A prediction method for the peak incidence of rice sheath blight based on meteorological and landscape information

Abstract

The present invention discloses a method for predicting the peak incidence of rice sheath blight based on meteorological and landscape information. This method collects the actual peak incidence data of past diseases in the area to be predicted, as well as the meteorological information and remote sensing landscape information in the corresponding period, and extracts the key meteorological factors and key landscape factors. Then, using the correlation analysis method, the extracted key factors are selected, and the factors that have a significant correlation with the peak incidence of rice sheath blight are used to establish a prediction model. The mathematical relationship between the historical peak incidence of rice sheath blight and meteorological and landscape factors is obtained by using the partial least squares regression method, and based on this, the peak incidence of rice sheath blight in the area to be predicted in the next year is predicted. This method can predict the future peak incidence at the stage when rice sheath blight has not occurred or has just started to occur, and can provide guidance for subsequent behaviors such as the prevention and control intensity of paddy fields.

Description

Technical Field

[0001] This method belongs to the technical field of vegetation disease prediction, and relates to a prediction method combining disease pathological characteristics, meteorological information and remote sensing landscape information, and specifically relates to a rice sheath blight incidence peak prediction method based on meteorological and landscape information. Background Art

[0002] With the update and iteration of mathematical modeling, meteorological sensors and satellite remote sensing technology, rice disease prediction has ushered in unprecedented opportunities. In the existing technology, most pest and disease predictions, including rice sheath blight, mainly rely on expert experience or are based on single conditions such as meteorological factors and remote sensing factors. The disadvantages of these prediction methods are that they are highly subjective, or the prediction model factors are single and do not fully consider the impact of human activities and changes in landscape patterns on the prevalence of diseases. Human adjustments to land use and crop planting patterns, including changes in farmland landscapes caused by reclamation, division or transformation of farmland landscapes, can extract surface cover types through remote sensing data, observe changes in host crop distribution, and provide spatial information for landscape pattern analysis for disease prevalence. Therefore, combining meteorological factors and remote sensing factors has great potential in the field of disease prediction. Summary of the invention

[0003] In view of the shortcomings of the existing technology, the present invention proposes a method for predicting the peak incidence of rice sheath blight based on meteorological and landscape information. By comprehensively using meteorological factors and remote sensing landscape information, the peak incidence of the disease is predicted in the pre-disease period or in the early stage of the disease, providing an important basis for the prevention and control efforts and prevention and control work of the disease in the later stage.

[0004] A method for predicting the peak incidence of rice sheath blight based on meteorological and landscape information specifically comprises the following steps:

[0005] Step 1: Collect the actual peak data of past disease occurrences, as well as the corresponding meteorological data and remote sensing data of land cover types, to construct a data set.

[0006] s1.1. Collect the peak occurrence rate P of rice sheath blight diseased plants in each sampled rice field during the growth period in the tested rice area in the past few years.

[0007] s1.2. Collect the daily average temperature and daily average precipitation of each sampled rice field in the past few years, and construct a plant protection-meteorological correlation data set based on the peak value P obtained in s1.1.

[0008] s1.3. Collect the transplanting period phenological data and land cover type data of the locations of the tested rice areas in the past few years, and establish the plant protection-phenological data set and the plant protection-landscape data set based on the peak value P obtained in s1.1.

[0009] Step 2: Based on the collected dataset of the association between plant protection and meteorology in Step 1, extract the active accumulated temperature, precipitation, and number of rainy days during different growth stages of rice as key meteorological factors.

[0010] Step 3: Based on the collected dataset of plant protection and landscape in Step 1, extract the landscape indices of the location of the measured rice-growing area, including patch density (PD), largest patch index (LPI), edge density (ED), landscape shape index (LSI), aggregation index (C), mean patch size (MPS), patch size standard deviation (PSSD), patch size coefficient of variation (PSCV), patch richness density (PRD), diversity index (H), class proportion (PLAND), mean patch fractal dimension (FRAC), aggregation index (AI), and clumpiness index (CLUMPY) as key landscape factors.

[0011] Step 4: Use the correlation analysis method to select the key meteorological factors and key landscape factors extracted in Steps 2 and 3.

[0012] s4.1: Conduct a correlation analysis between the peak occurrence of sheath blight in rice and the key meteorological factors during the growth stage, and extract the significantly correlated meteorological factors.

[0013] s4.2: Conduct a correlation analysis between the peak occurrence of sheath blight in rice and the key landscape factors of the location of the measured rice-growing area, and extract the significantly correlated landscape factors.

[0014] Step 5: Based on the significantly correlated meteorological factors and landscape factors extracted in Step 4, and the historical disease incidence of the measured rice-growing area, use the partial least squares regression method to establish the relationship between the peak occurrence of historical sheath blight in rice and meteorological and landscape factors, and obtain a prediction model for the peak occurrence of sheath blight in rice.

[0015] Step 6: Calculate the key meteorological factors and key landscape factors of the area to be predicted, and input them into the prediction model obtained in Step 5 to obtain the predicted value of the peak occurrence of sheath blight in rice in this area.

[0016] The present invention has the following beneficial effects:

[0017] This method can predict the peak occurrence of sheath blight in rice in the predicted area based on the meteorological conditions and landscape feature information during the past rice growth stages. It can predict the future peak occurrence of the disease when sheath blight has not occurred or is in the initial stage of occurrence, and can provide guidance for subsequent prevention and control measures in paddy fields. Description of the Drawings

[0018] Figure 1 It is a scatter plot comparison of the actual peak occurrence of sheath blight in rice in the measured rice-growing area in the embodiment and the prediction results.

[0019] Figure 2This is a comparison chart of the actual peak occurrence of rice sheath blight in the tested rice area in 2013 and the predicted result. DETAILED DESCRIPTION

[0020] The present invention is further explained below in conjunction with the accompanying drawings.

[0021] A method for predicting the peak incidence of rice sheath blight based on meteorological and landscape information specifically comprises the following steps:

[0022] Step 1: Collect the actual peak data of past disease occurrences, as well as the corresponding meteorological data and remote sensing data of land cover types, to construct a data set.

[0023] s1.1. Select n sampling paddies in the tested rice area, and sample the rice sheath blight disease rate in the n sampling paddies during the growth period of more than 5 years from June 30 to September 30 every year with a 5-day cycle, and record the measurement time, geographical information of the observation point, rice crop type and disease rate. Then, the peak incidence P of rice sheath blight during the growth period is counted based on the collected data.

[0024] s1.2. Based on the geographical location information of n sampled rice fields, the daily average temperature and precipitation grid daily products with a spatial resolution of 0.5°×0.5° from the National Meteorological Science Data Center were collected to collect the daily average temperature and precipitation of each sampled rice field over the past five years. Based on the peak value P obtained in s1.1, a plant protection-meteorological correlation dataset was constructed.

[0025] s1.3. Taking the county boundary where the rice-growing area is located as the boundary, using the MCD12Q1 data product of the MODIS series and the China CropPhen1km, a high-resolution rice phenology dataset produced based on the leaf area index product of the Global Land Surface Characteristics Parameters (GLASS), the phenology and land cover type data of the counties and cities where the rice-growing areas are located in the past five years or more are extracted. According to the peak value P obtained in s1.1, the plant protection-phenology dataset and the plant protection-landscape dataset are established.

[0026] Step 2: Meteorological factors have different degrees of influence on the occurrence and development of diseases at different growth stages of rice. Therefore, it is necessary to determine the critical growth stage time when meteorological factors act. After rice seedlings are transplanted, it is mainly divided into the seedling stage, tillering stage, jointing stage, booting stage, heading stage, etc. The sheath blight of rice usually shows disease symptoms at the tillering stage and reaches the maximum value of disease occurrence from the booting stage to the heading stage. Therefore, in this embodiment, the critical growth stages between rice transplanting and the booting stage are selected, including the early tillering stage of rice (0 - 15 days), the late tillering stage (15 - 30 days), the tillering stage (0 - 30 days), the jointing stage (30 - 45 days), and the late tillering stage to the jointing stage (15 - 45 days), to extract the key meteorological factors. The key meteorological factors include active accumulated temperature, rainfall, and rainy days.

[0027] Active Accumulated Temperature A a (Active Accumulated Temperature) refers to the sum of the daily average temperatures exceeding the biological lower limit temperature of the crop within a specified time period, which has biological significance in plant mechanism and is used to describe the development of the crop or crop disease during the growth period:

[0028]

[0029] s.t.TEM d ∈[TEM low ,TEM high

[0030] Where, D start is the start date of the growth period, D end is the end date of the growth period, TEM d is the average temperature on the d-th day, TEM low is the lower limit temperature of disease incidence of the disease, TEM high is the upper limit temperature of disease incidence of the disease.

[0031] Rainy days RD and rainfall PRE are two key meteorological factors highly related to moisture:

[0032]

[0033] s.t.d>0

[0034] Where, pre d is the precipitation on the d-th day, RD d is the number of rainy days of precipitation within d days.

[0035] ​Step 3: According to the collected plant protection-landscape dataset in Step 1, use the landscape spatial pattern quantitative analysis software FRAGSTATS to conduct spatial landscape pattern analysis at the patch type level and landscape level by calculating landscape indices, so as to obtain the landscape indices of the measured rice-growing area in the sampling year as the key landscape factors. Specifically as follows:

[0036] (1) Patch density PD

[0037]

[0038] In the formula, N is the number of patch types in the landscape, and A is the total patch area. Its value range is PD > 0, without an upper limit. As an important index to describe landscape fragmentation, the larger the PD, the higher the degree of landscape fragmentation.

[0039] (2) Largest patch index LPI

[0040]

[0041] Among them, a k is the area of the k-th patch in the landscape, and A is the total landscape area. Its value range is 0 < LPI ≤ 100. LPI characterizes the proportion of the largest patch in the entire landscape or a certain type.

[0042] (3) Edge density ED

[0043]

[0044] Among them, E is the total length of patch boundaries in the landscape, with the unit of meter. ED ≥ 0, without an upper limit.

[0045] (4) Landscape shape index LSI

[0046]

[0047] When the patch shape is irregular or deviates from a square, the landscape shape index LSI will increase.

[0048] (5) Aggregation index C

[0049]

[0050] In the formula, C max is the maximum value of the aggregation index 2ln(n). n is the total number of patch types in the landscape, and P ij is the probability of adjacency between patch types i and j. In the rasterized digital landscape pattern, P ij is calculated as follows:

[0051] P ij = P i Pj / i (9)

[0052] Among them, P i is the probability that the randomly selected grid pixel value belongs to patch type i, and this probability can also be estimated using the proportion of the area of patch type i in the entire landscape. And P j / i is the conditional probability that patch type j is adjacent to i when the patch type is i, and the calculation method of P j / i is as follows:

[0053] P j / i = m ij / m i (10)

[0054] Among them, m ij is the number of grid edges adjacent to patches i and j in the landscape pattern, and m i is the total number of edges of the grid of patch type i.

[0055] The aggregation index C provides a quantitative feedback on the degree of aggregation between different types of landscapes. If there are many discrete patches in the landscape, the value of the aggregation index C is small; otherwise, it is large. The aggregation index reflects the adjacent relationship between patch types and the spatial configuration characteristics of the landscape.

[0056] (6) Patch area

[0057] 1) Mean patch size MPS

[0058]

[0059] 2) Patch size standard deviation PSSD

[0060]

[0061] In the formula, a k is the area of the k-th patch, PSSD ≥ 0, with no upper limit. When the sizes of all patches are the same, or there is only one patch, PSSD = 0.

[0062] 3) Patch size coefficient of variation PSCV

[0063]

[0064] PSCV ≥ 0, with no upper limit.

[0065] (7) Patch richness density PRD

[0066]

[0067] In the formula, PRD > 0, with no upper limit.

[0068] (8) Diversity index H

[0069] The diversity index H is used to measure the complexity of the system structure composition. Common diversity indices include the Shannon diversity index and the Simpson diversity index.

[0070] 1) Shannon diversity index

[0071]

[0072] In the formula, P k is the probability of the occurrence of patches of type k. Generally, the proportion of the number of grids or pixels occupied by type k in the total number of grids or pixels of the overall landscape is used for estimation.

[0073] 2) Simpson diversity index

[0074]

[0075] The diversity index is affected by information in two aspects: one is the richness of patch types in the landscape pattern; the other is the evenness of the area distribution of different patch types. For a given n, when the area ratios of each patch are the same, that is, P k = 1 / n, the diversity index will reach the maximum value. Generally speaking, when H increases, to a certain extent, it can indicate that the complexity of the current landscape structure composition is also increasing.

[0076] (9) Class proportion PLAND

[0077]

[0078] In the formula, r is the number of patches of the target type, and a t is the area of the t-th patch of this type. 0 < PLAND ≤ 100. The fewer the patches of type i in the landscape, the closer PLAND is to 0; when the entire landscape consists of a single patch type, PLAND is 100. The class proportion PLAND quantifies the proportional abundance of each patch type in the landscape and is an important landscape composition index. Moreover, it is a relative measure and is suitable as a landscape index for comparing landscape patterns of different sizes.

[0079] (10) Average patch fractal dimension FRAC

[0080]

[0081] In the formula, D it is the perimeter of the t-th patch of patch type i. Generally speaking, 1 < FRAC < 2, which can measure the complexity of the shape of the landscape or patch type. The more complex the shape, the closer the fractal dimension is to 2; on the contrary, the simpler the shape, the closer it is to 1.

[0082] (11) Aggregation Index AI

[0083]

[0084] Wherein, G is the number of adjacent grids under a certain type of patch, G max is the maximum number of adjacent grids that all grids under this type can reach. The value range of AI is 0 ≤ AI ≤ 100. Each adjacent grid edge in AI is counted at most once, reflecting the degree of aggregation inside the patch and not being affected by the connection between the patch boundary and other patches.

[0085] (12) Clumpiness Index CLUMPY

[0086]

[0087] Then it is the number of grids of all patches under a certain type. CLUMPY can reflect the frequency of different types of patches being juxtaposed and aggregated in the landscape. Its value range is -1 ≤ CLUMPY ≤ 1. The more dispersed the patches are, the closer CLUMPY tends to -1; conversely, the more aggregated the patches are, the closer CLUMPY tends to 1.

[0088] Step 4: Take the key meteorological factors and key landscape factors extracted in Steps 2 and 3 as independent variables, and the peak value of rice sheath blight as the dependent variable, calculate the Pearson correlation coefficient R between the two, and analyze the correlation between different factors and the peak value of rice sheath blight. The calculation method of the Pearson correlation coefficient is as follows:

[0089]

[0090] Wherein, b is the number of independent variables, X a is the a-th independent variable, Y is the dependent variable, Cov(X a , Y) is the covariance between variables X a and Y, Var[X a is the variance of X a , and Var[Y] is the variance of Y. The interval of R is [-1, 1]. When R ≠ 0, it indicates that there is a correlation between variables. When R > 0, it indicates a negative correlation between the two, and when R < 0, it indicates a positive correlation between the two. The closer |R| is to 1, the stronger the correlation. If R = 0, there is no correlation between the two.

[0091] s4.1. First, conduct a correlation analysis on the peak value of rice sheath blight and the key meteorological factors during the growth period. The results are shown in Table 1:

[0092]

[0093]

[0094] Table 1

[0095] Among them, *p<0.05; **p<0.01; ***p<0.001. According to Table 1, the precipitation in the early tillering period was significantly positively correlated with the peak incidence of sheath blight, R=0.457, p<0.001, so this key meteorological factor was selected as a significantly correlated meteorological factor.

[0096] s4.2. Then, the correlation analysis was conducted between the peak occurrence of rice sheath blight and the key landscape factors of the location of the tested rice area. The results are shown in Table 2:

[0097]

[0098]

[0099] Table 2

[0100] According to Table 2, at the landscape level, the peak value and the landscape shape index LSI and the plate richness density PRD index have a significant negative correlation and a significant positive correlation with the peak value of rice sheath blight respectively; at the patch type level, there are 4 landscape indices that have a significant correlation with the peak value of rice sheath blight, namely, the maximum patch index LPI, the shape index LSI, the average patch area MPS, and the patch area standard deviation PSSD. Therefore, the above 6 key landscape factors are selected as significantly correlated landscape factors.

[0101] Step 5: In order to reflect the characteristics of the local fungus source, this embodiment also extracts the peak value of the historical plant protection data, calculates the average value of the peak values ​​of the tested rice area over the years, and uses the partial least squares regression method together with the significantly correlated meteorological factors and landscape factors extracted in step 4 to establish the relationship between the historical rice sheath blight occurrence peak value and the meteorological and landscape factors, and obtains the rice sheath blight incidence peak prediction model. The model coefficients are shown in Table 3:

[0102]

[0103] Table 3

[0104] Step 6: Calculate the key meteorological factors and key landscape factors of the area to be predicted, input them into the prediction model obtained in step 5, and obtain the predicted value of the peak incidence of rice sheath blight in the area. Figure 1 As shown in the figure, it can be seen that the actual peak value of rice sheath blight disease rate in the measured area and the predicted peak value show a strong linear relationship, and the prediction accuracy rate reaches 85%. Taking 2013 as an example, Figure 2As shown, the prediction effect of the model is good. It can make a better prediction of the peak incidence of rice fields in the measured area in the current year based on information such as meteorological conditions in the early growth stage of rice and local landscape pattern characteristics, indicating that this method is practical and effective.

Claims

1. A method for predicting the peak incidence of sheath blight in rice based on meteorological and landscape information, characterized in that: Specifically, it includes the following steps: Step 1: Collect the actual peak data of past disease occurrences, as well as the corresponding meteorological data and remote sensing data of land cover types, and construct a plant protection - meteorology correlation dataset, a plant protection - phenology dataset, and a plant protection - landscape dataset; Step 2: According to the collected dataset of the correlation between plant protection and meteorology in Step 1, extract the active accumulated temperature A, precipitation, and number of rainy days during different growth stages of rice as key meteorological factors; the different growth stages of rice include the early tillering stage, late tillering stage, tillering stage, jointing stage, and the stage from late tillering to jointing stage of rice. a , precipitation, and number of rainy days during different growth stages of rice as key meteorological factors; the different growth stages of rice include the early tillering stage, late tillering stage, tillering stage, jointing stage, and the stage from late tillering to jointing stage of rice. The active accumulated temperature A a is calculated as follows: s.t. TEM d ∈ [TEM low , TEM high ​ Among them, D start is the starting date of the growth period, D end is the ending date of the growth period, TEM d is the average temperature on the d-th day, TEM low is the lower limit of the disease incidence temperature, TEM high is the upper limit of the disease incidence temperature; The calculation methods for the number of rainy days RD and rainfall PRE are: Among them, pre d is the precipitation on the d-th day, and RD d is the number of rainy days within d days; Step 3: According to the plant protection - landscape dataset collected in Step 1, extract the landscape indices of the location where the rice area to be measured is located, including patch density PD, largest patch index LPI, edge density ED, landscape shape index LSI, aggregation index C, mean patch size MPS, patch size standard deviation PSSD, patch size coefficient of variation PSCV, patch richness density PRD, diversity index H, class proportion PLAND, mean patch fractal dimension FRAC, aggregation index AI, and clumpiness index CLUMPY, as key landscape factors. The calculation methods are: (1) Patch density PD In the formula, N is the number of type patches in the landscape, and A is the total patch area; (2) Largest patch index LPI where a k is the area of the k-th patch in the landscape, and A is the total area of the landscape; (3) Edge density ED Among them, E is the total length of the patch boundaries in the landscape, with the unit of meter; (4) Landscape shape index LSI (5) Aggregation index C where C max is the maximum value of the aggregation index, 2ln(n); n is the total number of patch types in the landscape, and P ij is the probability of adjacency between patch types i and j; P ij = P i P j / i (9) Among them, P i is the probability that the randomly selected grid pixel value belongs to patch type i, and P j / i represents the conditional probability that patch type j is adjacent to i when the patch type is i: P j / i = m ij / m i (10) where m ij is the number of grid edges adjacent to patches i and j in the landscape pattern, and m i is the total number of edges of the grid of patch type i; (6) Patch area 1) Mean patch size MPS MPS > 0; 2) Patch size standard deviation PSSD 3) Patch size coefficient of variation PSCV (7) Patch richness density PRD (8) Diversity index H The diversity index H is used to measure the complexity of the system structure composition, including the Shannon diversity index and the Simpson diversity index; 1) Shannon diversity index where P k is the probability of the occurrence of k-type patches; 2) Simpson diversity index (9) Class proportion PLAND where r is the number of patches of the target type, and a t is the area of the t-th patch of this type; (10) Mean patch fractal dimension FRAC where D it is the perimeter of the t-th patch of patch type i; (11) Aggregation index AI where G is the number of adjacent grids under a certain type of patch, and G max is the maximum number of adjacent grids that all grids under this type can reach; (12) Clumpiness index CLUMPY where \(N\) is the raster count of all patches of a certain type, and \(P\) is the proportion of patches of this type in the overall landscape; Step 4: Calculate the correlations between the key meteorological factors, key landscape factors extracted in Steps 2 and 3, and the peak incidence of sheath blight in rice, and select the factors with significant correlations among them; Step 5: According to the meteorological factors and landscape factors with significant correlations selected in Step 4, and the historical disease incidence of the rice area to be measured, use the partial least squares regression method to establish the relationship between the historical peak incidence of sheath blight in rice and meteorological and landscape factors, and obtain a prediction model for the peak incidence of sheath blight in rice; Step 6: Calculate the key meteorological factors and key landscape factors of the area to be predicted, and input them into the prediction model obtained in Step 5 to obtain the predicted value of the peak incidence of sheath blight in rice in this area.

2. The method for predicting the peak incidence of sheath blight in rice based on meteorological and landscape information as described in claim 1, characterized in that: The specific content of Step 1 includes: s1.1: Collect the peak incidence P of the sheath blight disease plant rate of each sampling paddy field during the growth period in the rice area to be measured for more than 5 years in the past; s1.2: Collect the daily average temperature and daily average precipitation of each sampling paddy field for more than 5 years in the past, and construct a plant protection - meteorology correlation dataset according to the peak P obtained in s1.1; S1.

3. Collect the phenological data of the transplanting period and the land cover type data at the location of the measured rice-growing area over the past 5 years or more, and establish a plant protection - phenology dataset and a plant protection - landscape dataset according to the peak value P obtained in S1.

1.

3. A method for predicting the incidence peak of rice sheath blight based on meteorological and landscape information as claimed in claim 1, characterized in that: in step four, the Pearson correlation coefficient is used to measure the correlation between the key meteorological factors or key landscape factors and the incidence peak of rice sheath blight.

Images