Plum brown rot early warning method fusing phenological and microenvironmental sensing

Abstract

Provided in the present invention is a plum brown rot early warning method fusing phenological and microenvironmental sensing. The method comprises: by defining the processes and stages of occurrence and evolution of specific plant diseases and insect pests, establishing the probability relationship between evolution stages of plum brown rot and occurrence times and durations thereof, and establishing a host phenological stage determination model (sub-model 1) of plum brown rot; determining the microenvironmental conditions and durations for the occurrence and evolution stages of plum brown rot; establishing a state transfer model; accessing monitoring data; calibrating and correcting model parameters; and performing model-based prediction and early warning. All of the seven steps are used for predicting the infection probability (risk) of plum brown rot. The described method establishes an artificial intelligence based AI early warning monitoring, so as to perform on-line automatic monitoring on the microenvironmental dependence conditions of the occurrence and evolution stages of plum brown rot, thus providing early warning in advance, pushing alarm information, and guiding targeted prevention and control in early stages.

Description

An early warning method for plum brown rot integrating phenology and microenvironmental perception Technical Field

[0001] The present invention relates to the field of crop disease and insect pest prediction, and in particular to an early warning method for plum brown rot that integrates phenology and microenvironment perception. Background Art

[0002] Symptoms of brown rot: Caused by fungi of the genera Monilinia and Monilia, with the primary pathogen being the Australian-American type, Monilinia fructicola. The disease infects the fruit surface and stem, producing yellow-brown lesions that gradually expand and become covered with gray, velvety protrusions. As the disease progresses, the fruit surface turns dark brown to black, and a gray mold layer gradually develops around the lesions. Eventually, the affected fruit rots, loses water, and dies, resulting in fruit drop and necrosis.

[0003] Brown rot overwinters as mycelium on trees, in fallen dead fruit, or in ulcers on branch tips. The following spring, it produces large numbers of conidia, which are spread by wind, rain, or insects. Initial infection occurs through insect damage, mechanical wounds, or natural openings. During rainy and foggy weather, large numbers of spores are produced on the surface of diseased parts, leading to secondary infection. Low temperatures, high humidity, and high rainfall during the flowering period are prone to causing blossom end rot. Warm weather, rain, and fog during the fruit ripening period are also prone to causing fruit end rot.

[0004] The limitations of current mainstream pest and disease detection technologies include: First, relying on drone-mounted optical or hyperspectral pods for aerial remote sensing, or high-resolution satellite remote sensing data, to detect pest and disease symptoms is of no practical value for prediction, early warning, and control. The first reason is the latency of data acquisition. The actual acquisition frequency of high-resolution satellite remote sensing data (data transmission to end users) typically takes from 10 days to half a month, and the timeliness of the information lags far behind the speed of disease outbreaks and spread. Second, the resolution is insufficient. Even with drone-mounted aerial remote sensing, the image resolution is insufficient to clearly identify lesions at the leaf level, and various obstructions often distort the image data. Third, the advantage of current technology lies in disease detection, not prediction, prevention, or early warning. This is because most (fungal) diseases have a latent period. By the time symptoms appear, the outbreak and spread of the disease have already begun. This makes it difficult to provide guidance for early warning and can only serve as a supplementary verification after the fact.

[0005] In my country, crop early warnings are typically implemented at very large spatial scales (city and provincial levels and above). However, due to regional disparities, large-scale early warnings are not practically valuable at smaller scales, where microclimate variations are significant. Another key factor is the significant bottlenecks in data acquisition, processing, analysis, and modeling. Summary of the Invention

[0006] The purpose of the present invention is to provide an early warning method for plum brown rot based on the integrated phenology and microenvironment perception, so as to solve the problem of early warning of brown rot.

[0007] To achieve the above object, the present invention is implemented through the following technical solutions:

[0008] The present invention proposes an early warning method for plum brown rot that integrates phenology and microenvironment perception, comprising the following steps:

[0009] Step 1: Define the process and stages of occurrence and evolution of specific pests and diseases;

[0010] Step 2: Establish the probabilistic relationship between each evolutionary stage of plum brown rot and the time of occurrence and duration, and establish a phenological stage judgment model for plum brown rot hosts (sub-model 1):

[0011] Define the probability distribution of the occurrence time and duration of each evolutionary stage of plum brown rot as the system information prior;

[0012] Measuring effective accumulated temperature is used to judge physiological status;

[0013] The lower temperature threshold and the upper temperature threshold are determined through observation and control experiments. The lower temperature threshold is the temperature at which development stops, and the upper temperature threshold is the temperature at which the growth or development rate begins to decline.

[0014] A relationship graph between the probability of germination of plum brown rot fungal spores and effective accumulated temperature under different meteorological conditions can be drawn.

[0015] Step 3: Determine the microenvironmental conditions and duration of plum brown rot occurrence and evolution stages:

[0016] Identify variables highly correlated with plum brown rot, including climate and phenological conditions, soil and environmental conditions, and establish a conversion model between daily rainfall accumulation and leaf wetness (sub-model 2);

[0017] Step 4: Establish a state transition model:

[0018] The evolutionary stages of plum brown rot were defined as the state set of a Markov chain. The basic transition probability matrix was calculated based on microenvironmental conditions and duration. The conditions for pest outbreaks were used as the initial probability to establish a Markov state transition model for pest occurrence and evolution.

[0019] Step 5: Monitoring data access:

[0020] Through micro-weather stations and soil detectors, we can obtain time series data on microclimate conditions closely related to the occurrence and evolution of plum brown rot. We can also obtain data on temperature, relative humidity, wind speed, and radiation intensity through real-time monitoring data from micro-weather stations.

[0021] Step 6: Model parameter calibration:

[0022] The distribution parameters of each model variable are fitted by measured data and the value range is given. The calibration of the model is divided into two stages:

[0023] First, trend and pattern consistency. By conducting statistical analysis on the judgment data of indicators such as infection risk and infection stage, if there is no statistically significant difference between the predicted indicators and the actual indicators, it is considered that the model can capture the trend of disease occurrence and evolution and can also output a relatively reliable disease evolution pattern.

[0024] The second is the consistency of numerical indicators such as recall rate. By tracking the performance of the model, if a high prediction accuracy can be achieved in 2 to 3 seasons in the case of out-of-sample prediction, the task model can work stably.

[0025] Step 7: Model-based forecasting and early warning:

[0026] Establish a process for predicting the probability (risk) of plum brown rot infection;

[0027] Based on meteorological environmental data, a shallow network training microclimate prediction model was established to predict meteorological change data for 5 to 7 days. At the same time, key data such as the phenological period of the plum tree was obtained through the tree management model and related monitoring data. Based on the critical values ​​and ranges of parameters highly correlated with plum brown rot obtained during the training and verification stages of the model, a parameterized disease prediction model was established to give the risk level of infection (0 to 5) and the possible infection stage (0 to 2) one week later.

[0028] Based on the above-mentioned definitions of the occurrence and evolution process and stages of specific pests and diseases, empirical data are used to divide the entire process of plum brown rot infection into multiple specific stages, including: the infected but asymptomatic stage (incubation period), the infected and symptomatic stage, the infectious stage, the transmission stage and the stage where diseased and rotten fruits appear in the fruit. Characteristic variables are defined and quantified.

[0029] According to the plum brown rot host phenological stage judgment model (sub-model 1), it is divided into two stages:

[0030] The first stage, model training and calibration:

[0031] Through 5 years of observation, the Julian days of the host's 12 phenological phase transitions were recorded. The Julian day of each phenological phase transition was averaged and rounded off by taking the average of 5 observation data to obtain the time range of each host phenological phase.

[0032] Using historical meteorological data from NOAA / NWS over the past five years, we obtained daily maximum and minimum temperature data for the desired altitude range. We used the minimum temperature as the effective accumulated temperature threshold to calculate the cumulative effective temperature on the day of each phenological phase transition. We averaged and rounded the five observations to obtain the effective accumulated temperature thresholds (ATi) for each phenological phase transition (where i = 1, 2, 3…12).

[0033] In the second stage, the model predicts:

[0034] According to the forecast accuracy requirements, obtain the daily maximum and minimum temperature data of NOAA / NWS for the next 7 days or 7 months, and start the cumulative calculation from Julian day 1. If the cumulative effective accumulated temperature exceeds ATi, the i-th phenological period output is the current phenological period.

[0035] According to the above, the microenvironmental conditions and duration of the occurrence and evolution stages of plum brown rot are determined.

[0036] The conversion model between daily rainfall accumulation and leaf wetness (sub-model 2):

[0037] The leaf wetness threshold is calculated using daily accumulated rainfall: the minimum daily rainfall is PL = 12 mm / day. If the rainfall is below 12, it is considered that the daily rainfall is insufficient to cause leaf wetness. If the daily rainfall exceeds PL for three consecutive days, the leaf wetness on the third day is set as susceptible and marked as WA. Otherwise, the state of the day is marked as WN. The number of consecutive WA days is accumulated as LWA. If the WN state occurs during the accumulation process, the LWA is reset to zero until a new WA state appears and the LWA accumulation starts again.

[0038] It is also necessary to determine the sufficient number of hours for leaf wetness (LWD) and the hourly temperature during the leaf wetness period (TLWD), create a schematic diagram of the time periods when the critical infection conditions (LWD and TLWD) are reached, and mark the time periods of susceptible conditions. Based on experience or historical data, fit the probability distribution of each relevant variable when plum brown rot occurs. Based on the correlation evaluation of the data, establish the joint probability distribution of each variable as the system's information prior.

[0039] According to the monitoring data access, including input and output parameters:

[0040] Input parameters:

[0041] The effective accumulated temperature for the development of plum brown rot pathogen (Monilia spp.) was calculated, and the lower and upper temperature thresholds were obtained. The threshold parameters were fitted with the values ​​of df, F, and the range of p based on the host's three-year observation experimental data, the highest temperature of the day, the lowest temperature of the day, and the rainfall of the day.

[0042] Starting date: Julian day 1;

[0043] End date: Julian day 245;

[0044] Output parameters:

[0045] The infection stage is either primary or secondary;

[0046] Infection probability, a decimal between 0 and 1;

[0047] Risk level: 4 levels: none, low, medium and high risk.

[0048] According to the above-mentioned plum brown rot infection probability (risk) prediction process,

[0049] Call the plum brown rot host phenology model (sub-model 1) to obtain the current phenological period:

[0050] If the current phenological period is "full bloom", "falling bloom" or "young fruit":

[0051] Call the conversion model of daily rainfall accumulation and leaf wetness (sub-model 2),

[0052] If LWA>=3 and the 3-day average temperature is between 10 and 25 degrees Celsius, the output is the primary infection probability P1, which obeys the normal distribution. where x is the average temperature of the day, with a mean of 15 degrees Celsius and a variance of 4.

[0053] The output infection stage is primary infection;

[0054] If the secondary infection probability P1 exceeds 0.8, the output risk level is high risk;

[0055] The secondary infection probability P1 is between 0.6 and 0.8, and the output risk level is medium risk;

[0056] The secondary infection probability P1 is lower than 0.6, and the output risk level is low risk;

[0057] If the current phenological period is "Hard Core Period":

[0058] Call the conversion model of daily rainfall accumulation and leaf wetness (sub-model 2),

[0059] If LWA>=3 and the 3-day average temperature is between 10 and 30 degrees Celsius, then the output is the secondary infection probability P2, which obeys the normal distribution. where x is the average temperature of the day, with a mean of 20 degrees Celsius and a variance of 5.

[0060] The output infection stage is secondary infection;

[0061] If the secondary infection probability P2 exceeds 0.8, the output risk level is high risk;

[0062] The secondary infection probability P2 is between 0.6 and 0.8, and the output risk level is medium risk;

[0063] The secondary infection probability P2 is lower than 0.6, and the output risk level is low risk;

[0064] If the current phenological period is not one of the above listed phenological periods, then

[0065] The output infection probability is 0,

[0066] The output risk level is no risk.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] By constructing a phenological stage judgment model for the host of plum brown rot and a conversion model for daily rainfall accumulation and leaf humidity, and then calculating or fitting the input and output parameters of the plum brown rot prediction and risk level discrimination model, the probability (risk) of plum brown rot infection is predicted. Through the above method, an AI early warning monitoring system based on artificial intelligence is constructed to conduct online automatic monitoring of the microenvironmental dependence conditions of plum brown rot at each stage of occurrence and evolution, and to provide early warning, push alarm information, and guide early targeted prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] A patent application titled "An early warning method for plum brown rot integrating phenology and microenvironment perception" includes six drawings, the descriptions of which are as follows.

[0070] Figure 1 is a schematic diagram of the infection stages of plum brown rot (fungal disease);

[0071] Figure 2 shows the relationship between the probability of germination of plum brown rot fungus spores and effective accumulated temperature under different meteorological conditions;

[0072] Figure 3 is a schematic diagram of the time periods (dashed boxes) when critical infection conditions LWD and TLWD are reached;

[0073] Figure 4 is a diagram of the real-time monitoring data of the micro-weather station: temperature, relative humidity, wind speed and radiation intensity;

[0074] Figure 5 shows the relationship between the risk of plum brown rot (fungal disease) infection and the duration of leaf wetness and the temperature during the leaf wet period;

[0075] FIG6 is a functional flow diagram of the plum brown rot prediction and early warning model. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0077] The present invention provides an early warning method for plum brown rot that integrates phenology and microenvironmental perception, comprising the following steps:

[0078] Step 1: Define the process and stages of occurrence and evolution of specific pests and diseases;

[0079] As shown in Figure 1, empirical data were used to divide the entire process of plum brown rot infection into several specific stages, including: the infected but asymptomatic stage (incubation period), the infected and symptomatic stage, the infectious stage, the transmission stage, and the stage where diseased and rotten fruits appeared. Characteristic variables were defined and quantified.

[0080] Step 2: Establish the probabilistic relationship between each evolutionary stage of plum brown rot and the time of occurrence and duration, and establish a phenological stage judgment model for plum brown rot hosts (sub-model 1):

[0081] Define the probability distribution of the occurrence time and duration of each evolutionary stage of plum brown rot as the system information prior;

[0082] Measuring effective accumulated temperature is used to judge physiological status.

[0083] The lower temperature threshold and the upper temperature threshold are determined through observation and control experiments. The lower temperature threshold adopts the temperature at which development stops, and the upper temperature threshold adopts the temperature at which the growth or development speed begins to decline.

[0084] As shown in Figure 2, a relationship diagram between the probability of germination of plum brown rot fungal spores and the effective accumulated temperature under different meteorological conditions can be drawn.

[0085] The host phenological stage judgment model for plum brown rot (sub-model 1) includes two stages:

[0086] The first stage, model training and calibration:

[0087] Through 5 years of observation, the Julian days of the host's 12 phenological phase transitions were recorded. The Julian day of each phenological phase transition was averaged and rounded off by taking the average of 5 observation data to obtain the time range of each host phenological phase.

[0088] The daily maximum and minimum temperature data for the required altitude range were obtained using historical meteorological data from NOAA / NWS over the past five years. The cumulative effective accumulated temperature on the day of each phenological stage transition was calculated, using the minimum temperature as the effective accumulated temperature threshold. The effective accumulated temperature thresholds ATi (where i = 1, 2, 3…12) for each phenological stage transition of the host were obtained by taking the average of five observations and rounding them up.

[0089] In the second stage, the model predicts:

[0090] According to the forecast accuracy requirements, obtain the daily maximum and minimum temperature data of NOAA / NWS for the next 7 days or 7 months, and start the cumulative calculation from Julian day 1. If the cumulative effective accumulated temperature exceeds ATi, the i-th phenological period output is the current phenological period.

[0091] Step 3: Determine the microenvironmental conditions and duration for the occurrence and evolution of plum brown rot:

[0092] Identify variables highly correlated with plum brown rot, including climate and phenological conditions, soil and environmental conditions, and establish a conversion model between daily rainfall accumulation and leaf wetness (sub-model 2);

[0093] The conversion model between daily rainfall accumulation and leaf wetness (sub-model 2):

[0094] The leaf wetness threshold is calculated using daily accumulated rainfall: the minimum daily rainfall is PL = 12 mm / day. If the rainfall is lower than 12, it is considered that the daily rainfall cannot cause leaf wetness. If the daily rainfall exceeds PL for three consecutive days, the leaf wetness on the third day is set as susceptible condition and marked as WA. Otherwise, the state of the day is marked as WN. The number of consecutive WA days is LWA. If the WN state occurs during the accumulation process, the LWA is reset to zero until a new WA appears and the LWA accumulation starts again.

[0095] As shown in Figure 3, it is also necessary to determine the sufficient number of hours for leaf wetness (LWD) and the hourly temperature during the leaf wet period (TLWD). A schematic diagram of the time periods when critical infection conditions (LWD and TLWD) are reached should be drawn, and the time periods of susceptible conditions should be marked. Based on experience or historical data, the probability distribution of each relevant variable when plum brown rot occurs should be fitted. Based on the correlation assessment of the data, the joint probability distribution of each variable should be established as the system information prior.

[0096] Step 4, establish a state transition model:

[0097] The evolutionary stages of plum brown rot were defined as the state set of a Markov chain. The basic transition probability matrix was calculated based on microenvironmental conditions and duration. The conditions for pest outbreaks were used as the initial probability to establish a Markov state transition model for pest occurrence and evolution.

[0098] Step 5, monitoring data access:

[0099] Through the micro-meteorological station and soil detector, the time series data of microclimate environmental conditions closely related to the occurrence and evolution of plum brown rot were obtained, as shown in Figure 4. The data of air temperature, relative humidity, wind speed and radiation intensity were obtained through the real-time monitoring data of the micro-meteorological station.

[0100] Input parameters include:

[0101] The effective accumulated temperature for the development of plum brown rot pathogen (Monilia spp.) was calculated, and the lower and upper temperature thresholds were obtained. The threshold parameters were fitted with the values ​​of df, F, and the range of p based on the host's three-year observation experimental data, the highest temperature of the day, the lowest temperature of the day, and the rainfall of the day.

[0102] Starting date: Julian day 1;

[0103] End date: Julian day 245;

[0104] Output parameters include:

[0105] The infection stage is either primary or secondary;

[0106] Infection probability, a decimal between 0 and 1;

[0107] Risk level: 4 levels: none, low, medium and high risk.

[0108] Step 6, model parameter calibration:

[0109] The distribution parameters of each model variable are fitted by measured data and the value range is given. The calibration of the model is divided into two stages:

[0110] First, trend and pattern consistency. By conducting statistical analysis on the judgment data of indicators such as infection risk and infection stage, if there is no statistically significant difference between the predicted indicators and the actual indicators, it is considered that the model can capture the trend of disease occurrence and evolution and can also output a relatively reliable disease evolution pattern.

[0111] The second is the consistency of numerical indicators such as recall rate. By tracking the performance of the model, if a high prediction accuracy can be achieved in 2 to 3 seasons in the case of out-of-sample prediction, the task model can work stably.

[0112] Step 7: Model-based prediction and early warning:

[0113] Establish a process for predicting the probability (risk) of plum brown rot infection;

[0114] As shown in Figure 6, based on meteorological environmental data, a shallow network training microclimate prediction model was established to predict meteorological change data for 5 to 7 days. At the same time, key data such as the phenological period of the plum tree was obtained through the tree management model and related monitoring data. As shown in Figure 5, based on the critical values ​​and ranges of parameters highly correlated with plum brown rot obtained during the training and verification stages of the model, a parameterized disease prediction model was used to give the risk level of infection (0 to 5) and the possible infection stage (0 to 2) one week later.

[0115] The described plum brown rot infection probability (risk) prediction process:

[0116] Call the plum brown rot host phenology model (sub-model 1) to obtain the current phenological period:

[0117] If the current phenological period is "full bloom", "falling bloom" or "young fruit":

[0118] Call the conversion model of daily rainfall accumulation and leaf wetness (sub-model 2),

[0119] If LWA>=3 and the 3-day average temperature is between 10 and 25 degrees Celsius, the output is the primary infection probability P1, which obeys the normal distribution. where x is the average temperature of the day, with a mean of 15 degrees Celsius and a variance of 4.

[0120] The output infection stage is primary infection.

[0121] If the secondary infection probability P1 exceeds 0.8, the output risk level is high risk;

[0122] The secondary infection probability P1 is between 0.6 and 0.8, and the output risk level is medium risk;

[0123] The secondary infection probability P1 is lower than 0.6, and the output risk level is low risk.

[0124] If the current phenological period is "Hard Core Period":

[0125] Call the conversion model of daily rainfall accumulation and leaf wetness (sub-model 2),

[0126] If LWA>=3 and the 3-day average temperature is between 10 and 30 degrees Celsius, then the output is the secondary infection probability P2, which obeys the normal distribution. where x is the average temperature of the day, with a mean of 20 degrees Celsius and a variance of 5.

[0127] The output infection stage is secondary infection.

[0128] If the secondary infection probability P2 exceeds 0.8, the output risk level is high risk;

[0129] The secondary infection probability P2 is between 0.6 and 0.8, and the output risk level is medium risk;

[0130] The secondary infection probability P2 is lower than 0.6, and the output risk level is low risk.

[0131] If the current phenological period is not one of the above listed phenological periods, then

[0132] The output infection probability is 0,

[0133] The output risk level is no risk.

[0134] By constructing a phenological stage judgment model for the host of plum brown rot and a conversion model for daily rainfall accumulation and leaf humidity, and then calculating or fitting the input and output parameters of the plum brown rot prediction and risk level discrimination model, the probability (risk) of plum brown rot infection is predicted. Through the above method, an AI early warning monitoring system based on artificial intelligence is constructed to conduct online automatic monitoring of the microenvironmental dependence conditions of plum brown rot at each stage of occurrence and evolution, and to provide early warning, push alarm information, and guide early targeted prevention and control.

Claims

1. An early warning method for plum brown rot that integrates phenology and microenvironment perception, characterized in that: The following steps are involved: Step 1: define the process and stages of occurrence and evolution of specific pests and diseases; Step 2: Establish the probability relationship between each evolutionary stage of plum brown rot and the occurrence time and duration, and establish the sub-model 1 for judging the phenological stage of plum brown rot host: Define the probability distribution of the occurrence time and duration of each evolutionary stage of plum brown rot as the system information prior; Measuring effective accumulated temperature is used to determine physiological status; The lower temperature threshold and the upper temperature threshold are determined by observation and control experiments. The lower temperature threshold is the temperature at which development stops, and the upper temperature threshold is the temperature at which the growth or development speed begins to decrease. Draw a relationship between the probability of germination of plum brown rot fungus spores and effective accumulated temperature under different meteorological conditions; Step 3, determine the microenvironmental conditions and duration of the occurrence and evolution stages of plum brown rot: Identify variables that are highly correlated with the occurrence of plum brown rot, including climate and phenological conditions, soil and environmental conditions, and establish a conversion sub-model for daily rainfall accumulation and leaf surface wetness2; Step 4: Establish a state transition model: The evolution stages of plum brown rot were defined as the state set of Markov chain, the basic transition probability matrix was calculated based on microenvironmental conditions and duration, and the conditions of pest outbreak were used as the initial probability to establish a Markov state transition model for the evolution of pests and diseases. Step 5: Monitoring data access: Through micro-weather stations and soil detectors, we can obtain time series data of microclimate environmental conditions closely related to the occurrence and evolution of plum brown rot, and obtain data on temperature, relative humidity, wind speed and radiation intensity through real-time monitoring data from micro-weather stations; Step 6: Model parameter calibration: The distribution parameters of each variable in the model are fitted through measured data and the value range is given. The calibration of the model is divided into two stages: First, the consistency of trends and patterns. By conducting statistical analysis on the judgment data of indicators such as infection risk and infection stage, if there is no statistically significant difference between the predicted indicators and the actual indicators, it is believed that the model can capture the trend of disease occurrence and evolution and can also output a more reliable disease evolution model. The second is the consistency of numerical indicators such as recall rate. By tracking the operation effect of the model, if a high prediction accuracy can be achieved in 2 to 3 seasons in the case of out-of-sample prediction, the task model can work stably. Step 7, Model-based prediction and early warning: Establish a prediction process for the probability (risk) of plum brown rot infection; Based on the meteorological environmental data, a shallow network training microclimate prediction model was established to predict the meteorological change data for 5 to 7 days. At the same time, the key phenological data of the plum trees were obtained through the tree management model and related monitoring data. According to the critical values ​​and ranges of the parameters highly correlated with plum brown rot obtained during the training and verification stages of the model, the disease prediction model was parameterized to give the risk level of infection (0 to 5 levels) and the possible infection stage (0 to 2) one week later.

2. The early warning method for plum brown rot integrating phenology and microenvironment perception according to claim 1, characterized in that: In step 1, empirical data are used to divide the entire process of plum brown rot infection into several specific stages, including: the infected but asymptomatic stage (latent period), the infected and symptomatic stage, the infectious stage, the transmission stage, and the stage where diseased and rotten fruits appear in the fruit, and characteristic variables are defined and quantified.

3. The early warning method for plum brown rot integrating phenology and microenvironment perception according to claim 1, characterized in that: The establishment of sub-model 1 for judging the phenological stage of plum brown rot host is divided into two stages: The first stage, model training and calibration: Through 5 years of observation, the Julian days of the host's 12 phenological phase transitions were recorded; the Julian day of each phenological phase transition was averaged and rounded off by taking the average of 5 observation data to obtain the time range of each phenological phase of the host; The daily maximum and minimum temperature data at the required altitude range were obtained through the historical meteorological data of NOAA / NWS in the past five years; the minimum temperature was used as the effective accumulated temperature threshold to calculate the cumulative effective accumulated temperature on the day of each phenological stage transition; the effective accumulated temperature threshold ATi (where i=1, 2, 3…12) of each phenological stage transition of the host was obtained by taking the average of 5 observation data and rounding them off; In the second stage, the model predicts: According to the prediction accuracy requirements, obtain the daily maximum and minimum temperature data of NOAA / NWS for the next 7 days or 7 months, and start the cumulative calculation from Julian day 1. If the cumulative effective accumulated temperature exceeds ATi, the output i-th phenological period is the current phenological period.

4. The early warning method for plum brown rot integrating phenology and microenvironment perception according to claim 1, characterized in that: In step 3, the conversion sub-model 2 for the microenvironmental conditions and duration of plum brown rot occurrence and each evolution stage is a conversion model for daily rainfall accumulation and leaf surface humidity: The leaf wetness threshold is calculated using the daily accumulated rainfall: the minimum daily rainfall is PL = 12 mm / day. If it is lower than 12, it is considered that the rainfall on that day cannot cause the leaf surface to be wet. If the daily rainfall for three consecutive days exceeds PL, the leaf wetness on the third day is set as susceptible condition and marked as WA. Otherwise, the state of that day is marked as WN. The number of days of accumulated continuous WA is LWA. If WN state occurs during the accumulation process, LWA will be reset to zero until a new WA occurs and LWA accumulation starts again. It is also necessary to determine the sufficient number of hours for the leaf surface to be wet (LWD) and the hourly temperature of the leaf during the wet period (TLWD), make a schematic diagram of the time period for reaching the critical infection conditions LWD and TLWD, and mark the time period for susceptible conditions. Through experience or historical data, fit the probability distribution of each related variable when plum brown rot occurs, and based on the correlation evaluation of the data, establish the joint probability distribution of each variable as the system information prior.

5. The early warning method for plum brown rot integrating phenology and microenvironment perception according to claim 3, characterized in that: In step 2, the second stage, the model prediction includes input and output parameters: The input parameters are: Calculate the effective accumulated temperature for the development of plum brown rot pathogen (Monilia spp.), obtain the lower and upper temperature thresholds, and the threshold parameters are fitted according to the host's three-year observation experimental data to obtain the values ​​of df, F, and the range of p, the highest temperature of the day, the lowest temperature of the day, and the rainfall of the day. Starting date: Julian day 1; End date: Julian day 245; Output parameters: The infection stage, which can be primary or secondary; Infection probability, a decimal between 0 and 1; Risk level: 4 levels: none, low, medium and high risk.

6. The early warning method for plum brown rot integrating phenology and microenvironment perception according to claim 1, characterized in that: In step 7, Call the plum brown rot host phenological model (sub-model 1) to obtain the current phenological period: If the current phenological period is "full bloom" or "falling bloom" or "young fruit stage": Call the conversion model of daily rainfall accumulation and leaf surface wetness (sub-model 2), If LWA>=3 and the 3-day average temperature is between 10 and 25 degrees Celsius, the output is the primary infection probability P1, which follows a normal distribution. A random number, where x is the average temperature of the day, with a mean of 15 degrees Celsius and a variance of 4; The output infection stage is primary infection; The secondary infection probability P1 exceeds 0.8, and the output risk level is high risk; The secondary infection probability P1 is between 0.6 and 0.8, and the output risk level is medium risk; The secondary infection probability P1 is lower than 0.6, and the output risk level is low risk; If the current phenological period is "Hard Core Period": Call the conversion model of daily rainfall accumulation and leaf surface wetness (sub-model 2), If LWA>=3 and the 3-day average temperature is between 10 and 30 degrees Celsius, the output is the secondary infection probability P2, which follows a normal distribution. A random number, where x is the average temperature of the day, with a mean of 20 degrees Celsius and a variance of 5; The output infection stage is secondary infection; The secondary infection probability P2 exceeds 0.8, and the output risk level is high risk; The secondary infection probability P2 is between 0.6 and 0.8, and the output risk level is medium risk; The secondary infection probability P2 is lower than 0.6, and the output risk level is low risk; If the current phenological period is not one of the above listed phenological periods, the output infection probability is 0 and the output risk level is no risk.

Images