Kiwi fruit brown spot prevention and control decision-making system

Abstract

The invention belongs to the technical field of agricultural disease prevention and control, and particularly relates to a kiwi fruit brown spot prevention and control decision-making system which comprises a data acquisition module, a suitable-to-life division module, an epidemic prediction module, a loss evaluation module and a prevention and control decision-making module. Multi-source data are acquired through the data acquisition module, and suitable-to-life division is performed by using the MaxEnt model, so that different risk level areas of the brown spot of the kiwi fruit can be accurately positioned, a visual disease occurrence risk map is provided for kiwi fruit planters, and the risk level of the brown spot of the kiwi fruit is improved. The problem of blindness in prevention and treatment caused by insufficient understanding of disease occurrence rules in a traditional prevention and treatment method is effectively solved; the prevalence prediction module adopts a BP neural network model to realize accurate prediction of the prevalence stage of the kiwi fruit brown blotch, and the short-term prediction accuracy and the middle-term prediction accuracy of the prevalence stage of the kiwi fruit brown blotch reach 90.26% and 90.11% respectively; the loss evaluation module maps the disease index into a fruit quality comprehensive score based on a Boltzmann model, presets a disease index threshold value 36.14, and can trigger economic loss early warning in time when the disease index exceeds the threshold value.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of agricultural disease prevention and control, and specifically relates to a kiwifruit brown spot disease prevention and control decision system. The system integrates multiple modules such as data collection, suitable area division, prevalence prediction, loss assessment and prevention and control decision, aiming to provide comprehensive, scientific and accurate brown spot disease prevention and control decision support for kiwifruit growers and agricultural managers, so as to effectively reduce the impact of the disease on kiwifruit yield and quality and ensure the sustainable development of the kiwifruit industry. BACKGROUND

[0002] As a high-value fruit, the kiwifruit planting industry has developed rapidly in recent years. However, kiwifruit brown spot disease, as a major fungal disease, poses a serious threat to kiwifruit yield and quality. The prevalence of brown spot disease not only causes premature leaf drop, affecting photosynthesis and reducing fruit yield, but also leads to decreased fruit quality, resulting in significant economic losses.

[0003] Traditional methods of preventing and controlling kiwifruit brown spot disease mainly rely on the spraying of chemical pesticides. However, due to the lack of in-depth understanding and accurate prediction of disease occurrence patterns, there are often problems such as improper timing of prevention and control, excessive use of pesticides, etc., which not only lead to pesticide residues and environmental pollution, but also increase production costs and easily cause pathogens to develop drug resistance. Therefore, developing a system that can accurately predict disease occurrence, scientifically assess disease losses and develop reasonable prevention and control strategies is of great significance for improving the prevention and control effect of kiwifruit brown spot disease, reducing prevention and control costs, reducing environmental pollution and ensuring the sustainable development of the kiwifruit industry.

[0004] In terms of disease prediction, although some prediction models based on meteorological data and historical disease data have been proposed, these models often have low prediction accuracy and insufficient timeliness, making it difficult to meet the demand for accurate disease prediction in actual production. In addition, existing disease loss assessment methods are mostly rough and do not fully consider the multi-faceted impact of disease on fruit quality, making it difficult to provide accurate economic loss assessment for growers.

[0005] In actual field management, kiwifruit growers and agricultural managers urgently need a prevention and control decision system that takes into account disease occurrence patterns, environmental factors and specific orchard management conditions to guide them to reasonably arrange prevention and control measures and effectively prevent and control kiwifruit brown spot disease. SUMMARY

[0006] In view of this, the present application aims to provide a kiwifruit brown spot disease prevention and control decision system, which aims to provide comprehensive disease prevention and control solutions for the kiwifruit industry by integrating multi-source data and using advanced prediction models and assessment methods.

[0007] The inventor provides a technical scheme for solving the above technical problems through long-term exploration and trial, and multiple experiments and efforts, that is, a kiwifruit brown spot disease prevention and control decision system, comprising: A data acquisition module is configured to acquire meteorological data, geographic information data, and field pathogen monitoring data. A suitable area division module is connected to the data acquisition module, and is configured to process climate variables and species distribution data based on a MaxEnt model to output a spatial distribution map of high, medium, low, and unsuitable areas for the kiwifruit brown spot disease. A prevalence prediction module is configured to internally store a BP neural network model, and to output a disease prevalence stage prediction result after inputting key meteorological factors. A loss assessment module is configured to map a disease index to a comprehensive fruit quality score based on a Boltzmann model, and to preset a disease index threshold of 36.14. A prevention and control decision module is configured to generate a zoned management plan based on the results of the suitable area division, the prevalence prediction, and the loss assessment.

[0008] Compared with the prior art, the present application has the following advantages: The present application can accurately locate different risk level areas of the kiwifruit brown spot disease by acquiring multi-source data through the data acquisition module and dividing the suitable areas by using the MaxEnt model, and can provide a disease occurrence risk map for kiwifruit growers, thereby effectively solving the problem of blindness in traditional prevention and control methods due to insufficient understanding of the disease occurrence regularity.

[0009] The prevalence prediction module uses the BP neural network model to accurately predict the prevalence stage of the kiwifruit brown spot disease, and the short-term and medium-term prediction accuracy rates reach 90.26% and 90.11%, respectively, which provides a sufficient time window for growers to develop a scientific and reasonable prevention and control plan in advance, and avoids missing the best prevention and control opportunity due to untimely or inaccurate prediction.

[0010] The loss assessment module maps the disease index to a comprehensive fruit quality score based on the Boltzmann model, and presets a disease index threshold of 36.14, so that the economic loss warning can be triggered in time when the disease index exceeds the threshold, and the growers can intuitively understand the potential impact of the disease on fruit quality and yield, thereby achieving accurate prevention and control and reducing economic losses caused by the disease.

[0011] On the basis of the above technical scheme, the present application can also be improved as follows: Further, the data acquisition module performs: Monitoring the dynamics of pathogenic spores in the air by using a spore trap. The pathogen is detected by using a qPCR detection system and using primers CC1F: 5'-CAAGTGTTGGGCGTCTGTCC-3' and CC1R: 5'-TGTGGGGGGGCTCATGGTG-3'; Field disease fallen leaves are collected every 15 days to monitor overwintering of the pathogen.

[0012] Compared with the prior art, the beneficial effects of the above further technical solutions are: The data acquisition module can more comprehensively and accurately obtain pathogen dynamic information, and provide more reliable data support for early warning and prevention and control decision of the kiwi brown spot disease.

[0013] On the basis of the above technical solutions, the present application can also be improved as follows: Further, the overwintering monitoring of the pathogen comprises: Identifying the ground disease fallen leaves as the initial infection source; Detecting the fallen leaf pathogen load from December to the following April; Implementing disease fallen leaf removal or soil deep ploughing to a depth of 10-20 cm after fallen leaves.

[0014] Compared with the prior art, the beneficial effects of the above further technical solutions are: The present application can effectively reduce the overwintering base of the pathogen, cut off the initial infection source of the disease, significantly reduce the risk of occurrence of the kiwi brown spot disease, and provide a scientific basis and effective means for source prevention and control of the disease.

[0015] On the basis of the above technical solutions, the present application can also be improved as follows: Further, the suitable area division module performs the following operations: Dividing the suitable area into high suitable area, medium suitable area, low suitable area and unsuitable area; Marking the high suitable area and the high suitable core planting area.

[0016] Compared with the prior art, the beneficial effects of the above further technical solutions are: The present application enables the kiwi grower to more accurately identify the high-risk area of the disease, thereby optimizing the planting layout and prevention and control resource allocation, and improving the disease prevention and control efficiency.

[0017] On the basis of the above technical solutions, the present application can also be improved as follows: Further, the BP neural network model of the prevalence prediction module comprises: The input layer receives 4 key meteorological factors: extreme maximum temperature in mid-June, extreme minimum temperature in early May, extreme maximum rainfall in late April, and rainfall in early August; The number of hidden layer neurons in the logistic growth phase is 3, and the number of hidden layer neurons in the decline phase is 4; Output layer: predicted logistic growth phase and decline phase disease index; Training parameters: Lenberg-Marquardt algorithm, 70% training set, 15% validation set, and 15% test set.

[0018] Compared with the prior art, the beneficial effects of the above further technical solutions are: The epidemic prediction module of the present application can accurately predict the disease index in the logistic growth phase and the decline phase of kiwifruit brown spot disease by innovatively designing the input layer, hidden layer and output layer of the BP neural network model and using the Lenberg-Marquardt algorithm for training, providing high timeliness and high precision disease epidemic prediction for growers, and effectively improving the scientificity and timeliness of prevention and control decisions.

[0019] On the basis of the above technical solutions, the present application can also be improved as follows: Further, the epidemic prediction module outputs: Short-term prediction: disease index prediction 10 days in advance, accuracy ≥ 90.26%; Medium-term prediction: disease index prediction 30 days in advance, accuracy ≥ 90.11%.

[0020] Compared with the prior art, the beneficial effects of the above further technical solutions are: The epidemic prediction module of the present application can output disease index prediction results, providing more accurate and timely disease epidemic trend prediction for kiwifruit growers, so that they can take effective prevention and control measures in advance, reducing the risk and loss of disease outbreaks.

[0021] On the basis of the above technical solutions, the present application can also be improved as follows: Further, the training method of the BP neural network model comprises: Normalizing meteorological data: ; wherein X norm represents the normalized value, X is the original value, X max and X min are the maximum and minimum values in the data set, respectively; Stepwise regression is used to screen key meteorological factors; The number of hidden layer neurons is adjusted until the mean square error is less than 0.01.

[0022] Compared with the prior art, the beneficial effects of the above further technical solutions are: The application significantly improves the training efficiency and prediction accuracy of the BP neural network model by normalizing the weather data, stepwise regression screening of key weather factors and dynamic adjustment of the number of hidden layer neurons, ensures that the accuracy of short-term and medium-term prediction reaches 90.26% and 90.11% or more, and provides reliable prediction support for accurate prevention and control of brown spot disease of kiwifruit.

[0023] Based on the above technical solutions, the application can also be improved as follows: Further, the loss evaluation module executes: Through principal component analysis, the seven quality indicators of single fruit weight, fruit shape index, hardness, soluble solid content, dry matter content, diseased fruit rate and shelf life are reduced in dimension; The data is standardized by using the membership function method; The Boltzmann model of disease index and quality comprehensive score is established: ; Wherein y is the comprehensive score, i.e., fruit quality, and x is the disease index; When the disease index is greater than 36.14, the economic loss warning is triggered.

[0024] Compared with the prior art, the beneficial effects of the above further technical solutions are: The loss evaluation module reduces the dimension of multiple fruit quality indicators through principal component analysis, standardizes the data by using the membership function method, accurately quantifies the relationship between the disease index and the comprehensive score of fruit quality by using the Boltzmann model, and sets a clear disease index threshold to trigger the economic loss warning, providing a more scientific and accurate loss evaluation method for kiwifruit growers, and effectively improving the economic rationality and feasibility of the prevention and control decision.

[0025] Based on the above technical solutions, the application can also be improved as follows: Further, the prevention and control decision module generates a scheme including: In high-adaptation areas, disease-resistant varieties are compulsorily introduced; In the exponential growth period, the first pesticide application is started; When the incidence rate reaches 5%, chemical control is performed no later than June 19; When the disease index is greater than 36.14, emergency prevention and control is started.

[0026] Compared with the prior art, the beneficial effects of the above further technical solutions are: The application provides a set of accurate and time node-specific prevention and control strategies, effectively improving the prevention and control efficiency and effect, and reducing the damage of kiwifruit brown spot disease to yield and quality.

[0027] Based on the technical scheme, the application can be further improved as follows. Further, the chemical control scheme comprises: Spraying 200g / L fluopicolide·difenoconazole during the dormancy period; The interval between two sprays is 10 days, and the spraying is performed twice.

[0028] Compared with the prior art, the beneficial effects of the above further technical scheme are: The application provides clear and efficient spraying guidance for chemical control of kiwifruit brown spot disease, effectively improves the effect of chemical control, reduces the amount of pesticide used and the number of spraying times, reduces the risk of pesticide residues and environmental pollution, helps to delay the generation of pathogen resistance, and ensures the sustainable development of the kiwifruit industry. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to more clearly illustrate the technical scheme of the embodiments of the application, the drawings needed in the embodiments will be briefly introduced as follows, and it should be understood that the following drawings only show some embodiments of the application, and therefore should not be regarded as a limitation on the scope, and other related drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0030] Figure 1 PCR detection results of the primer CC1F / CC1R pair on Corynespora cassiicola.

[0031] Figure 2 Distribution points of kiwifruit brown spot disease samples.

[0032] Figure 3 Response curve between kiwifruit brown spot disease distribution probability and climate variables.

[0033] Figure 4 Regularized training scores of climate variables affecting the distribution of kiwifruit brown spot disease.

[0034] Figure 5 Potential distribution simulation of kiwifruit brown spot disease in Sichuan.

[0035] Figure 6 BP neural network model structure diagram of the logistic growth period.

[0036] Figure 7 BP neural network model structure diagram of the recession period.

[0037] Figure 8 Neural network model training process of the logistic growth period.

[0038] Figure 9 Correlation of the training function of the logistic growth period.

[0039] Figure 10 is a neural network model training process in a recession period.

[0040] Figure 11 is the correlation of the logistic growth period training function. DETAILED DESCRIPTION

[0041] The following will be described in conjunction with specific embodiments.

[0042] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work under the premise that the embodiments in the present application fall within the scope of protection of the present application. Therefore, the following detailed description of the embodiments of the present application is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application.

[0043] In the present application, all the equipment and raw materials can be purchased from the market or commonly used in the industry, unless specified. The methods in the following examples are all conventional methods in the art, unless specified.

[0044] The kiwi brown spot disease prevention and control decision system described in this embodiment is illustrated by the epidemic and prevention and control data of kiwi brown spot disease in Sichuan Province. The pathogenic bacteria of kiwi brown spot disease is Corynespora cassiicola. C. cassiicola ).

[0045] The kiwi brown spot disease prevention and control decision system comprises: A data acquisition module for acquiring meteorological data, geographic information data and field pathogen monitoring data; A suitable area division module connected to the data acquisition module, which processes climate variables and species distribution data based on a MaxEnt model, and outputs spatial distribution maps of high, medium, low and unsuitable areas for kiwi brown spot disease; A prevalence prediction module with a built-in BP neural network model that outputs disease prevalence stage prediction results after inputting key meteorological factors; A loss assessment module that maps disease index to fruit quality comprehensive score based on a Boltzmann model, and presets a disease index threshold of 36.14; A prevention and control decision module that generates a zonal management plan based on the results of suitable area division, prevalence prediction and loss assessment.

[0046] 1. Data acquisition module Data acquisition module is the key component of the kiwifruit brown spot disease prevention and control decision system. Its main function is to obtain various types of data related to the occurrence of kiwifruit brown spot disease, and to provide reliable data support for subsequent disease prediction, loss assessment and control decision. This module focuses on collecting meteorological data, geographic information data and field pathogen monitoring data to fully understand the various factors affecting the occurrence and development of the disease.

[0047] 1.1. Collection of meteorological data Meteorological data is obtained from the Worldclim database, which is based on global weather record information and integrates interpolated raster data. The collected meteorological data includes: annual average temperature, monthly average diurnal temperature difference, isothermality, standard deviation of seasonal temperature variation, maximum temperature of the warmest month, minimum temperature of the coldest month, annual temperature variation range, average temperature of the wettest quarter, average temperature of the driest quarter, average temperature of the warmest quarter, average temperature of the coldest quarter, annual precipitation, precipitation in the wettest month, precipitation in the driest month, precipitation variation coefficient, precipitation in the wettest quarter, precipitation in the driest quarter, precipitation in the warmest quarter, precipitation in the coldest quarter, monthly average precipitation, monthly average temperature, monthly average minimum temperature, monthly average maximum temperature, etc.

[0048] The spatial resolution of these meteorological data is 2.5 arc-minutes, and they are stored in TIFF format raster files. The format conversion function of ArcGIS is used to convert the TIFF format variable layers to ASCⅡ format layers required by MaxEnt, providing basic data for subsequent suitable area analysis.

[0049] 1.2. Collection of geographic information data The collection of geographic information data covers the geographic coordinates (longitude, latitude and elevation) of kiwifruit planting areas and other information such as planting varieties. Detailed geographic information of the main kiwifruit producing areas in Sichuan Province was obtained through field surveys. The survey areas include Chengdu City, Ya'an City, Guangyuan City, Mianyang City, Deyang City and Meishan City, and the survey was conducted at the end of the disease peak period. The survey used a 5-point sampling method to record the geographic information of each orchard, including longitude, latitude and elevation, as well as cultivation varieties and disease occurrence. For distribution points without specific longitude and latitude, the corresponding coordinate information was extracted from the Global Geographic Information Integration Database GeoName and Baidu Coordinate Capture System.

[0050] 1.3. Collection of field pathogen monitoring data The collection of field pathogen monitoring data is mainly carried out through the following methods: 1.3.1. Spore trap monitoring The spore trap was installed at different locations and heights within the kiwifruit orchard to comprehensively monitor the dynamic changes of pathogenic spores in the air. The instrument automatically collects air samples, captures spores in the air through adhesive tape, and sets the collection period to collect at regular intervals, such as in the morning and evening, two active periods of pathogenic spore diffusion, each for 15-30 minutes to ensure effective capture of spore transmission dynamics.

[0051] 1.3.2. qPCR detection system The qPCR detection system was used to accurately detect the pathogen of kiwifruit brown spot disease, using specific primers CC1F: 5'-CAAGTGTTGGGCGTCTGTCC-3' and CC1R: 5'-TGTGGGGGGGCTCATGGTG-3', with an amplification product size of 137bp.

[0052] Periodically collect kiwifruit leaf, branch and other samples from the field, extract DNA from the samples according to the standard DNA extraction method, and use it as the template for the qPCR reaction. Reaction system (25μl): Max Buffer 12.5μl, dNTP Mix 0.5μl, Polymerase 0.5μl, upstream and downstream primers 1μl each, template 1μl, ddH2O 8.5μl. Reaction program: pre-denaturation, 95℃, 3min; denaturation, 95℃, 15s, annealing, 60℃, 15s, extension, 72℃, 30s, 30 cycles; complete extension, 72℃, 5min. Take 5μl of the obtained PCR product and detect it by 1% agarose gel electrophoresis.

[0053] The generation of specific amplification products during the PCR reaction process is monitored in real-time using fluorescence signals, thereby rapidly and accurately detecting the presence or absence of pathogens and their relative content. The detection results are quantitatively analyzed by standard curve method, and the pathogen content in the sample is quantified as a specific numerical value, providing a basis for the occurrence degree assessment of the disease.

[0054] The results of PCR detection specificity of the primer CC1F / CC1R on Corynespora cassiicola are shown in Figure 1 . Figure 1 Among them, M is 2000bp marker, 1 and 2 are Corynespora cassiicola , 3-8 are Botryosphaeria dothidea, Botrytis cinerea , Colletotrichum gloeosporioides , Diaporthe actinidiae, Alternaria alternata , Exserohilum turcicum , 9 and 10 are negative controls.

[0055] 1.3.3. Pathogen overwintering monitoring By analyzing the potential overwintering sites of the pathogen in the field, such as diseased leaves, soil, buds, and tree bark, it was found that the main overwintering site of the pathogen was diseased leaves in the field.

[0056] The overwintering monitoring of the pathogen began in December of each year and continued until April of the following year. During this period, the ground diseased leaves in the kiwifruit orchard were observed and recorded regularly to identify the possibility of these leaves as the initial source of infection. Every 15 days, samples of diseased leaves in the field were collected and taken back to the laboratory for detailed isolation and culture of the pathogen. By dilution plating or tissue isolation method, the diseased leaf tissues were inoculated on suitable culture medium and incubated in a constant temperature incubator for a certain period of time. The growth and morphological characteristics of the pathogen were observed and recorded, and the amount of pathogen in the fallen leaves was counted. Based on the monitoring results, the overwintering site and survival of the pathogen were determined, providing key information for predicting the occurrence trend of the disease in the coming year.

[0057] 1.3.4. Diseased leaf treatment After the kiwifruit leaves fall, timely measures should be taken to remove the diseased leaves and collect them for harmless treatment, such as burning or deep burial. For soil deep plowing treatment, professional agricultural machinery equipment is used to plow the soil to a depth of 10-20 cm, and by turning the soil, part of the diseased leaves are buried deep in the soil, destroying the overwintering environment of the pathogen, reducing the number of initial infection sources in the coming year, and reducing the risk of disease occurrence.

[0058] 2. Suitable area division module The suitable area division module connects the data collection module and processes the climate variables and species distribution data based on the MaxEnt model to output the spatial distribution map of different suitable grades of kiwifruit brown spot disease, providing key geographical spatial reference for kiwifruit planting planning and disease prevention and control.

[0059] 2.1. Data preparation Climate variable data Climate data for Sichuan Province was obtained from the WorldClim database, including 19 bioclimatic factors, monthly average precipitation, monthly average temperature, and maximum and minimum temperature, etc. Climate indicators, with a spatial resolution of 2.5 arc-minutes, provided in TIFF format raster files.

[0060] Using the format conversion function of ArcGIS software, the TIFF format climate variable layer was converted to the ASCⅡ format layer required by MaxEnt.

[0061] Species distribution data Integrate multi-channel data sources, including published papers, local plant protection station issued pest forecast report, agricultural weather service coordination service report and inquiry local plant protection staff, etc. Collect the distribution records of kiwifruit brown spot in Sichuan area.

[0062] Use the buffer analysis function of ArcGIS to calculate the distance between the grid center and the distribution points. Only one record with the closest distance to the center is retained in the grid. Import the above brown spot distribution records into an Excel table. The north latitude and east longitude are positive, and the south latitude and west longitude are negative. Record and save them in the order of species name (Species), longitude (Longitude), and latitude (Latitude). After the above program screening, 225 valid distribution points are obtained, see Figure 2 , which is used to build MaxEnt model and convert species distribution data into CSV file required by MaxEnt software.

[0063] 2.2. Model building Data integration and input Integrate the collected kiwifruit brown spot distribution data and climate variable data into MaxEnt software. Create a climate response curve to analyze the relationship between each climate variable and the probability of kiwifruit brown spot existence. Draw a prediction map. Select the jackknife method to measure the importance of variables. Set the random test percentage to 25% and repeat 10 times. The rest of the parameters are the default values of the model. Set the output path to build the model.

[0064] Distribution data is recorded in the form of species name, longitude and latitude, and climate data is provided in ASCⅡ format.

[0065] Model training and running Run the MaxEnt model to calculate the potential distribution probability of kiwifruit brown spot under different climate conditions using the maximum entropy algorithm.

[0066] The model generates a climate response curve based on the input distribution data and climate variables, analyzing the relationship between each climate variable and the probability of kiwifruit brown spot existence.

[0067] Optimize model performance and improve prediction accuracy by repeated training and parameter adjustment.

[0068] 2.3. Results output and analysis Suitable grade division Use the format conversion tool of ArcGIS software to convert the ASCⅡ format file output by MaxEnt software to Raster format file.

[0069] Taking the administrative map of Sichuan Province as the base map, the potential suitable distribution map of kiwifruit brown spot in Sichuan was extracted. In this embodiment, see Figure 5 The suitable grade was divided into high suitable area (P ≥ 0.66, red), medium suitable area (0.33 ≤ P <0.66, orange), low suitable area (0.05 ≤ P < 0.33, yellow) and unsuitable area (P < 0.05, white), and different colors were used for display, directly presenting the suitable degree distribution of kiwifruit brown spot in different regions of Sichuan Province.

[0070] Using the statistical analysis function of ArcGIS software, the distribution area of each city (state) and each district (county) was calculated, providing a quantitative basis for kiwifruit planting planning and disease control.

[0071] Results verification To verify the accuracy of the prediction results of the MaxEnt model, the supplementary investigation distribution records were introduced for field test. The actual distribution points were mapped to the reclassified map, the corresponding suitable index was extracted by ArcGIS, and compared with the actual occurrence grade of the disease. If the grades are the same, it is considered that the prediction is accurate.

[0072] The accuracy of the model was calculated to evaluate the reliability of the model. In this study, the accuracy of the model reached 84.44%, indicating that the model had high credibility and could provide scientific basis for the suitable area division of kiwifruit brown spot.

[0073] 2.4. Key climate variable screening Jackknife test Using the Jackknife test function in MaxEnt software, each climate variable was measured respectively to evaluate the contribution rate and importance of each variable to the model.

[0074] The importance and contribution rate of environmental variables were determined by the jackknife test method, and variables with a contribution rate less than 1% were removed. The attribute values of the climate variables of the 225 distribution points were extracted using the extraction and analysis tool of ArcGIS software, the Pearson correlation between the climate variables was calculated using SPSS 21.0 software, and variables with strong correlation were removed. Considering the relationship between kiwifruit brown spot and meteorological factors, the variables were screened. According to the results of the jackknife test, the key climate variables that had important influence on the distribution of kiwifruit brown spot were determined, and a total of 8 variables, including isothermality (bio3), April precipitation (prec4), June precipitation (prec6), August precipitation (prec8), October average temperature (tavg10), July maximum temperature (tmax7), June minimum temperature (tmin6), and August minimum temperature (tmin8), were selected to establish the final model. The Pearson correlation analysis results of the climate variables affecting the distribution of kiwifruit brown spot are shown in Table 1. Table 1 Pearson correlation analysis results of climate variables affecting the distribution of kiwifruit brown spot The regularization training score results of the climate variables affecting the distribution of kiwifruit brown spot are shown in Figure 3 .

[0075] Response curve analysis Draw the response curve between the key climate variables and the probability of kiwifruit brown spot distribution, see Figure 4 , and analyze the suitability of kiwifruit brown spot to different climate conditions.

[0076] Through the response curve, determine the suitable range and optimal value of each climate variable, and provide a reference for the climate suitability evaluation of kiwifruit planting.

[0077] 2.5. Case application and display Taking Sichuan Province as an example, the actual application effect of the kiwifruit brown spot suitable area division module is shown. According to the prediction of the MaxEnt model, see Figure 5 , the high-suitability area of kiwifruit brown spot in Sichuan Province is mainly located in the eastern part of Chengdu City, the central and western part of Ya'an City, the southern part of Yibin City, the central part of Leshan City, and the eastern part of Meishan City, with a total area of 21849.83 km 2, accounting for 4.49% of Sichuan's land area. Among them, Chengdu and Ya'an have the largest proportion of high-adaptability areas, accounting for 28.12% and 20.94% respectively. Dazhou, Ganzi, Guang'an, Nanchong, Neijiang, Panzhihua, Suining, Ziyang and Zigong have no high-adaptability areas. Guangyuan, Yibin and Mianyang have more than 10% of their areas in the medium-adaptability area. Guang'an, Panzhihua, Suining and Ziyang have no medium-adaptability areas. Low-adaptability areas are widely distributed in the province, except for Panzhihua, they are distributed in the other 20 cities (states). Ziyang is a low-adaptability area. The largest area of unsuitable area is 318185.83km 2 , mainly distributed in Ganzi, Aba and Liangshan, with Ganzi accounting for the largest proportion, nearly 50%; Panzhihua is a whole unsuitable area for brown spot of kiwifruit.

[0078] Table 2 Potential distribution area prediction of brown spot of kiwifruit (key counties) Note: The unit of area is km 2 , and the unit of area ratio is %.

[0079] Through the test of 45 supplementary investigation distribution points on the model, 38 points were simulated accurately. The prediction accuracy of the model is 84.44%, and the simulation result of the model has high reliability. Among the 7 inaccurate simulation distribution points, all points are over-fitted, and the predicted adaptability level is higher than the actual occurrence of the disease.

[0080] 3. Epidemic prediction module The epidemic prediction module is a key component of the kiwifruit brown spot disease prevention and control decision system, aiming to accurately predict the future epidemic trend of the disease based on historical disease data and meteorological information.

[0081] 3.1. Data collection and processing Disease data collection: Collect disease data from monitoring points in Sichuan's main kiwifruit producing areas for many years, including disease onset time, disease index, disease leaf rate, etc.

[0082] Ensure that the data covers the dynamic changes of disease in different years, different regions and different planting conditions.

[0083] Meteorological data collection: Obtain meteorological data within the corresponding time range of disease data, including temperature, humidity, rainfall, sunshine hours and other key meteorological factors.

[0084] Meteorological data should have high spatial and temporal resolution to accurately reflect the changes in meteorological conditions during the disease epidemic.

[0085] 3.2. BP neural network model construction Model structure design, see Figure 6 and Figure 7 .

[0086] Input layer: contains 4 neurons, corresponding to 4 key meteorological factors: extreme maximum temperature in mid-June, extreme minimum temperature in early May, extreme maximum rainfall in late April, and rainfall in early August.

[0087] Hidden layer: set 3 neurons for logistic growth period and 4 neurons for decline period.

[0088] Output layer: output the predicted disease index.

[0089] Model training: use the collected historical disease data and meteorological data to train the model, and use Lenberg-Marquardt algorithm for optimization.

[0090] Divide the data set into 70% training set, 15% validation set and 15% test set.

[0091] Normalize the meteorological data to eliminate the dimensional difference between different variables and improve the training efficiency and accuracy of the model. The normalized meteorological data of different epidemic stages as input samples, and the corresponding disease index as output samples, use MATLAB R2018b software to establish the prediction model.

[0092] , where Xnorm represents the normalized value, X is the original value, Xmax and Xmin are the maximum and minimum values in the data set respectively.

[0093] During training, the number of hidden layer neurons and network parameters are constantly adjusted until the mean square error is less than 0.01 to ensure the prediction accuracy of the model.

[0094] Import the sorted data sample set into the software and save it in the mat file. Then open the neural network toolbox, put the input samples into Inputs and the output samples into Targets. Select the proportion of training sample set as 70%, validation sample set as 15%, and test sample set as 15%. Select different number of hidden layer neurons for training, and determine the best number of neurons according to the mean square error (MSE) and coefficient of determination (R 2 ) of model simulation. Select Lenberg-Marquardt algorithm for operation, get the training result, then save the function for prediction. Import the input data for prediction into the software, use the simulation function sim to run the saved function model for calculation, and get the predicted value after inverse normalization of the calculated value.

[0095] The neural network model function for the logistic growth phase is implemented using MATLAB scripts, which creates a two-layer feedforward neural network for processing time series data for prediction tasks (specifically, the logistic growth phase model). The core functionality is described as follows: A. Network structure Input layer: 4 nodes (corresponding to 4-dimensional input features) Hidden layer: 3 nodes (using the hyperbolic tangent activation function) Output layer: 1 node (linear activation function) % Layer 1 (hidden layer) b1 = [-1.986; -0.080; -0.242]; % Bias vector (3x1) IW1_1 = [2.155, 0.607, -0.390, 0.719;...] % Input weights (3x4) % Layer 2 (output layer) b2 = 0.960; % Bias scalar LW2_1 = [0.683, 0.531, 0.934]; % Hidden layer to output layer weights (1x3) B. Data processing flow (1) Input preprocessing Xp1 = mapminmax_apply(X{1,ts}, x1_step1); % Execution: % y = (x - xoffset) * gain + ymin % Parameters: % xoffset = [0;0;0;0], gain=[2;2;2;2], ymin=-1 Normalize the original input to the [-1, 1] interval (2) Forward propagation a1 = tansig_apply(b1 + IW1_1 * Xp1); % Hidden layer output (using tanh activation) a2 = b2 + LW2_1 * a1; % Output layer (linear activation) (3) Output post-processing Y{1,ts} = mapminmax_reverse(a2, y1_step1); % Execution: x = (y - ymin) / gain + xoffset % Parameters: % ymin=-1, gain=2, xoffset=0 Map network output from [-1,1] back to original data range.

[0096] C. Key Features (1) Time series support: Input X is a 1xTS cell array (TS is the number of time steps) Data format at each time step: Qx4 matrix (Q is the number of samples) Output Y is a 1xTS cell array (each element is Qx1) (2) Activation function function a = tansig_apply(n) a = 2. / (1+exp(-2*n)) - 1; % equivalent to tanh(n) (3) Independent time step processing: for ts=1:TS % Each time step is calculated independently.

[0097] The neural network model function in the recession period uses MATLAB scripts to implement a two-layer feedforward neural network for multiple input single output (MISO) prediction tasks. The main function is to map 4 input features to 1 output value. The core function description is as follows: A. Network architecture Input layer: 4 nodes (corresponding to 4 input features) Hidden layer: 4 neurons (using hyperbolic tangent activation function) Output layer: 1 neuron (linear activation) B. Data processing flow (1) Input preprocessing: Xp1 = mapminmax_apply(X{1,ts},x1_step1); Perform normalization operation: (x - xoffset) * gain + ymin Offset xoffset = [0;0;0;0], gain gain = [2;2;2;2] Map input data to [-1, 1] range (2) Hidden layer calculation: a1 = tansig_apply(repmat(b1,1,Q) + IW1_1*Xp1); Weight matrix IW1_1 (4 x 4) Bias vector b1 (4 x 1) Using hyperbolic tangent activation function: a = 2 / (1+exp(-2n)) - 1 (3) Output layer calculation: a2 = repmat(b2,1,Q) + LW2_1*a1; Weight vector LW2_1 (1 x 4) Bias scalar b2 = 0.1022 Linear output (no activation function) (4) Post-processing of output: Y{1,ts} = mapminmax_reverse(a2,y1_step1); Inverse normalization operation: (y - ymin) / gain + xoffset Map output from [-1, 1] back to original range.

[0098] In the process of completing the present application, the inventors compared the SVM prediction model. The optimal penalty coefficient c and kernel function width g of different epidemic stages of kiwifruit brown spot were screened out by cross-validation. The results showed that when the penalty coefficient c and kernel function width g of the SVM prediction model of the logistic growth period were 724.0773 and 0.0055 respectively, the simulation error of the RBF training function was the smallest, the mean square error (MSE) was 0.102381, and the correlation coefficient (R 2 ) was 0.820130. When the penalty coefficient c and kernel function width g of the recession period were 181.0193 and 0.00098 respectively, the mean square error (MSE) was the smallest, which was 0.166581, and the correlation coefficient (R 2 ) was 0.669975. The average prediction accuracy of the SVM prediction model for the occurrence degree of kiwifruit brown spot in the logistic growth period was 90.56%, and the mean square error (MSE) was 0.018273; the accuracy of the recession period was 81.85%, and the mean square error (MSE) was 0.108539.

[0099] Compared with the SVM prediction model, the training function of the BP neural network model can better simulate the occurrence degree of different epidemic stages, with lower mean square error (MSE) and stronger correlation (R 2 ). At the same time, the BP neural network model achieved better prediction results on the validation data, with higher accuracy and lower mean square error (MSE).

[0100] The data sample set of different epidemic stages of kiwifruit brown spot was trained with different number of hidden layer neurons. The results showed that when the number of hidden layer neurons of the neural network prediction model in the logistic growth and decline stages was 3 and 4 respectively, the simulation error of the trainlm training function was the smallest ( Figure 6 and Figure 7 ). Therefore, 3 and 4 hidden layer neurons were selected for training, respectively. The logistic growth stage model established by the training function was stopped after 5 iterations through the network performance and gradient change curve, which meant that the neural network reached the expected and achieved the best validation effect in the second training process ( Figure 8 ). The mean square error (MSE) tended to be stable at 0.006144, and the correlation coefficient (R 2 ) reached 0.98264 ( Figure 9 ). The decline stage model achieved the best validation effect in the third training process after 7 iterations ( Figure 10 ), with a mean square error (MSE) of 0.004581 and a correlation coefficient (R 2 ) of 0.99002 ( Figure 11 ). As shown in Table 3, the average prediction accuracy of the BP neural network prediction model for the occurrence degree of kiwifruit brown spot in the logistic growth stage was 94.59%, with a mean square error (MSE) of 0.001466; and the accuracy in the decline stage was 96.89%, with a mean square error (MSE) of 0.000787.

[0101] Table 3 Validation results of BP neural network prediction model Note: 14DJY is Dujiangyan in 2014, 19AZ, 20AZ and 21AZ are Anzhou in 2019, 2020 and 2021 respectively; 21LS is Lushan in 2021; DI is disease index.

[0102] 3.3. Model validation and evaluation Short-term prediction verificationSelect the latest 10 days of weather data as input to predict the disease index in the next 10 days. Compare the prediction results with the actual survey data, calculate the prediction accuracy, and ensure that the accuracy is ≥90.26%.

[0103] Medium-term prediction verificationUse the latest 30 days of weather data to predict the disease index in the next 30 days.

[0104] Compare the predicted value with the actual value to verify the accuracy of medium-term prediction, and ensure that the accuracy is ≥90.11%.

[0105]

[0106] 3.4. Screening of key meteorological factors Stepwise regression analysis: Using stepwise regression analysis method, the key meteorological factors with significant influence on disease index change were screened from a large number of meteorological factors.

[0107] Through the analysis of the correlation between meteorological factors and disease index, the logistic growth period is Y2=-302.726+14.930X 97 -10.677X 46 +0.747X 36 +0.053X 165 , and the recession period is Y3=-62.178+4.636X 97 +0.199X 114 -1.453X 148 +0.593X 188 , the combination of key meteorological factors is determined to improve the prediction ability of the model.

[0108] The meteorological factors of the logistic growth period are the extreme maximum temperature in mid-June (X 97 ), the extreme minimum temperature in early May (X 46 ), the extreme maximum rainfall in late April (X 36 ) and the rainfall in early August (X 165 ). The mid-term prediction of the prevalence of brown spot disease in the logistic growth period needs to obtain X 165 from the meteorological website, and the short-term prediction can be obtained by measurement. The meteorological factors of the recession period are the extreme maximum temperature in mid-June (X 97 ), the extreme maximum rainfall in late June (X 114 ), the minimum average temperature in late July (X 144 ) and the extreme maximum temperature in late August (X 188 ). The mid-term prediction needs to obtain X 144 and X 188 , the short-term prediction needs to obtain X 188 , and the rest are obtained by measurement.

[0109] The prediction results of kiwifruit brown spot disease occurrence by BP neural network model in 2023 (Table 4) show that the average accuracy of short-term prediction and mid-term prediction of the logistic growth period is 90.26% and 90.11% respectively; the average accuracy of short-term prediction of the recession period is 98.68%, and the mid-term prediction is slightly lower, which is 97.60%. The error of short-term prediction is slightly lower than that of mid-term prediction, and the mean square error (MSE) is 0.014426 and 0.014964 respectively, which has achieved good prediction effect. In summary, BP neural network can well predict the prevalence of kiwifruit brown spot disease, and can be considered for extensive testing and application in production.

[0110] Table 4: Prediction results of neural network model Note: 23CZ refers to Chongzhou in 2023, 23DJY refers to Dujiangyan in 2023, and 23PJ refers to Pujiang in 2023. DI refers to disease index.

[0111] 3.5. Visualization of epidemic prediction results Draw the prediction curve: based on the output results of the BP neural network model, draw the disease epidemic prediction curve in the future period.

[0112] The prediction curve includes the trend of disease index over time, which intuitively shows the epidemic situation of the disease.

[0113] Comparison with historical data: compare the prediction curve with historical disease data to evaluate the rationality of the prediction results.

[0114] Through comparison, the degree of agreement between the prediction results and the actual disease changes can be found, and the model parameters can be further optimized.

[0115] 3.6. Output and application of epidemic prediction results Prediction result output: display the predicted disease index, disease epidemic stage, and other information to users in the form of intuitive charts.

[0116] Provide short-term (within 10 days) and medium-term (within 30 days) prediction results to help users understand the disease development dynamics in a timely manner.

[0117] Adjustment of prevention and control measures: according to the prediction results, adjust the prevention and control strategy in advance, and reasonably arrange chemical control and agricultural control measures.

[0118] When it is predicted that the disease will enter the high epidemic stage, increase the frequency and intensity of control in a timely manner to reduce disease loss.

[0119] 4. Loss assessment module The loss assessment module is used to quantify the impact of kiwifruit brown spot disease on fruit quality and yield, and to convert disease loss into specific fruit quality comprehensive score and economic loss assessment based on disease index, providing scientific basis for formulating reasonable prevention and control strategies.

[0120] 4.1. Fruit quality index determination Index selection: determine the key indicators of fruit quality, including single fruit weight, fruit shape index, hardness, soluble solids content, dry matter content, disease fruit rate, and shelf life.

[0121] These indicators can comprehensively reflect the appearance quality, internal quality, and storage resistance of kiwifruit.

[0122] 4.2. Principal component analysis (PCA) Data Standardization: Use the membership function method to standardize the fruit quality indicators. For positively correlated indicators (single fruit weight, fruit shape index, hardness, soluble solids content, dry matter content, shelf life), calculate according to the formula, and for negatively correlated indicators (disease fruit rate), calculate according to the formula to eliminate the influence of different indicator dimensions.

[0123] Principal Component Extraction: Use SPSS 21.0 software for principal component analysis to extract principal components with eigenvalues greater than 1.0 and cumulative variance contribution rates greater than 85%.

[0124] Usually extract the first few principal components, which can explain most of the variation in the data.

[0125] Comprehensive Score Calculation: According to the characteristic vector and contribution rate of each principal component, calculate the principal component score of each sample.

[0126] Take the contribution rate as the weight, calculate the weighted sum of the first few principal components of each sample, and get the comprehensive score of fruit quality.

[0127] 4.3. Constructing Boltzmann Model Data Preparation: Collect fruit quality comprehensive score data under different disease index.

[0128] Ensure that the data covers a wide range of disease index to accurately reflect the relationship between disease index and fruit quality.

[0129] Model Construction: Use Origin 2019b software to construct the Boltzmann model with disease index as the independent variable and comprehensive score as the dependent variable.

[0130] The Boltzmann model can describe the nonlinear relationship between disease index and fruit quality, and the model form is: Where y is the comprehensive score, i.e. fruit quality, and x is the disease index.

[0131] The model is an inverted S-shaped curve. With the increase of disease index, the comprehensive score decreases slowly, and the impact on fruit quality is small. After reaching the inflection point, the comprehensive score rapidly decreases with the increase of disease index. When the disease develops to a certain extent, the impact on fruit quality tends to be stable, and the loss rate is below 5%, which is considered harmless.

[0132] 4.4. Economic Loss Warning Threshold Setting: According to the Boltzmann model, set the disease index threshold to 36.14. When the disease index exceeds this threshold, it is considered that the impact of kiwifruit brown spot disease on fruit quality is significant, and it may cause significant economic losses.

[0133] Early warning trigger: Integrate economic loss early warning function in the system, automatically trigger early warning when the monitored disease index approaches or exceeds the threshold.

[0134] In a preferred embodiment, the user is timely notified to take appropriate prevention and control measures to reduce potential economic losses through a visual interface or alarm information.

[0135] 5. Prevention and control decision module The prevention and control decision module is the core part of the kiwifruit brown spot disease prevention and control decision system, which integrates the results of the suitable area division, epidemic prediction and loss assessment modules to provide zonal management solutions and precise prevention and control measures for kiwifruit growers and agricultural managers.

[0136] 5.1. Data integration and analysis Multi-source data fusion: Integrate the spatial distribution map of different suitable grades of kiwifruit brown spot disease output by the suitable area division module, the disease epidemic stage prediction results provided by the epidemic prediction module, and the fruit quality comprehensive score and economic loss assessment data obtained by the loss assessment module.

[0137] Establish a unified data analysis platform to ensure that various data can be integrated and analyzed in the same spatial and temporal dimension.

[0138] 5.2. Zonal management plan development High-suitable area management strategy: Recommend planting disease-resistant varieties such as Ruiyu and Jinyan in high-suitable areas to reduce the risk of disease occurrence. Adjust the layout of varieties to improve the resistance of the host in the field.

[0139] Chemical control optimization: According to the short-term and medium-term prediction results of the epidemic prediction module, start the first application during the disease index growth period, and choose the latest chemical control before June 19 to effectively control the disease epidemic.

[0140] Agricultural measures strengthening: Spray 200g / L fluazinam·difenoconazole during the dormant period, with a 10-day interval and a total of 2 times, to reduce the overwintering base of the pathogen.

[0141] 5.3. Implementation of precise prevention and control measures Chemical control scheme refinement Pesticide selection and proportioning: Recommend using existing high-efficiency, low-toxicity, and environmentally friendly fungicides, and optimize the proportioning of pesticides according to different growth periods and disease epidemic stages.

[0142] Application timing and frequency: Combine the short-term prediction results of the epidemic prediction module to forecast the disease index 10 days in advance to ensure timely application before the disease occurs. The medium-term prediction results are used to plan the number of control times throughout the growing season, and preparations are made 30 days in advance.

[0143] 5.4. Dynamic adjustment of prevention and control decisions Real-time monitoring and feedback: Set up multiple fixed monitoring points in the field, regularly collect disease data and weather data, and feed back to the system in time.

[0144] According to the latest monitoring data, dynamically adjust the prevention and control strategy, such as advancing or delaying the spraying time, increasing or reducing the number of prevention and control, etc.

[0145] The above is only the preferred embodiment of the present application, it should be pointed out that the above preferred embodiment should not be regarded as limiting the present application, the protection scope of the present application should be limited by the scope defined by the claims. For ordinary skilled in the art, without departing from the spirit and scope of the present application, can also make a number of improvements and refinements, these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A kiwifruit brown spot disease prevention and control decision system, characterized in that, The system comprises: a data acquisition module for acquiring meteorological data, geographic information data and field pathogen monitoring data; a suitable area division module connected to the data acquisition module, which processes climate variables and species distribution data based on a MaxEnt model to output spatial distribution maps of high, medium, low and unsuitable areas for brown spot of kiwifruit; a prevalence prediction module which internally stores a BP neural network model and outputs disease prevalence stage prediction results after inputting key meteorological factors; a loss assessment module which maps disease index to fruit quality comprehensive score based on a Boltzmann model and presets a disease index threshold of 36.14; a control decision module which generates a zoned management plan based on the results of suitable area division, prevalence prediction and loss assessment.

2. The kiwifruit brown patch disease prevention and control decision system according to claim 1, characterized in that, The data acquisition module performs: monitoring the dynamics of pathogen spores in the air through a spore trap; detecting the pathogen using primers CC1F: 5'-CAAGTGTTGGGCGTCTGTCC-3' and CC1R: 5'-TGTGGGGGGGCTCATGGTG-3' in a qPCR detection system; collecting field disease fallen leaf samples every 15 days for overwintering pathogen monitoring.

3. The kiwifruit brown patch disease prevention and control decision system according to claim 2, characterized in that, The overwintering pathogen monitoring comprises: identifying fallen leaves on the ground as the initial source of infection; detecting the amount of fallen leaves carrying the pathogen from December to the following April; implementing fallen leaf removal or deep plowing of the soil to a depth of 10-20 cm after the fallen leaves.

4. The kiwifruit brown patch disease prevention and control decision system according to claim 1, characterized in that, The suitable area division module performs the following operations: dividing the suitable areas into high, medium, low and unsuitable areas; labeling the high suitable areas and high suitable core planting areas.

5. The kiwifruit brown patch disease prevention and control decision system according to claim 1, characterized in that, The BP neural network model of the prevalence prediction module comprises: the input layer receives four key meteorological factors: extreme maximum temperature in mid-June, extreme minimum temperature in early May, extreme maximum rainfall in late April, and rainfall in early August; the number of neurons in the hidden layer increases logically in the growth period and decreases in the decline period; the output layer predicts the disease index in the growth period and the decline period; the training parameters are Lenberg-Marquardt algorithm, 70% training set, 15% validation set and 15% test set.

6. The kiwifruit brown patch disease prevention and control decision system according to claim 5, characterized in that, The prevalence prediction module outputs: short-term prediction: predicting the disease index 10 days in advance with an accuracy rate of ≥90.26%; medium-term prediction: predicting the disease index 30 days in advance with an accuracy rate of ≥90.11%.

7. The kiwifruit brown patch disease prevention and control decision system according to claim 6, characterized in that, The training method of the BP neural network model comprises: Normalization processing weather data: ; wherein X norm represents the normalized value, X is the original value, X max and X min are the maximum and minimum values in the data set, respectively; selecting key meteorological factors through stepwise regression; adjusting the number of neurons in the hidden layer until the mean square error is less than 0.

01.

8. The kiwifruit brown patch disease prevention and control decision system according to claim 1, characterized in that, The loss assessment module performs: dimensionality reduction of seven quality indicators: single fruit weight, fruit shape index, hardness, soluble solids content, dry matter content, disease fruit rate and shelf life, through principal component analysis; standardizing the data using the membership function method; establishing a Boltzmann model of disease index and quality comprehensive score: ； where y is the comprehensive score, i.e. fruit quality, and x is the disease index; triggering economic loss warning when the disease index is >36.

14.

9. The kiwifruit brown patch disease prevention and control decision system according to claim 1, characterized in that, The control decision module generates a plan comprising: introducing disease-resistant varieties in high suitable areas; starting the first application of pesticides in the exponential growth period; performing chemical control no later than June 19 when the incidence rate reaches 5%. Emergency prevention and control is started when the disease index is greater than 36.

14.

10. The kiwifruit brown patch disease prevention and control decision system according to claim 1, characterized in that, The chemical control scheme comprises: Spraying 200 g / L fluopicolide·difenoconazole during the dormancy period; The interval between two times of application is 10 days, and the application is performed twice in total.

Images