Tea tree anthracnose disease prediction method based on sensitive period of active accumulated temperature extraction

Abstract

The application discloses a tea tree anthracnose prediction method based on an active accumulated temperature disease sensitive period extraction, which comprises the following steps: 1: constructing a plurality of different regions of plant protection and meteorology, remote sensing and geographical correlation data sets; 2: determining the active accumulated temperature threshold corresponding to the optimal sensitive period of tea tree anthracnose, and the time window length; 3: according to the optimal active accumulated temperature threshold and the time window length, the optimal sensitive period of tea tree anthracnose in different regions is determined; the habitat factors in the optimal sensitive period of tea tree anthracnose are extracted as the optional prediction factor set; 4: selecting the optimal prediction factor from the optional prediction factor set, and constructing a tea tree anthracnose prediction model. The application determines the sensitive period of tea tree anthracnose based on the active accumulated temperature, the sensitive periods of different regions are different, the spatial and temporal differences of the tea garden data in different regions can be extracted, and the disease prediction feature data set constructed based on this can significantly improve the accuracy of tea tree anthracnose prediction.

Description

Technical Field

[0001] This invention belongs to the field of biological disaster prediction technology, specifically involving a large-scale prediction method for anthracnose in tea trees based on the extraction of characteristics of sensitive periods of active accumulated temperature diseases. Background Technology

[0002] Anthracnose is a typical leaf disease of tea trees that seriously threatens the yield and quality of Chinese tea. Early and accurate control of the disease relies on large-scale disease prediction. The occurrence of anthracnose is influenced by a variety of habitat factors, resulting in complex relationships. However, current disease prediction models are mostly based on single meteorological data representing crop environmental conditions, and the representation of disease habitat information is still insufficiently systematic and comprehensive. The rapidly developing satellite remote sensing technology in recent years can provide ideal information reflecting host physiological conditions and geographical environmental differences. However, the response relationship between environmental and host factors and disease occurrence and prevalence is usually a dynamic process that changes over time, thus posing a significant challenge to the extraction and modeling of characteristic factors for crop disease prediction. Summary of the Invention

[0003] The purpose of this invention is to address the problem that the representation of disease habitat information in disease prediction is still not systematic and sufficient. It proposes a method for determining the sensitive period of disease based on active accumulated temperature, and applies the sensitive period window of disease to extract the spatiotemporally differentiated disease influencing factors represented by multi-source data such as meteorology, remote sensing, and geography, to construct a disease prediction feature dataset and establish a tea anthracnose prediction model based on multi-source data such as remote sensing.

[0004] A method for predicting anthracnose in tea trees based on sensitive periods of active accumulated temperature includes the following steps:

[0005] Step 1: Construct multiple disease and plant protection datasets linked with meteorology, remote sensing, and geography from different regions. These datasets include tea anthracnose data and three habitat factors: meteorological data, geographical data, and satellite remote sensing vegetation indices.

[0006] Step 2: Determine the active accumulated temperature threshold corresponding to the optimal sensitive period for anthracnose in tea trees, as well as the length of the time window.

[0007] 2.1 Set up alternative active accumulated temperature ranges; select multiple alternative active accumulated temperature thresholds within the active accumulated temperature ranges; calculate the active accumulated temperature for different dates for each region, and record the dates when each alternative active accumulated temperature threshold is reached as the starting point of the sensitive period.

[0008] 2.2 For each candidate accumulated temperature threshold, a time window length is set. Using the start point of the sensitive period and the time window length, the candidate disease sensitive period is obtained for each region for each candidate accumulated temperature threshold. The habitat factors within the candidate disease sensitive period corresponding to each accumulated temperature threshold in each region are taken as a sample; the samples corresponding to the same accumulated temperature threshold are formed into a sample set.

[0009] 2.3 Sensitivity analysis was performed on each sample set for the occurrence and severity of anthracnose in tea trees. The accumulated temperature threshold corresponding to the sample set with the highest sensitivity level was taken as the optimal accumulated temperature threshold, and the corresponding time window length was taken as the optimal time window length.

[0010] Step 3: Based on the optimal accumulated temperature threshold and time window length, determine the optimal sensitive period for anthracnose in tea trees in different regions; extract habitat factors within the optimal sensitive period for anthracnose in tea trees as a set of candidate predictive factors.

[0011] Step 4: Select some or all of the candidate predictors from the set of candidate predictors as the optimal predictors and construct a tea anthracnose prediction model.

[0012] Step 5: Based on the optimal accumulated temperature threshold, determine the starting point of the optimal sensitive period for the tested area; then, based on the optimal time window length and the starting point of the optimal sensitive period for the tested area, determine the ending point of the optimal sensitive period for the tested area. Extract the optimal predictive factors for the tested area within the optimal sensitive period and input them into the tea anthracnose prediction model obtained in Step 4 to obtain the prediction results of the occurrence of tea anthracnose in the tested area.

[0013] Preferably, the meteorological data includes temperature and precipitation.

[0014] Preferably, the geographic data includes elevation, slope aspect, and topography.

[0015] Preferably, the satellite remote sensing vegetation index uses the enhanced vegetation index and the photosynthetically active radiation component.

[0016] Preferably, the active accumulated temperature range set in step 2.1 is 50℃~300℃.

[0017] Preferably, the candidate active accumulated temperature thresholds in step 2.1 are obtained by enumerating multiple active accumulated temperature thresholds at 50°C intervals within the active accumulated temperature range.

[0018] Preferably, the formula for calculating the accumulated temperature in step 2.1 is:

[0019]

[0020] Where AAT represents the accumulated temperature of activity, Ti Let represent the average daily temperature on day i of the year, and s represent the number of days for calculating the accumulated temperature.

[0021] Preferably, in step 2.2, for each accumulated temperature threshold, the length of the period from the latest starting point of the sensitive period in the corresponding m regions to August 31st is taken as the time window length corresponding to that accumulated temperature threshold. m is the number of regions corresponding to the disease and plant protection, meteorology, remote sensing, and geographic association dataset constructed in step one.

[0022] Preferably, step 2.3 specifically performs sensitivity analysis on the ten-day average temperature corresponding to each sample set and the occurrence and severity of anthracnose in tea trees, and records the sensitivity level for each sample set. Sensitivity level is characterized by significant correlation over continuous time. Significant correlation indicates a p-value less than 0.05 in the sensitivity analysis.

[0023] As a preferred option, the specific process of step four is as follows: calculate the weights of all candidate predictors using the Relief-F algorithm and sort them from largest to smallest weight; select candidate predictors from largest to smallest according to different proportions to obtain multiple different combinations of predictors; use the RF algorithm to establish a tea anthracnose prediction model, take different combinations of predictors as inputs to the tea anthracnose prediction model, and use two accuracy indicators, OA and kappa coefficient, to evaluate the prediction model. Take the combination of predictors with the highest prediction accuracy as the optimal predictor to obtain the final tea anthracnose prediction model.

[0024] The beneficial effects of this invention are as follows:

[0025] 1. This invention determines the sensitive period of anthracnose in tea trees based on accumulated temperature. The sensitive period varies in different regions, enabling the extraction of spatiotemporal differences in tea garden data from different areas. The disease prediction feature dataset constructed based on this can significantly improve the accuracy of anthracnose prediction in tea trees.

[0026] 2. Based on the pathogenesis of anthracnose in tea trees, this invention selects meteorological data such as temperature and precipitation, geographical data such as altitude, slope aspect, and topography, and combines multi-source spatiotemporal data such as satellite remote sensing vegetation indices (EVI and FPAR) to characterize the environmental and host factors affecting the occurrence of anthracnose in tea trees. These data are then used as inputs to the model for disease prediction and modeling, thereby improving the accuracy of anthracnose prediction in tea trees. Attached Figure Description

[0027] Figure 1 This is a graph showing the results of the ten-day average temperature and the sensitivity of tea trees to anthracnose under different accumulated temperature (AAT) conditions in this invention.

[0028] Figure 2This is a schematic diagram illustrating how, in this invention, different regions determine and extract habitat factors within the optimal sensitive period for anthracnose in tea trees as a set of candidate predictive factors based on the optimal accumulated temperature threshold and time window length.

[0029] Figure 3 This is a comparison chart of the accuracy of prediction models based on different proportions of sorted predictor combinations using the Relief-F algorithm as input in this invention. Detailed Implementation

[0030] A method for predicting anthracnose in tea trees based on sensitive periods of active accumulated temperature includes the following steps:

[0031] Step 1: Construct multiple disease and plant protection datasets with meteorological, remote sensing, and geographic association data for different regions (specifically, multiple tea gardens within Zhejiang Province in this example). The number of regions is m.

[0032] Anthracnose in tea trees spreads rapidly under conditions of high humidity and high temperature, and its occurrence and development are closely related to the climatic background. Geographical environmental factors can, to some extent, characterize the differences in microclimate within tea gardens. Furthermore, the occurrence of anthracnose depends not only on environmental conditions but also on the host's physiological and growth status. When the host is growing vigorously, well-developed tea trees are much less likely to be infected with anthracnose than poorly developing tea trees.

[0033] The disease and plant protection dataset, which integrates meteorological, remote sensing, and geographic data, includes tea anthracnose data and three habitat factors: meteorological data, geographic data, and satellite remote sensing vegetation indices. Meteorological data includes temperature and precipitation. Geographic data includes elevation, aspect, and topography. Satellite remote sensing vegetation indices utilize EVI (Enhanced Vegetation Index) and FPAR (Photosynthetically Active Radiation Component). These meteorological, geographic, and satellite remote sensing vegetation indices collectively constitute multi-source spatiotemporal data, capable of characterizing environmental and host factors influencing the occurrence of tea anthracnose.

[0034] In this embodiment, anthracnose data for tea trees was provided by the Agricultural Technology Extension Center of Zhejiang Provincial Department of Agriculture and Rural Affairs, and was investigated and collected by tea plant protection professionals. Temperature and precipitation data were obtained from the daily grid value products of the National Meteorological Science Data Center. Altitude data were obtained from NASA's NASADEM data products. Slope aspect data was extracted and generated based on altitude data using ArcGIS 10.6 geographic information system software. Topographic data were obtained from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences. EVI and FPAR were obtained from the annual MODIS series MOD13Q1 and MOD15A2 data products.

[0035] Step 2: Determine the active accumulated temperature threshold corresponding to the optimal sensitive period for anthracnose in tea trees, as well as the length of the time window.

[0036] 2.1 Determine the start date of the disease-sensitive period by traversing different accumulated temperature windows.

[0037] The reproduction and invasion of anthracnose fungi in tea trees require certain accumulated temperature conditions. Environmental and host factors must meet these conditions to significantly influence the disease. Therefore, the period during which tea trees are susceptible to anthracnose when the accumulated temperature requirement is met is called the anthracnose sensitive period. Determining the anthracnose sensitive period is crucial for predicting the occurrence trend of anthracnose in tea trees. Based on the characteristics of accumulated temperature activity during the tea tree's growth period, within a relatively wide range of 50-300℃, the accumulated temperature thresholds required for the start of the disease sensitive period are enumerated at 50℃ intervals, resulting in n accumulated temperature thresholds; in this embodiment, n = 6.

[0038] Using January 1st as the starting time for calculating accumulated temperature, the accumulated temperature above biological zero (i.e., 10℃) in each region is calculated, and the date when the accumulated temperature threshold is reached is recorded. The corresponding date is used as the starting point of the sensitive period. For each accumulated temperature threshold, m regions are obtained as the starting points of the sensitive period, resulting in a total of m·n starting points of the sensitive period.

[0039] The formula for calculating the accumulated temperature is shown in equation (1).

[0040]

[0041] Where AAT represents the accumulated temperature of activity, T i Let represent the average daily temperature on day i of the year, and s represent the number of days for calculating the accumulated temperature.

[0042] 2.2 Determining the Sensitive Period for Diseases

[0043] Due to regional environmental differences, the starting dates of the disease-sensitive periods in different tea gardens during Step 1 vary under the same accumulated temperature threshold. To facilitate subsequent feature factor extraction analysis and predictive modeling, this embodiment standardizes the time window length for the sensitive period of tea anthracnose in different regions with the same accumulated temperature threshold. The peak incidence of tea anthracnose in the study area is in autumn, and tea farmers perform pesticide application and pruning to ensure the quality and yield of the next batch of tea. Considering the relationship between disease prediction and environmental factors, as well as the timeliness of tea anthracnose prediction in the study area, the end date of the disease-sensitive period should be no later than August 31st.

[0044] Therefore, for each accumulated temperature threshold, the period from the latest start date of the sensitive period in the corresponding m regions to August 31st is taken as the time window length corresponding to that accumulated temperature threshold. Thus, each of the n accumulated temperature thresholds obtains its corresponding time window length. For each of the m·n sensitive period start dates, the period within the time window length following the start date is considered as a candidate disease-sensitive period, resulting in m·n candidate disease-sensitive periods.

[0045] For each region, the habitat factors within the candidate disease-sensitive period corresponding to each accumulated temperature threshold are considered as one sample; m samples corresponding to the same accumulated temperature threshold form a sample set.

[0046] 2.3 Determine the required accumulated temperature threshold for the onset of the sensitive period for anthracnose in tea trees based on sensitivity analysis.

[0047] Temperature is the primary influencing factor for the occurrence of anthracnose in tea trees. To select relatively stable sensitive periods for anthracnose while controlling for data randomness, this invention performs sensitivity analysis on the ten-day average temperature corresponding to each sample set and the occurrence and severity of anthracnose, recording the sensitivity level for each sample set. Sensitivity levels are characterized by significant correlation over continuous time. Significant correlation indicates a p-value less than 0.05 in the sensitivity analysis. During the sensitivity analysis, samples within the same sample set are aligned on the time scale. The accumulated temperature threshold corresponding to the sample set with the highest sensitivity level is taken as the optimal accumulated temperature threshold, and the corresponding time window length is taken as the optimal time window length.

[0048] Step 3: Determine the optimal sensitive period for anthracnose in tea trees in different regions and extract alternative predictive factors for anthracnose in tea trees.

[0049] In hilly areas where tea gardens are distributed, temperature conditions vary significantly among different tea gardens due to topographical factors, and the time required to reach certain accumulated temperature conditions also differs. Therefore, extracting habitat factors for the same period across the entire region would ignore the differences in the disease-sensitive periods among different tea gardens within the region, affecting the effectiveness of disease prediction. Considering the spatiotemporal heterogeneity of disease-sensitive periods, to improve the relevance and effectiveness of multi-source habitat information in disease prediction, thereby enhancing the model's performance and robustness,

[0050] Based on the optimal accumulated temperature threshold, the starting point of the optimal sensitive period for different regions is determined; then, based on the starting point of the optimal sensitive period and the optimal time window length, the ending point of the optimal sensitive period for different regions is determined. Habitat factors within the optimal sensitive period for each region are extracted as a set of candidate predictive factors.

[0051] Step 4: Select predictive feature factors for anthracnose in tea trees and construct a predictive model for anthracnose in tea trees.

[0052] Based on the spatiotemporal extraction for disease-sensitive periods, to further reduce information redundancy among feature factors and address the computational complexity of disease prediction models, the Relief-F algorithm is used to calculate the weights of all candidate predictors and sort them from largest to smallest weight. Feature combinations with different sorting ratios (top 20%, 30%, 40%, 50%, 60%) are selected to form subsequent predictor combinations. The top 20% predictor combination represents the set of candidate predictors whose weights rank within the top 20% of the candidate predictor set.

[0053] Using the RF algorithm, a tea anthracnose prediction model was established. Different combinations of predictive factors ranked in different proportions (top 20%, 30%, 40%, 50%, 60%) were used as inputs to the tea anthracnose prediction model. The prediction model was evaluated using two accuracy indicators: OA and kappa coefficient. The combination of predictive factors with the highest prediction accuracy was selected as the optimal predictive factor data, and the final tea anthracnose prediction model was obtained.

[0054] Step 5: Based on the optimal accumulated temperature threshold and the accumulated temperature changes in the tested area, determine the starting point of the optimal sensitive period for the tested area; then, based on the optimal time window length and the starting point of the optimal sensitive period for the tested area, determine the ending point of the optimal sensitive period for the tested area. Extract the optimal predictive factor data for the tested area within the optimal sensitive period and input it into the tea anthracnose prediction model obtained in Step 4 to obtain the prediction results of tea anthracnose occurrence in the tested area.

[0055] In this embodiment, the number of regions m is 235; the optimal sensitive time period corresponding to one region for the six accumulated temperature thresholds and the sensitivity analysis results are as follows: Figure 1 As shown; from Figure 1 The data shows that the longest consecutive period (7 consecutive ten-day average temperatures) between the ten-day average temperature and tea anthracnose reached a significant level during the disease-sensitive period starting from the 150℃ accumulated temperature threshold indicates a stable sensitivity between temperature and disease occurrence within this temperature threshold standard. Therefore, the time point reaching the 150℃ accumulated temperature threshold is taken as the starting point of the optimal sensitive period for tea anthracnose.

[0056] Figure 2 The study shows that the optimal sensitive period for anthracnose in tea trees in different regions was determined based on the optimal accumulated temperature threshold and time window length; habitat factors within the optimal sensitive period for anthracnose in tea trees were extracted as a set of candidate predictive factors.

[0057] Figure 3The results show the accuracy of anthracnose predictions for tea plants obtained by inputting candidate predictor sets with different weightings (top 20%, 30%, 40%, 50%, 60%) into the prediction model based on the Relief-F algorithm. The prediction results indicate that the accuracy of the model increases as the number of input predictors decreases. When the input is a combination of predictors ranked in the top 20% by weighting, the model achieves the highest OA and kappa coefficients, with an OA accuracy of 0.85 and a kappa accuracy of 0.77.

Claims

1. A method for predicting anthracnose in tea trees based on sensitive periods of active accumulated temperature, characterized in that: Includes the following steps: Step 1: Construct multiple disease and plant protection datasets that are linked with meteorology, remote sensing, and geography in different regions; the disease and plant protection datasets linked with meteorology, remote sensing, and geography include anthracnose data for tea trees and three habitat factors; the three habitat factors are meteorological data, geographical data, and satellite remote sensing vegetation indices; Step 2: Determine the active accumulated temperature threshold corresponding to the optimal sensitive period for anthracnose in tea trees, as well as the length of the time window; 2.1 Set alternative active accumulated temperature ranges; Multiple candidate active temperature thresholds are selected from the active temperature range; the active temperature for different dates is calculated for each region, and the dates when each candidate active temperature threshold is reached are recorded as the starting point of the sensitive period. 2.2 For each candidate accumulated temperature threshold, a time window length is set; Using the start point of the sensitive period and the length of the time window, each region obtains a candidate disease sensitive period for each candidate accumulated temperature threshold; the habitat factors within the candidate disease sensitive period corresponding to each accumulated temperature threshold in each region are taken as a sample; the samples corresponding to the same accumulated temperature threshold are combined into a sample set. 2.3 Sensitivity analysis was performed on each sample set for the occurrence and severity of anthracnose in tea trees; the active accumulated temperature threshold corresponding to the sample set with the highest sensitivity level was taken as the optimal active accumulated temperature threshold, and the corresponding time window length was taken as the optimal time window length. Step 3: Based on the optimal accumulated temperature threshold and time window length, determine the optimal sensitive period for anthracnose in tea trees in different regions; extract habitat factors within the optimal sensitive period for anthracnose in tea trees as a set of candidate predictive factors; Step 4: Select some or all of the candidate predictors from the candidate predictor set as the optimal predictors and construct a tea anthracnose prediction model. Step 5: Determine the starting point of the optimal sensitive period in the measured area based on the optimal accumulated temperature threshold; then, determine the ending point of the optimal sensitive period in the measured area based on the optimal time window length and the starting point of the optimal sensitive period in the measured area. Extract the optimal predictive factors for the tested area during the optimal sensitive period and input them into the tea anthracnose prediction model obtained in step four to obtain the prediction results of the occurrence of tea anthracnose in the tested area.

2. The method for predicting anthracnose in tea trees based on the sensitive period of active accumulated temperature disease extraction according to claim 1, characterized in that: The meteorological data mentioned includes temperature and precipitation.

3. The method for predicting anthracnose in tea trees based on the sensitive period of active accumulated temperature disease extraction according to claim 1, characterized in that: The geographic data includes elevation, slope aspect, and topography.

4. The method for predicting anthracnose in tea trees based on the sensitive period of active accumulated temperature disease extraction according to claim 1, characterized in that: The satellite remote sensing vegetation index mentioned uses the enhanced vegetation index and the photosynthetically active radiation component.

5. The method for predicting anthracnose in tea trees based on the sensitive period of active accumulated temperature disease extraction according to claim 1, characterized in that: The active accumulated temperature range set in step 2.1 is 50℃~300℃.

6. The method for predicting anthracnose in tea trees based on the sensitive period of active accumulated temperature disease extraction according to claim 1, characterized in that: The candidate active accumulated temperature thresholds in step 2.1 are obtained by enumerating multiple active accumulated temperature thresholds at 50°C intervals within the active accumulated temperature range.

7. The method for predicting anthracnose in tea trees based on the sensitive period of active accumulated temperature disease extraction according to claim 1, characterized in that: The formula for calculating the accumulated temperature in step 2.1 is: if T i ≥10℃,T' i =T i ,elseT' i =0 Where AAT represents the accumulated temperature of activity, T i Let represent the average daily temperature on day i of the year, and s represent the number of days for calculating the accumulated temperature.

8. The method for predicting anthracnose in tea trees based on the sensitive period of active accumulated temperature disease extraction according to claim 1, characterized in that: In step 2.2, for each accumulated temperature threshold, the length of the period from the latest time point of the sensitive period start point in the corresponding m regions to August 31 is taken as the time window length corresponding to that accumulated temperature threshold. m represents the number of regions corresponding to the disease and plant protection, meteorology, remote sensing, and geographic association dataset constructed in step one.

9. The method for predicting anthracnose in tea trees based on the sensitive period of active accumulated temperature disease extraction according to claim 1, characterized in that: In step 2.3, sensitivity analysis was specifically performed on the ten-day average temperature corresponding to each sample set and the occurrence and severity of anthracnose in tea trees, and the sensitivity level corresponding to each sample set was recorded. The sensitivity level was characterized by significant correlation over continuous time. Significant correlation indicates that the P value in the sensitivity analysis is less than 0.

05.

10. The method for predicting anthracnose in tea trees based on the sensitive period of active accumulated temperature disease extraction according to claim 1, characterized in that: The specific process of step four is as follows: calculate the weights of all candidate predictors using the Relief-F algorithm and sort them from largest to smallest weights; select candidate predictors from largest to smallest according to different proportions to obtain multiple different combinations of predictors; Using the RF algorithm, a tea anthracnose prediction model was established. Different combinations of predictive factors were used as inputs to the tea anthracnose prediction model. The prediction model was evaluated using two accuracy indicators: OA and kappa coefficient. The combination of predictive factors with the highest prediction accuracy was selected as the optimal predictive factors, thus obtaining the final tea anthracnose prediction model.

Images