard deviation is often used to represent the central tendency of data, while bias represents the relative deviation of data. When used for model evaluation, if the overall average bias between the model's predictions under normal temperature and rainfall scenarios and the actual plant protection data is calculated, and the calculated overall bias is compared with the standard deviation of the actual plant protection data, this bias index can be used to measure the accuracy of the model's output. By calculating the corresponding bias index under other scenarios, we can understand the occurrence of pests and diseases under different meteorological conditions, and assess the robustness of the model's output based on changes in the index.

[0140] The specific calculation expression for the static deviation index (Dev_S) is as follows:

[0141]

[0142]

[0143]

[0144] In the formula: X k Y is the output of the prediction model. k The data represents actual plant protection data; k is the prediction period number; num is the number of prediction periods; Mean is the mean; and Std is the standard deviation.

[0145] The static deviation index (Dev_S) is a real number. The value represents the degree of deviation between the model's prediction results and the actual results, that is, the degree of deviation in accuracy. The larger the value, the greater the model deviation and the lower the accuracy. Positive and negative values ​​indicate that the prediction results for pests and diseases are too high or too low, respectively. Positive values ​​indicate too high and negative values ​​indicate too low.

[0146] 5-2 Evaluation Methods for Dynamic Prediction Models

[0147] Dynamic prediction models are mainly used to predict the occurrence of pests and diseases in a certain location at different times. The evaluation method for dynamic prediction models is similar to that for static prediction models. The specific calculation expression for the dynamic deviation index (Dev_D) is as follows:

[0148]

[0149]

[0150]

[0151] In the formula: X kl Y is the output of the prediction model. kl The data represents actual plant protection data; k is the prediction period number; l is the specific date number within the prediction period; num is the number of prediction periods; num1 is the number of days within the prediction period; Mean is for calculating the mean; and Std is for calculating the standard deviation.

[0152] The dynamic deviation index (Dev_D) is a real number. The value represents the degree of deviation between the model's prediction and the actual result, that is, the degree of deviation in accuracy. The larger the value, the greater the model deviation and the lower the accuracy. Positive and negative values ​​indicate that the prediction results for pests and diseases are too high or too low, respectively. Positive values ​​indicate too high and negative values ​​indicate too low.

[0153] This embodiment evaluates a tea green leafhopper pest prediction model based on high-temperature and rainy, low-temperature and low-rainy, and normal-temperature and normal-rainy scenarios output by a meteorological multi-scenario simulator. This model is a dynamic prediction model, and a dynamic prediction model evaluation method is used to evaluate it. The model outputs disease stress values ​​and levels. The predicted severity of tea green leafhopper infestation in 2010 under different scenarios is shown below. Figure 5 As shown, the predicted pest infestation stress values ​​are as follows: Figure 6 As shown.

[0154] The dynamic deviation index Dev_D of the pressure value prediction output under normal temperature and rainfall conditions can be calculated. 压力值 =0.0047, the dynamic bias index Dev_D of the model level prediction output. 等级 = -0.0359; Dev_D, the dynamic deviation index of the model's pressure prediction output under the high temperature and heavy rainfall scenario. 压力值 == 0.0262, the dynamic bias index Dev_D of the model level prediction output. 等级 =0.0539; Dev_D, the dynamic deviation index of the model's pressure prediction output under the low temperature and low rainfall scenario. 压力值 == -0.0815, the dynamic bias index Dev_D of the model-level prediction output. 等级 = -0.0179.

[0155] It is known that the occurrence of tea green leafhopper infestation is closely related to temperature and rainfall. Suitable temperature and rainfall will promote the growth of tea green leafhopper, while otherwise they will inhibit its growth. Furthermore, there are two peak periods for the occurrence of tea green leafhopper infestation each year, in May and September.

[0156] By evaluating the dynamic bias index Dev_D of the model output under various scenarios and comparing it... Figure 5 , Figure 6The predicted occurrence levels and pressure values ​​output by the prediction models under various scenarios show that the overall pest trend predicted by the models remains unchanged, with peak pest periods occurring in May and September in all scenarios. Due to the influence of temperature and precipitation, the pest occurrence period is slightly prolonged in February and March under the high-temperature and rainy scenario, and the model's dynamic bias indicators all show a positive shift under this scenario, with the overall predicted pest pressure value being higher than that under the normal temperature and rainy scenario. Under the low-temperature and low-rainy scenario, the pest occurrence period is slightly shortened in March, and the model's dynamic bias indicators all show a negative shift, with the pressure value being lower than that under the normal temperature and rainy scenario. The prediction results are largely consistent with the habits of pests and diseases, and the trends of the results have maintained a high degree of consistency in multiple simulations, indicating that the model output is reasonable.

[0157] Based on the above calculation results, the model can be considered to have high reliability, reasonable parameter settings, and good predictive performance and robustness.

Claims

1. A meteorological multi-scenario simulation method for predicting and evaluating plant diseases and pests, characterized in that: The method specifically includes the following steps: Step 1: Acquisition of local long-term historical meteorological data Collect local historical meteorological data, including historical precipitation and temperature data; Step 2: Extraction of local meteorological data statistics Statistical analysis was performed on the collected local historical meteorological data, including the daily average temperature on the same date over the years. Standard deviation of temperature on the same date over the years σ i Average daily precipitation on the same date over the years Percentage difference (PA) on the same date in previous years ij ; Step 3: Setting up meteorological scenarios for plant disease and pest prediction and evaluation 3-1. Setting Standards for Different Temperature Scenarios Based on historical data and extracted temperature statistics, the following temperature scenario standards are set: High temperature scenario standard: original input daily temperature value T ij Increase ΔT i Numerical value, 2.0σ i <ΔT i ≤3.0σ i ; Standard for normal temperature scenario: Maintain the original daily input temperature value T ij constant; Low temperature scenario standard: original input daily temperature value T ij Increase ΔT i Numerical value, -3.0σ i <ΔT i ≤-2.0σ i ; 3-2. Standard settings for different precipitation scenarios Based on historical data and extracted precipitation statistics, the following precipitation scenario standards are set: Rainy Scenario Standard: Original Daily Input Precipitation Value P ij Increase in PA ij The percentage, 80% ≤ PA ij ≤95%; Standard for constant rainfall scenario: Maintain the original daily input precipitation value P ij constant; Low rainfall scenario standard: original daily input precipitation value P ij Increase in PA ij Percentage, -95% ≤ PA ij ≤-80%; 3-3. Specific Classification of Meteorological Scenarios Based on the established temperature and precipitation scenario standards, nine meteorological scenarios were set up: high temperature and abundant rainfall, high temperature and normal rainfall, high temperature and little rainfall, normal temperature and normal rainfall, normal temperature and abundant rainfall, normal temperature and little rainfall, low temperature and little rainfall, low temperature and normal rainfall, and low temperature and abundant rainfall. 3-4. Setting the degree of meteorological scenario anomaly Among the nine meteorological scenarios, setting three consecutive days of high temperature or heavy rain is considered a mild anomaly, while four or five consecutive days are considered moderate and severe anomalies, respectively. Step 4: Specific simulation of each meteorological scenario 4-1 Preprocessing of Input Data for Various Meteorological Scenario Simulations Based on the different meteorological scenarios set in step three, select the temperature and precipitation values ​​to be simulated, and then set the degree of anomaly; then, based on the set abnormal meteorological scenarios, update the temperature and precipitation in the historical meteorological data, and recalculate the statistics on the updated data, including the temperature T on the same date in previous years. ij daily average and standard deviation σ i And precipitation P on the same date in previous years ij daily average and the percentage of anomalies PA ij ; 4-2. Temperature Simulation Based on the updated meteorological data from April 1st, daily maximum and minimum temperatures for each year were selected, and the residual χ² of the temperature was generated using a first-order linear autoregressive model. j，i (h): x j，i (h)=Ax j,i-1 (h)+Bε j,i (h) Where, χ j,i-1 (h) and ε j,i (h) are all matrices of size 2*1. When h = 1 or 2, χ j,i-1 (h) represent the residuals of the highest temperature and the lowest temperature on the i-th day of the j-th year, respectively, and ε j，i (h) represents a random sequence with a mean of 0 and a variance of 1; A and B are both matrices of size 2x2. Wherein, the superscript -1 and T are the inverse matrix and transpose matrix, respectively; M0 and M1 are the covariance matrix of temperature and the covariance matrix of temperature data sequence lagged by one day, respectively; The daily maximum temperature T is calculated based on the residuals generated by the first-order linear autoregressive model. imax The lowest daily temperature T imin : Where, χ j,i The temperature residuals generated by the first-order linear autoregressive model; 4-3 Precipitation Simulation Using 0.1 mm of precipitation as the criterion for precipitation occurrence, a first-order Markov chain is used to determine whether precipitation has occurred based on the updated meteorological data. When precipitation occurs, a probability distribution model is used to simulate the specific amount of precipitation.

2. The meteorological multi-scenario simulation method for predicting and evaluating plant diseases and pests as described in claim 1, characterized in that: Historical meteorological data from the past thirty years were collected for the local area, with precipitation data in millimeters and temperature data in degrees Celsius.

3. The meteorological multi-scenario simulation method for predicting and evaluating plant diseases and pests as described in claim 1, characterized in that: Average daily temperature on the same date over the years Where N is the number of years selected for calculation, and N is a multiple of 10; j is the year number, j = 1, 2, 3, N; i is the date, i = 1, 2, ..., 365; T ij These are the temperature values ​​for the same date over the years; Standard deviation of temperature on the same date over the years σ i : Average daily precipitation on the same date over the years Among them, P ij This refers to the rainfall on the same date in previous years; Percentage of anomalies on the same date over the years (PA) ij :

4. The meteorological multi-scenario simulation method for predicting and evaluating plant diseases and pests as described in claim 1, characterized in that: The criteria for the nine meteorological scenarios are as follows: ① High temperature and heavy rainfall scenario Temperature: High temperature scenario standard is adopted, and the original input daily temperature value T ij Increase ΔT i Numerical value, 2.0σ i <ΔT i ≤3.0σ i And randomly generate ΔT for three consecutive days each month. i =3.0σ i High temperatures; Precipitation: Using the rainy scenario standard, the original input daily precipitation value P ij Increase in PA ij The percentage, 80% ≤ PA ij ≤95%, and generates PA for three consecutive days each month. ij =95% chance of heavy rainfall; ②High Temperature and Frequent Rainfall Scenario Temperature: High temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, 2.0σ i <ΔT i ≤3.0σ i And randomly generate ΔT for three consecutive days each month. i =3.0σ i High temperatures; Precipitation: The standard of constant rainfall scenario is adopted, and the original daily input precipitation value P is maintained. ij constant; ③ High temperature and low rainfall scenario Temperature: High temperature scenario standard is adopted, and the original input daily temperature value T ij Increase ΔT i Numerical value, 2.0σ i <ΔT i ≤3.0σ i And randomly generate ΔT for three consecutive days each month. i =3.0σ i High temperatures; Precipitation: Using the low-rainfall scenario standard, the original input daily precipitation value P ij Increase in PA ij Percentage, -95% ≤ PA ij ≤-80%; PA is generated randomly for three consecutive days each month. ij = -95% chance of light precipitation; ④ Normal temperature and normal rainfall scenario Temperature: Adopting the normal temperature scenario standard, maintaining the original daily input temperature value T. ij constant; Precipitation: The standard of constant rainfall scenario is adopted, and the original daily input precipitation value P is maintained. ij constant; ⑤ Normal temperature and rainy weather scenario Temperature: Adopting the normal temperature scenario standard, maintaining the original daily input temperature value T. ij constant; Precipitation: Using the rainy scenario standard, the original input daily precipitation value P ij Increase in PA ij The percentage, 80% ≤ PA ij ≤95%, and generates PA for three consecutive days each month. ij =95% chance of heavy rainfall; ⑥ Normal temperature and low rainfall scenario Temperature: Adopting the normal temperature scenario standard, maintaining the original daily input temperature value T. ij constant; Precipitation: Using the low-rainfall scenario standard, the original input daily precipitation value P ij Increase in PA ij Percentage, -95% ≤ PA ij ≤-80%; PA is generated randomly for three consecutive days each month. ij = -95% chance of light precipitation; ⑦ Low temperature and low rainfall scenario Temperature: Low temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, -3.0σ i <ΔT i ≤-2.0σ i And randomly generate ΔT = -3.0σ for three consecutive days each month. i Low temperatures; Precipitation: Using the low-rainfall scenario standard, the original input daily precipitation value P ij Increase in PA ij Percentage, -95% ≤ PA ij ≤-80%; PA is generated randomly for three consecutive days each month. ij = -95% chance of light precipitation; ⑧ Low temperature and frequent rain scenario Temperature: Low temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, -3.0σ i <ΔT i ≤-2.0σ i And randomly generate ΔT = -3.0σ for three consecutive days each month. i Low temperatures; Precipitation: The standard of constant rainfall scenario is adopted, and the original daily input precipitation value P is maintained. ij constant; ⑨ Low temperature and rainy weather scenario Temperature: Low temperature scenario standard is adopted, and the original input daily temperature value T is used. ij Increase ΔT i Numerical value, -3.0σ i <ΔT i ≤-2.0σ i And randomly generate ΔT = -3.0σ for three consecutive days each month. i Low temperatures; Precipitation: Using the rainy scenario standard, the original input daily precipitation value P ij Increase in PA ij The percentage, 80% ≤ PA ij ≤95%, and generates PA for three consecutive days each month. ij = 95% chance of heavy rainfall.

5. The meteorological multi-scenario simulation method for predicting and evaluating plant diseases and pests as described in claim 1, characterized in that: The specific steps for simulating precipitation are as follows: 4-3-1 Precipitation Simulation A wet day is defined as a day with precipitation ≥ 0.1 mm, denoted by the symbol W; a dry day is defined as a day with precipitation < 0.1 mm, denoted by the symbol D. Let P(WW) represent the probability of remaining a wet day given a previous wet day, and P(WD) represent the probability of becoming a wet day given a previous dry day. When the length of historical meteorological data exceeds thirty years, P(WW) and P(WD) are directly calculated using probability formulas. When the length is less than thirty years, P(WW) and P(WD) are calculated using the following empirical formulas: P(WD) = a * Pw P(WW) = 1 - a + P(WD) Where a is a constant, ranging from 0.6 to 0.9; Pw represents the probability of a wet day occurring within a month; 4-3-2 Random Simulation of Precipitation Generate a random number between [0, 1]. Based on whether the previous day was a wet or dry day, compare this random number with the precipitation transition probability P(WW) or P(WD). If the random number is not greater than the precipitation transition probability, then the current day is a wet day. The precipitation deviation for that day is then generated using a skewed normal distribution, with the precipitation amount X. ij From daily precipitation P ij Add the standard normal deviation λ of precipitation iJ get: X ij =P ij +λ ij Where, λ ij P represents the standard normal deviation. ij Daily precipitation; s i These are the mean and standard deviation of daily precipitation, respectively; g ij is the skewness coefficient; n is the data date length.

6. The meteorological multi-scenario simulation method for predicting and evaluating plant diseases and pests as described in claim 1, characterized in that: The value of the constant a is set to 0.75.

Images