A method for predicting oil-tea damage
Through comprehensive analysis and XGBoost algorithm, the oil tea disaster loss prediction model was constructed, and the problem of oil tea disaster loss prediction was solved, more accurate disaster loss assessment and climate risk assessment were achieved, and agricultural meteorological disaster prevention and mitigation capabilities were improved.
Patent Information
- Application Number
- CN202111297931.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-04
- Publication Date
- 2024-12-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing technology is difficult to effectively predict and evaluate the damage to oil tea, which affects oil tea production and quality, and lacks accurate meteorological disaster warnings and disaster reduction measures.
Using four methods: correlation analysis, analysis of variance, XGBoost factor importance analysis and nuclear density fitting, the key climate risk factors of oil tea production were screened out, and the disaster loss prediction model for each major growth period of oil tea was constructed through the XGBoost algorithm to establish oil tea. Total disaster loss prediction model to achieve disaster loss warning and prevention.
It improves the accuracy and accuracy of oil tea disaster damage prediction, can effectively evaluate the probability and rate of disaster damage, and provides better climate risk assessment and agricultural meteorological disaster prevention and mitigation services.
Smart Images

Figure CN113947031B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of agriculture, and in particular to a method for predicting oil-tea disaster losses. Background Art
[0002] Camellia oleifera is a high-grade woody edible oil plant unique to my country. At present, my country's camellia oleifera is mainly distributed in the Yangtze River Basin and 14 provinces to its south. Among them, Hunan Province ranks first in the country in terms of camellia oleifera planting area, tea oil production and output value. Like all crops produced in the open air, meteorological conditions are an important factor affecting the yield, quality and cost of camellia oleifera. The entire growth period of camellia oleifera from spring shoot growth to fruit ripening lasts for nearly 2 years. The vegetative growth and reproductive growth of camellia oleifera influence and intertwine with each other, and the meteorological conditions in each phenological period will affect the growth and development of camellia oleifera, and then affect the yield of camellia oleifera. Therefore, studying the impact of various meteorological disasters on the yield of Hunan camellia oleifera and providing a quantitative disaster loss assessment model are of great importance to carrying out agricultural meteorological disaster prevention and mitigation and improving the level of meteorological services for agriculture. Summary of the invention
[0003] In order to solve the above technical problems, the present invention discloses a method for predicting oil-tea disaster losses. The present invention obtains key climate risk factors for oil-tea production by comprehensively using four methods: correlation analysis, variance analysis, XGBoost factor importance analysis, and kernel density fitting, and establishes a related disaster loss prediction model, which is convenient for disaster loss warning, prevention, and remediation.
[0004] In order to achieve the above object, the technical solution of the present invention is as follows:
[0005] A method for predicting oil-tea damage, comprising the following steps:
[0006] Step 1: Collect the climate influencing factors of oil-tea during each major yield-influencing period and the corresponding relative meteorological yield of oil-tea;
[0007] Step 2: Analyze and obtain the key climate influencing factors of each major yield influencing period and combine them with the corresponding oil-tea relative meteorological yield to form a dataset of the major yield influencing period;
[0008] Step 3: Establish a disaster loss prediction model for each major production influencing period based on the relative meteorological yield of oil-tea and the key climate influencing factors of each major production influencing period, and then obtain the final disaster loss prediction model for each major production influencing period through training of the main production influencing period data set;
[0009] Step 4: Select the climate influencing factors of the whole growth period of oil-tea from the key climate influencing factors of each major yield influencing period and combine them with the corresponding relative meteorological yield of oil-tea to form a prediction data set; establish a total disaster loss prediction model for oil-tea based on the selected climate influencing factors and the corresponding relative meteorological yield of oil-tea, and then obtain the final total disaster loss prediction model for oil-tea through training of the prediction data set;
[0010] Step 5. When the tea oil tree is actually growing, the key climate influencing factors of the corresponding growth period of the tea oil tree are input into the disaster loss prediction model of the corresponding yield influence period to predict the probability of disaster losses in the tea oil tree during the yield influence period, thereby evaluating the disaster loss situation; when the tea oil tree fruits are ripe, the climate influencing factors of the entire growth period of the tea oil tree are input into the total disaster loss prediction model of the tea oil tree to predict the probability and loss rate of disaster losses in that year, thereby evaluating the disaster loss situation.
[0011] As a further improvement, the disaster loss prediction model for each major yield-influencing period and the total disaster loss prediction model for oil-tea are both established using the XGBoost algorithm.
[0012] According to a further improvement, the main yield-influencing periods include spring shoot sprouting period, flowering period, young fruit period, fruit enlargement peak period, oil accumulation period and fruit ripening period.
[0013] Further improvement, the key climate influencing factors of the spring shoot budding period include daily average temperature <5°C accumulated temperature, average maximum temperature, average minimum temperature, and accumulated precipitation;
[0014] The key climate factors affecting the flowering period include daily average temperature <10℃ accumulated temperature, daily average temperature ≥10℃ active accumulated temperature, average minimum temperature, and few days of sunshine;
[0015] The key climate factors affecting the young fruit stage include daily average temperature ≥5℃ active accumulated temperature, average temperature, average maximum temperature and extreme minimum temperature;
[0016] The key climate factors affecting the peak period of fruit expansion include the maximum temperature ≥35℃ effective accumulated temperature, the daily maximum temperature ≥37℃ effective accumulated temperature, the cumulative sunshine and the number of days with daily precipitation ≥1.0mm;
[0017] The key climate influencing factors during the oil accumulation period include effective accumulated temperature with daily maximum temperature ≥35℃, number of days with daily maximum temperature ≥35℃, the longest consecutive days with daily maximum temperature ≥35℃, and the longest consecutive days without precipitation;
[0018] The key factors affecting fruit maturity include the number of days with daily precipitation ≥ 0.1 mm, cumulative precipitation, the longest continuous precipitation days with daily precipitation ≥ 0.1 mm, and the number of days with daily precipitation ≥ 10 mm.
[0019] As a further improvement, the final predicted climate influencing factors include the daily average temperature during the flowering period <10℃ negative accumulated temperature, the daily average temperature during the spring shoot budding period <5℃ negative accumulated temperature, the average maximum temperature during the young fruit period, the daily precipitation during the peak period of fruit expansion ≥37℃ effective accumulated temperature and the number of days with daily precipitation ≥0.1mm during the maturity period.
[0020] As a further improvement, both the main yield impact period data set and the total prediction data set use 80% of the data as a training set and 20% of the data as a test set.
[0021] Advantages of the present invention:
[0022] (1) A set of key climate risk factors for tea oil production was screened by comprehensively using four methods: correlation analysis, variance analysis, XGBoost factor importance analysis, and kernel density fitting.
[0023] (2) The XGBoost algorithm was used to construct disaster loss probability models for the six main growth periods of oil-tea, and the thresholds of the key influencing factors that cause production reduction were obtained based on the contribution and influence of each influencing factor on the predicted value in each model.
[0024] (3) The Hunan oil-tea disaster loss probability model and regression model constructed using the XGBoost algorithm have better prediction accuracy than the linear model.
[0025] (4) The oil-tea tree disaster loss model can be applied to the climate risk assessment work of oil-tea tree production. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Figure 1 The importance distribution map of 16 main climate factors affecting low temperature and heavy rainfall during the spring shoot sprouting period;
[0027] Figure 2 Importance distribution map of 16 main climate factors affecting flowering period, including low temperature, cloudy and rainy weather and little sunshine;
[0028] Figure 3 Importance distribution map of the nine main climate factors affecting low temperature during the young fruit stage;
[0029] Figure 4 Distribution map of the importance of 15 main climate factors affecting drought during the peak period of fruit expansion;
[0030] Figure 5 Importance distribution map of 14 main climate factors affecting high temperature during oil accumulation period;
[0031] Figure 6 The importance distribution map of the 10 main climate factors affecting the rainy and cloudy weather during the fruit ripening period;
[0032] Figure 7Importance distribution map of 24 major climate influencing factors;
[0033] Figure 8 ROC curve and AUC value are used to evaluate the model;
[0034] Fig. 9 Ranking diagram of the average absolute value of the SHAP values of each impact factor;
[0035] Fig.10 Comparison chart of measured and predicted relative meteorological output; DETAILED DESCRIPTION
[0036] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0037] 1. Established a meteorological element data set corresponding to the relative meteorological yield information of oil-tea
[0038] 1.1.1 Establishment of the relative meteorological yield data series of oil-tea camellia
[0039] The data on oil-tea camellia production were obtained from 31 oil-tea camellia plots in different counties and districts in Hunan Province from 2006 to 2016, provided by the Hunan Academy of Forestry Sciences. Figure 1 ) The per-acre yield data of fresh oil-tea fruit was determined by technicians from Hunan Academy of Forestry Sciences who selected representative 1-acre plots of various oil-tea sample plots and picked fresh fruit and weighed them. Since the planting years, varieties and soils of oil-tea in different plots are different, these factors will affect the yield of oil-tea. Therefore, when analyzing the relationship between climate risk factors and oil-tea yield, the oil-tea yield data is first processed to separate the influence of the above factors. The processing method of oil-tea yield data is as follows:
[0040] (1) There are 31 oil-tea sample plots. Due to the large difference in yield between improved oil-tea forests (referring to improved oil-tea varieties approved by the Forest Variety Approval Committee of the State Forestry Administration) and non-improved oil-tea forests (referring to other varieties that have not been approved), the yield data of 11 non-improved oil-tea sample plots were excluded.
[0041] (2) It takes 7-8 years for camellia oleifera to enter the high-yield period (referring to the stage when the camellia oleifera forest enters the stage of high and stable yield, generally referring to the eighth year after the planting of high-quality camellia oleifera). Before entering the high-yield period, as the camellia oleifera trees continue to grow, their yield also increases accordingly. Therefore, the yield data of camellia oleifera that has not entered the high-yield period are excluded.
[0042] (3) The difference in tea oil yield between different improved forest varieties is also large. In order to reduce the difference between different regions,
[0043] The impact of productivity levels in different periods on oil-tea production is characterized by relative meteorological production. Positive relative meteorological production is defined as an increase in production, while negative relative meteorological production is defined as a decrease in production, i.e., the loss rate. The relative meteorological production calculation formula is as follows:
[0044]
[0045] Where: y u is the relative meteorological output; s is the actual output, y t is the trend yield. The trend yield y is calculated using the 5-point moving average method. t The trend yield mainly reflects the impact of the improvement of agricultural technology level on yield, which is gradual and relatively stable.
[0046] After the above processing, a total of 83 sets of relative meteorological yield data of oil-tea measuring points were obtained (Table 1), including 41 sets of increase data and 42 sets of decrease data. The relative meteorological yield separated the interference of oil-tea planting year, variety and soil on the oil-tea disaster yield data, and it can be considered that the oil-tea yield is only affected by meteorological factors.
[0047] Table 1 Relative meteorological yield of oil-tea production points in Hunan from 2006 to 2016 (%)
[0048]
[0049]
[0050]
[0051] 1.1.2 Establishment of meteorological element grid data set
[0052] The meteorological element data used in previous related research on Hunan oil tea were mainly data from meteorological stations and farmland climate observation stations. In order to make the meteorological element data obtained at the oil tea production points more objective and true, the meteorological element small grid extrapolation technology was adopted to perform small grid interpolation on the daily average temperature, maximum temperature, minimum temperature, precipitation, and sunshine data of 97 meteorological stations and automatic observation stations in Hunan from 2006 to 2019.
[0053] The interpolation method adopted in the present invention performs 500m×500m grid interpolation on the daily average temperature, maximum temperature, minimum temperature, precipitation, and sunshine data from 2006 to 2019.
[0054] 1.1.3 Establishment of meteorological element dataset corresponding to the relative meteorological yield information of oil-tea
[0055] Based on the daily meteorological element grid data with a resolution of 500m, and according to the principle of proximity, the grid element data closest to each oil-tea production measurement point were selected as the element data of each production measurement point, and a data set of daily average temperature, maximum temperature, minimum temperature, precipitation, and sunshine for 31 production measurement points from 2006 to 2019 was established.
[0056] The entire growth period of oil-tea camellia from spring shoot growth to fruit maturity lasts for nearly 2 years. The vegetative growth and reproductive growth of oil-tea camellia influence and intertwine each other, and the meteorological conditions of each phenological period will affect the growth and development of oil-tea camellia, and then affect the yield of oil-tea camellia. Through a lot of previous research work, it is known that the spring shoot sprouting period (2.19-3.12), flowering period (10.28-12.20), young fruit period (2.28-3.31), fruit expansion peak period (6.5-8.8), oil accumulation period (8.7-10.23) and fruit maturity period (10.8-11.23) are the main growth periods that affect the yield of oil-tea camellia. When there are abnormal climates and meteorological disasters during these growth periods, oil-tea camellia often suffers from a reduction in yield.
[0057] In order to facilitate the subsequent work of finding climate risk factors affecting oil-tea yield, a daily meteorological element dataset for each major growth period of 31 production measurement points from 2006 to 2019 was established based on the existing daily meteorological element dataset for 31 production measurement points from 2006 to 2019.
[0058] 1.2 Key climate risk factors affecting oil-tea production
[0059] According to previous research results, the main meteorological disasters affecting oil-tea are low temperature and heavy rain during the spring shoot sprouting period, low temperature and rainy weather during the flowering period, low temperature during the young fruit period, drought during the peak period of fruit expansion, high temperature during the oil accumulation period, and rainy weather during the fruit ripening period. It can be seen that the climate factors that are not conducive to rapeseed production also mainly appear in these growth periods. In order to comprehensively find out the climate influencing factors that may cause the reduction of oil-tea production, based on the main meteorological disasters in the above growth periods, by consulting literature, combining previous research results and expert experience, as many influencing factors related to meteorological disasters as possible that can reflect the degree of disasters and characterize disasters are found, and the key climate influencing factors are screened out by correlation analysis, variance analysis, kernel density fitting and XGBoost modeling methods.
[0060] 1.2.1 Factors affecting low temperature and heavy rainfall during the spring shoot budding period
[0061] Low temperature and heavy precipitation during the spring shoot budding period will reduce the number of effective buds in the spring shoots and affect flower bud differentiation. The number of flower bud differentiation on the spring shoots is closely related to the probability of oil-tea damage. By consulting literature, combining previous research results and expert experience, 16 major climate influencing factors associated with low temperature and heavy rainfall during the spring shoot budding period were found (Table 2).
[0062] Table 2 Main climate influencing factors of low temperature and high rainfall during spring shoot budding period and results of variance analysis and correlation analysis
[0063]
[0064] (1) Correlation analysis
[0065] The relative meteorological yield series of the oil-tea measuring points in Hunan from 2006 to 2016 were correlated with the main climate influencing factors listed in Table 2. From the correlation results, it can be seen that the eight factors, including daily average temperature ≥5℃ active accumulated temperature, daily average temperature <5℃ negative accumulated temperature, average temperature, average maximum temperature, average minimum temperature, cumulative precipitation, number of days with daily precipitation ≥25mm and number of days with the longest continuous precipitation with daily precipitation ≥0.1mm, have a good correlation with the relative meteorological yield of oil-tea, and passed the significance test of α=0.05.
[0066] (2) Analysis of variance
[0067] The main climate influencing factors (Table 2) corresponding to the two types of samples with increased and decreased oil-tea production in Table 1 were analyzed by variance analysis to find out the climate factors with significant differences between the two types of samples with increased and decreased production. As can be seen from Table 2, after variance analysis, the probability values of the six factors, namely, the extreme minimum temperature, the number of days with the minimum temperature <0℃, the cumulative sunshine hours, the number of days with little sunshine, the longest continuous number of days with little sunshine, and the number of days with daily precipitation ≥10mm, were all greater than 0.05, indicating that there were no significant differences in these factors between the years with increased and decreased oil-tea production. The P values of the other 10 factors were all less than 0.05, indicating that these factors were significantly different between the years with increased and decreased oil tea production. Among them, the three factors of average maximum temperature, number of days without sunshine and number of days with sunshine passed the significance test of α=0.05, and the seven factors of daily average temperature ≥5℃ active accumulated temperature, daily average temperature <5℃ negative accumulated temperature, average temperature, average minimum temperature, accumulated precipitation, number of days with daily precipitation ≥25mm and number of days with daily precipitation ≥0.1mm and longest continuous precipitation passed the significance test of α=0.01.
[0068] (3) Importance analysis of climate influencing factors based on XGBoost model
[0069] Based on the 16 influencing factors in Table 2, the XGBoost method was used to construct a binary classification model for oil-tea production increase and reduction. 80% of the total samples were selected to train the model, and 20% of the samples were used to verify the model. The model obtained after training was verified, and the accuracy rate reached 82.4%. At this time, the importance of the factors output by XGBoost is shown in Figure 1 As can be seen from the figure, negative accumulated temperature with daily average temperature <5℃ plays the most important role in modeling, followed by average maximum temperature, extreme minimum temperature, average minimum temperature, accumulated precipitation, etc. According to the order of importance of the 16 factors from large to small, 1 factor, 2 factors, and 16 factors were selected to re-model. The model constructed by selecting the first 10 or 12 factors according to the importance of the factors has the highest accuracy, reaching 88.2%.
[0070] The kernel density estimation method was used to fit the cumulative probability density distribution of the top 12 factors. Among the 12 factors, only the extreme minimum temperature and the longest continuous few days had no significant difference in the probability density distribution of the yield increase and yield decrease samples. According to the results of the factor importance analysis, it is believed that the 10 factors of daily average temperature <5℃ negative accumulated temperature, average maximum temperature, average minimum temperature, cumulative precipitation, cumulative sunshine, sunny days, few days, average temperature, daily average temperature ≥5℃ active accumulated temperature, and heavy rain and rainy days have a greater impact on the low temperature and heavy rain during the spring shoot budding period.
[0071] Based on the results of XGBoost factor importance analysis, combined with the results of variance and correlation analysis, factors with good correlation, significant variance test and relatively important were selected as key influencing factors of low temperature and rainy weather during the spring shoot sprouting period. After comprehensive analysis, four factors, namely daily average temperature <5℃ negative accumulated temperature, average maximum temperature, average minimum temperature and accumulated precipitation, were selected as key influencing factors.
[0072] 1.2.2 Factors affecting low temperature, rainy weather and low sunlight during flowering period
[0073] Flowering period is the key phenological period that affects the probability of oil-tea damage. Low temperature, heavy rainfall and little sunshine during the flowering period will cause a large number of oil-tea flowers to fall and low fruit setting rate, resulting in a reduction in production in the next year. If low temperature, rain, snow and freezing weather occur during the flowering and pollination stage of oil-tea, on the one hand, pollen cracking will be inhibited and the flower buds will be frozen; on the other hand, it will affect the activities of insects that serve as cross-pollination media. At the same time, precipitation will also wash away the stigma fluid and pollen, causing pollination and fertilization to fail to proceed normally, thereby causing a large reduction in oil-tea production. Low temperatures below -3.5 to -2.5℃ during the flowering period will freeze the flower buds and kill pollinating bees, resulting in a reduction in production in the next year. As the minimum temperature drops and the duration increases, the reduction in production increases, generally by more than 30%, or even by nearly 90%. By consulting literature, combining previous research results and expert experience, 16 major climate influencing factors associated with low temperature, cloudy and rainy weather and little sunshine during the flowering period were found (Table 3).
[0074] Table 3 Main climate factors affecting flowering period, low temperature, cloudy and rainy weather and low sunshine, and the results of variance analysis and correlation analysis
[0075]
[0076]
[0077] (1) Correlation analysis
[0078] The relative meteorological yield series of the oil-tea measuring points in Hunan from 2006 to 2016 were correlated with the main climate influencing factors listed in Table 3. From the relevant results, it can be seen that only the three factors of daily average temperature ≥ 10℃ active accumulated temperature, daily average temperature 10℃ negative accumulated temperature, and average temperature have a good correlation with the relative meteorological yield of oil-tea, passing the significance test of α = 0.05. The two factors of average maximum temperature and average minimum temperature only passed the significance test of α = 0.1, and the correlation between other factors and the relative meteorological yield of oil-tea is not obvious.
[0079] (2) Analysis of variance
[0080] The main climate influencing factors (Table 3) corresponding to the two types of samples of oil-tea increase and decrease in Table 1 were analyzed by variance analysis to find out the climate factors with significant differences between the two types of samples. As can be seen from Table 3, after variance analysis, the probability values of the six factors of daily average temperature ≥ 10℃ active accumulated temperature, daily average temperature 10℃ negative accumulated temperature, average temperature, average maximum temperature, average minimum temperature, and number of days with sunshine are all less than 0.05. These factors have significant differences in the years of oil-tea increase and decrease. The probability values of cumulative sunshine hours, days without sunshine, and days with little sunshine are all less than 0.1. There are certain differences in these three factors in the years of oil-tea increase and decrease. The other seven factors are not significantly different in the years of oil-tea increase and decrease.
[0081] (3) Importance analysis of climate influencing factors based on XGBoost model
[0082] Based on the 16 influencing factors in Table 3, the XGBoost method was used to construct a binary classification model for oil-tea production increase and reduction. 80% of the total samples were selected to train the model, and 20% of the samples were used to verify the model. The model obtained after training was verified, and the accuracy rate reached 82.4%. At this time, the importance of the factors output by XGBoost is shown in Figure 2As can be seen from the figure, negative accumulated temperature with daily average temperature <10℃ plays the most important role in modeling, followed by active accumulated temperature with daily average temperature ≥10℃, extreme minimum temperature, average minimum temperature, cumulative precipitation, and days with little sunshine. According to the order of importance of the 16 factors from large to small, 1 factor, 2 factors, and 16 factors were selected to re-model. The model constructed by selecting the first 3 factors according to the importance of the factors has the highest accuracy, reaching 88.2%.
[0083] Table 4 Accuracy of model construction with different numbers of influencing factors
[0084]
[0085]
[0086] Since the first three factors mainly reflect the low temperature condition, the characteristics of cloudy, rainy and few sunshine are not reflected, so the selection of important factors is expanded to the first 7 factors, and the accuracy of model fitting can reach 82.4%. The kernel density method is used to fit the cumulative probability density distribution of the first 7 factors. Among the 7 factors, only the probability density distribution of the increase and decrease samples of the extreme minimum temperature and cumulative precipitation factors has no obvious difference. According to the results of the importance analysis of the factors, it is believed that the five factors of daily average temperature <10℃ negative accumulated temperature, daily average temperature ≥10℃ active accumulated temperature, average minimum temperature, few sunshine days, and cumulative sunshine have a greater impact on low temperature, cloudy, rainy and few sunshine during flowering.
[0087] Based on the results of XGBoost factor importance analysis, combined with the results of variance and correlation analysis, factors with good correlation, significant variance test and relatively important were selected as key influencing factors of low temperature, rainy and low light during flowering period. After comprehensive analysis, four factors were selected as key influencing factors: daily average temperature <10℃ negative accumulated temperature, daily average temperature ≥10℃ active accumulated temperature, average minimum temperature and low light days.
[0088] 1.2.3 Factors affecting low temperature during the young fruit stage
[0089] Frost, ice and snow in late winter and early spring, especially frost in March and April, will seriously affect the pollination and young fruit formation of tea oil trees. If low temperature weather occurs, it is easy to cause the young fruit to freeze and produce abnormal fruit drop, which has a great impact on the probability of damage to tea oil trees. By consulting literature, combining previous research results and expert experience, 9 major climate influencing factors associated with low temperature during the young fruit stage were found (Table 5).
[0090] Table 5 Main climate influencing factors of low temperature during young fruit stage and results of variance analysis and correlation analysis
[0091]
[0092]
[0093] (1) Correlation analysis
[0094] The relative meteorological yield series of the oil-tea measuring points in Hunan from 2006 to 2016 were correlated with the main climate influencing factors listed in Table 5. From the relevant results, it can be seen that the three factors of daily average temperature ≥5℃ active accumulated temperature, average temperature, and average maximum temperature have a good correlation with the relative meteorological yield of oil-tea, passing the significance test of α = 0.05. The extreme minimum temperature only passed the significance test of α = 0.1, and the correlation between other factors and the relative meteorological yield of oil-tea is not obvious.
[0095] (2) Analysis of variance
[0096] The main climate influencing factors (Table 5) corresponding to the two types of samples with increased and decreased oil-tea production in Table 1 were analyzed by variance analysis to find out the climate factors with significant differences between the two types of samples with increased and decreased production. As can be seen from Table 5, after variance analysis, the probability values of the four factors of daily average temperature ≥ 5℃ active accumulated temperature, average temperature, average maximum temperature and extreme minimum temperature were all less than 0.05, and these factors were significantly different in the years with increased and decreased oil-tea production. The other five factors were not significantly different in the years with increased and decreased oil-tea production.
[0097] (3) Importance analysis of climate influencing factors based on XGBoost model
[0098] Based on the 9 influencing factors in Table 5, the XGBoost method was used to construct a binary classification model for oil-tea production increase and reduction. 80% of the total samples were selected to train the model, and 20% of the samples were used to verify the model. The model obtained after training was verified, and the accuracy rate reached 70.6%. At this time, the importance of the factors output by XGBoost is shown in Figure 3 As can be seen from the figure, the average maximum temperature plays the most important role in modeling, followed by the extreme minimum temperature, daily average temperature <5℃ negative accumulated temperature, average minimum temperature, average temperature, etc. According to the order of importance of the 9 factors from large to small, 1 factor, 2 factors, and 9 factors were selected to re-model. The model constructed by selecting the first 4 factors according to the importance of the factors has the highest accuracy, reaching 82.4%.
[0099] Analysis of the cumulative probability density distribution of the first four factors shows that there is no significant difference in the probability density distribution of the increase and decrease samples of daily average temperature <5℃ negative accumulated temperature and average minimum temperature. In order to obtain more important factors, the important factor analysis was expanded to the first 6 factors, and the fitting accuracy of the model was 76.5%. There is a certain difference in the probability density distribution of average temperature and daily average temperature ≥5℃ active accumulated temperature. According to the results of the importance analysis of the factors, it is believed that the four factors of average maximum temperature, extreme minimum temperature, average temperature and daily average temperature ≥5℃ active accumulated temperature have a greater impact on low temperature during the young fruit stage.
[0100] 1.2.4 Factors affecting drought during the peak period of fruit expansion
[0101] The peak period of fruit expansion requires more water. If there is little rain and high temperature at this time, the fruit volume cannot expand normally, which is not conducive to fruit formation or the fruit is not full, resulting in small fruit or even "dry fruit", affecting the formation of yield. By consulting literature, combining previous research results and expert experience, 15 major climate influencing factors associated with drought during the peak period of fruit expansion were found (Table 6).
[0102] Table 6 Main climate influencing factors of drought during the peak period of fruit expansion and the results of variance analysis and correlation analysis
[0103]
[0104] (1) Correlation analysis
[0105] The relative meteorological yield series of the oil-tea measuring points in Hunan from 2006 to 2016 were correlated with the main climate influencing factors listed in Table 6. From the relevant results, it can be seen that the six factors of daily maximum temperature ≥35℃ effective accumulated temperature, daily maximum temperature ≥37℃ effective accumulated temperature, number of days with daily maximum temperature ≥35℃, number of days with daily maximum temperature ≥37℃, cumulative sunshine and number of days with daily precipitation ≥1.0mm have a good correlation with the relative meteorological yield of oil-tea, passing the significance test of α=0.05. The correlation between other factors and the relative meteorological yield of oil-tea is not obvious.
[0106] (2) Analysis of variance
[0107] The main climate influencing factors (Table 6) corresponding to the two types of samples with increased and decreased oil-tea production in Table 1 were analyzed by variance analysis to find out the climate factors with significant differences between the two types of samples with increased and decreased production. As can be seen from Table 6, after variance analysis, only the probability values of the two factors, namely, the daily maximum temperature ≥ 37℃ effective accumulated temperature and the number of days with daily precipitation ≥ 1.0mm, were less than 0.05. These factors had significant differences in the years with increased and decreased oil-tea production. The other factors were not significantly different in the years with increased and decreased oil-tea production.
[0108] Comparing the results of variance analysis with the results of correlation analysis, it can be seen that the two factors with significant differences in the years of increased production and decreased production have good correlation, but the four factors with good correlation have no obvious differences in the years of increased production and decreased production, and the years of increased production and decreased production are not easy to distinguish.
[0109] (3) Importance analysis of climate influencing factors based on XGBoost model
[0110] The XGBoost method was used to construct a binary classification model for increasing and decreasing oil-tea production based on the 15 influencing factors in Table 6. 80% of the total samples were selected to train the model, and 20% of the samples were used to verify the model. The model obtained after training was verified to have an accuracy rate of 70.6%. At this time, the importance of the factors output by XGBoost is shown in Figure 4 As can be seen from the figure, the number of days with daily precipitation ≥ 1.0 mm plays the most important role in modeling, followed by extreme maximum temperature, rainy days, cumulative sunshine, cumulative precipitation, etc. According to the order of importance of the 15 factors from large to small, 1 factor, 2 factors, and 15 factors are selected to re-model. The model constructed by selecting the first 6 factors or all factors according to the importance of the factors has the highest accuracy, which can reach 70.6%.
[0111] Based on the results of XGBoost factor importance analysis, combined with the results of variance and correlation analysis, factors with good correlation, significant variance test and relatively important were selected as the key influencing factors of drought during the peak period of fruit expansion. After comprehensive analysis, four factors were selected as the key influencing factors: the number of days with daily precipitation ≥ 1.0 mm, the effective accumulated temperature of the daily maximum temperature ≥ 35 °C, the effective accumulated temperature of the daily maximum temperature ≥ 37 °C and the accumulated sunshine.
[0112] 1.2.5 Factors affecting high temperature during oil accumulation
[0113] The oil accumulation period is an important stage for the probability of oil-tea damage and the formation of quality. Insufficient water is the main reason for the high probability of oil-tea damage, the disparity between good and bad years, and the fluctuation of oil content. High temperature and drought will lead to the "dry oil" phenomenon with a decrease in oil content, which will affect the increase in oil yield and cause a decrease in yield. By consulting literature, combining previous research results and expert experience, 14 major climate influencing factors associated with high temperature during the oil accumulation period were found (Table 7).
[0114] Table 7 Main climate influencing factors of high temperature during oil accumulation period and results of variance analysis and correlation analysis
[0115]
[0116] (1) Correlation analysis
[0117] The relative meteorological yield series of the oil-tea measuring points in Hunan from 2006 to 2016 were correlated with the main climate influencing factors listed in Table 7. From the relevant results, it can be seen that the six factors, namely, effective accumulated temperature of daily maximum temperature ≥35℃, effective accumulated temperature of daily maximum temperature ≥37℃, extreme maximum temperature, number of days with daily maximum temperature ≥35℃, number of days with daily maximum temperature ≥37℃, and the longest consecutive number of days with daily maximum temperature ≥35℃, have a good correlation with the relative meteorological yield of oil-tea, and pass the significance test of α=0.05. The correlation between other factors and the relative meteorological yield of oil-tea is not obvious.
[0118] (2) Analysis of variance
[0119] The main climate influencing factors (Table 7) corresponding to the two types of samples with increased and decreased oil-tea production in Table 1 were analyzed by variance analysis to find out the climate factors with significant differences between the two types of samples with increased and decreased production. As can be seen from Table 7, after variance analysis, only the probability values of the three factors of daily maximum temperature ≥ 35℃ effective accumulated temperature, daily maximum temperature ≥ 37℃ effective accumulated temperature and number of days with daily maximum temperature ≥ 37℃ were less than 0.05. These factors had significant differences in the years with increased and decreased oil-tea production. The other factors were not significantly different in the years with increased and decreased oil-tea production.
[0120] (3) Importance analysis of climate influencing factors based on XGBoost model
[0121] Based on the 14 influencing factors in Table 7, the XGBoost method was used to construct a binary classification model for oil-tea production increase and reduction. 80% of the total samples were selected to train the model, and 20% of the samples were used to verify the model. The model obtained after training was verified, and the accuracy rate reached 64.7%. At this time, the importance of the factors output by XGBoost is shown in Figure 5 As can be seen from the figure, the effective accumulated temperature with daily maximum temperature ≥ 35℃ plays the most important role in modeling, followed by the longest continuous number of days without precipitation, rainy days, average maximum temperature, rainless days, etc. According to the order of importance of the 14 factors from large to small, 1 factor, 2 factors, and 14 factors were selected to re-model. The model constructed by selecting the first 1 or 4 factors according to the importance of the factors has the highest accuracy, which can reach 70.6%.
[0122] Based on the results of XGBoost factor importance analysis, combined with the results of variance and correlation analysis, factors with good correlation, significant variance test and relatively important were selected as key influencing factors of high temperature during the accumulation period of oil and fat. After comprehensive analysis, four factors, namely, effective accumulated temperature with daily maximum temperature ≥ 35℃, the longest consecutive number of days without precipitation, the longest consecutive number of days with daily maximum temperature ≥ 35℃, and the number of days with daily maximum temperature ≥ 35℃, were selected as key influencing factors.
[0123] 1.2.6 Factors affecting rainy weather during fruit ripening
[0124] If there is a lot of precipitation during the fruit ripening period, the oil-tea fruit that is not picked in time will become moldy and the oil will become rancid, which will have a certain impact on the probability of damage and quality of oil-tea. By consulting literature, combining previous research results and expert experience, 10 major climate influencing factors associated with cloudy and rainy weather during the fruit ripening period were found (Table 8).
[0125] Table 8 Main climate influencing factors of cloudy and rainy weather during fruit ripening period and results of variance analysis and correlation analysis
[0126]
[0127]
[0128] (1) Correlation analysis
[0129] The relative meteorological yield series of the oil-tea measuring points in Hunan from 2006 to 2016 were correlated with the main climate influencing factors listed in Table 8. From the relevant results, it can be seen that the number of days without sunshine has a good correlation with the relative meteorological yield of oil-tea, passing the significance test of α=0.05. The number of days with daily precipitation ≥0.1mm has a certain correlation with the relative meteorological yield of oil-tea, passing the significance test of α=0.1.
[0130] (2) Analysis of variance
[0131] The main climate influencing factors (Table 8) corresponding to the two types of samples with increased and decreased oil-tea production in Table 1 were analyzed by variance analysis to find out the climate factors with significant differences between the two types of samples with increased and decreased production. As can be seen from Table 8, after variance analysis, only the probability values of the two factors, cumulative precipitation and the number of days with daily precipitation ≥ 10 mm, were less than 0.05. These two factors had significant differences in the years with increased and decreased oil-tea production. The other factors were not significantly different in the years with increased and decreased oil-tea production.
[0132] (3) Importance analysis of climate influencing factors based on XGBoost model
[0133] The XGBoost method was used to construct a binary classification model for increasing and decreasing oil-tea production using the 10 influencing factors in Table 8. 80% of the total samples were selected to train the model, and 20% of the samples were used to verify the model. The model obtained after training was verified to have an accuracy rate of 70.6%. At this time, the importance of the factors output by XGBoost is shown in Figure 6As can be seen from the figure, rainy days play the most important role in modeling, followed by cumulative precipitation, moderate rainy days, cumulative sunshine, etc. According to the order of importance of the 10 factors from large to small, 1 factor, 2 factors, and 10 factors are selected to re-model the model. The model constructed by selecting the first 4 factors according to the importance of the factors has the highest accuracy, which can reach 76.5%.
[0134] The cumulative probability density distribution of the first four factors was analyzed. Except for the cumulative sunshine, the probability density distribution of the yield increase and yield decrease samples of the other three factors was significantly different. The cumulative probability density distribution of the last six factors was analyzed. The probability density distribution of the yield increase and yield decrease samples of the longest continuous precipitation days, the number of days with little sunshine, and the number of days with no sunshine was significantly different. Based on the results of XGBoost factor importance analysis, combined with the results of variance and correlation analysis, factors with good correlation, significant variance test, and relatively important were selected as the key influencing factors of rainy days during fruit ripening. After comprehensive analysis, the four factors of rainy days, cumulative precipitation, moderate rainy days, and the longest continuous precipitation days were the key influencing factors.
[0135] 1.2.7 Result analysis:
[0136] Four methods, namely correlation analysis, variance analysis, XGBoost factor importance analysis and kernel density fitting, were used to conduct a comprehensive analysis of the climate influencing factors in the six main growth periods (spring shoot budding period, flowering period, young fruit period, peak period of fruit expansion, oil accumulation period and fruit maturity period) that affect the probability of oil-tea damage. Finally, four key climate influencing factors were selected in each main growth period, and a total of 24 factors were selected as the key climate influencing factors affecting the probability of oil-tea damage (Table 9).
[0137] Table 9 Key climate factors affecting the probability of oil-tea disaster damage
[0138]
[0139] 2.3 Construction of Camellia oleifera disaster damage model
[0140] The main meteorological disasters affecting camellia oleifera are low temperature and heavy rain during the spring shoot sprouting period, low temperature and rainy weather with little sunlight during the flowering period, low temperature during the young fruit period, drought during the peak period of fruit expansion, high temperature during the oil accumulation period, and rainy weather during the fruit ripening period. These disasters are the main reasons for the reduction in camellia oleifera production. Based on the key climate factors that affect the probability of camellia oleifera disasters during each major growth period (Table 9), a camellia oleifera climate disaster loss model was constructed.
[0141] 1.3.1 Damage model of major meteorological disasters for oil-tea plants
[0142] 1.3.1.1 Low temperature and heavy rain damage model during the spring shoot sprouting period
[0143] According to statistical analysis, daily average temperature <5℃ negative accumulated temperature, average maximum temperature, average minimum temperature, and accumulated precipitation are the key influencing factors causing low temperature and heavy rain during the spring shoot budding period. The positive value of relative meteorological yield in Table 1 is defined as the yield increase rate, and the negative value is defined as the disaster loss rate (yield reduction rate), indicating that disaster losses have occurred. Using the four key influencing factors, the XGBoost algorithm is used to construct a probability model for low temperature and heavy rain disaster losses during the spring shoot budding period.
[0144] Similar to the factor importance analysis using XGBoost, 80% of the total samples were selected to train the model, and 20% of the samples were used to verify the model. The model obtained after training was verified, and the accuracy rate reached 82.4%.
[0145] 1.3.1.2 Low temperature, rainy and low sunlight damage model during flowering period
[0146] According to statistical analysis, daily average temperature <10℃ negative accumulated temperature, daily average temperature ≥10℃ accumulated temperature, average minimum temperature, and few days of sunshine are the key influencing factors causing low temperature, rainy and few sunshine during flowering period. The positive value of relative meteorological yield in Table 1 is defined as the yield increase rate, and the negative value is defined as the disaster loss rate (yield reduction rate), indicating that disaster losses have occurred. Using the four key influencing factors, the XGBoost algorithm is used to construct a probability model for the occurrence of low temperature, rainy and few sunshine disaster losses during flowering period. 80% of the total samples are selected to train the model, and 20% of the samples are used to verify the model. The model obtained after training is verified, and the accuracy rate reaches 82.4%.
[0147] 1.3.1.3 Low temperature damage model during young fruit stage
[0148] According to statistical analysis, the average maximum temperature, the extreme minimum temperature, the average temperature, and the daily average temperature ≥ 5℃ accumulated temperature are the key influencing factors causing low temperature in the young fruit stage. The positive value of the relative meteorological yield in Table 1 is defined as the yield increase rate, and the negative value is defined as the disaster loss rate (yield reduction rate), indicating that a disaster loss has occurred. Using the four key influencing factors, the XGBoost algorithm is used to construct a probability model for low temperature disasters in the young fruit stage. 80% of the total samples are selected to train the model, and 20% of the samples are used to verify the model. The model obtained after training is verified, and the accuracy rate reaches 82.4%.
[0149] 1.3.1.4 Drought damage model during the peak fruit expansion period
[0150] According to statistical analysis, the daily maximum temperature ≥35℃ effective accumulated temperature, the daily maximum temperature ≥37℃ effective accumulated temperature, the cumulative sunshine, and the number of days with daily precipitation ≥1.0mm are the key influencing factors causing drought during the peak period of fruit expansion. The positive value of the relative meteorological yield in Table 1 is defined as the yield increase rate, and the negative value is defined as the disaster loss rate (yield reduction rate), indicating that disaster losses have occurred. Using the four key influencing factors, the XGBoost algorithm is used to construct a probability model for drought damage during the peak period of fruit expansion. 80% of the total samples are selected to train the model, and 20% of the samples are used to verify the model. The model obtained after training is verified to have an accuracy rate of 76.5%.
[0151] 1.3.1.5 High temperature damage model during oil accumulation period
[0152] According to statistical analysis, the effective accumulated temperature of daily maximum temperature ≥ 35℃, the longest continuous number of days without precipitation, the longest continuous number of days with daily maximum temperature ≥ 35℃, and the number of days with daily maximum temperature ≥ 35℃ are the key influencing factors causing high temperature during the oil accumulation period. The positive value of relative meteorological yield in Table 1 is defined as the yield increase rate, and the negative value is defined as the disaster loss rate (yield reduction rate), indicating that disaster losses have occurred. Using the four key influencing factors, the XGBoost algorithm is used to construct a probability model for drought disaster losses during the peak period of fruit expansion. 80% of the total samples are selected to train the model, and 20% of the samples are used to verify the model. The model obtained after training has been verified to have an accuracy rate of 70.6%.
[0153] 1.3.1.6 Rainy weather damage model during fruit ripening period
[0154] According to statistical analysis, the number of days with daily precipitation ≥ 0.1 mm, cumulative precipitation, the longest continuous precipitation days with daily precipitation ≥ 0.1 mm, and the number of days with daily precipitation ≥ 10 mm are the key influencing factors causing rainy weather during the fruit ripening period. The positive value of the relative meteorological yield in Table 1 is defined as the yield increase rate, and the negative value is defined as the disaster loss rate (yield reduction rate), indicating that disaster losses have occurred. Using the four key influencing factors, the XGBoost algorithm is used to construct a probability model for rainy weather damage during the fruit ripening period. 80% of the total samples are selected to train the model, and 20% of the samples are used to verify the model. The model obtained after training is verified to have an accuracy rate of 88.2%.
[0155] 1.3.1.7 Result Analysis
[0156] Based on the XGBoost algorithm, the damage probability models of six major meteorological disasters were constructed, including low temperature and heavy rain during the spring shoot sprouting period of Camellia oleifera, low temperature and rainy weather during the flowering period, low temperature during the young fruit period, drought during the peak period of fruit expansion, high temperature during the oil accumulation period, and rainy weather during the fruit ripening period. The prediction accuracy of all models was above 70%, with the highest reaching 88.2%. The SHAP value was used to explain the prediction results of the model, and the contribution and influence of each influencing factor in each model to the predicted value were analyzed. Some key influencing factor thresholds that can be used to determine whether the increase or decrease in production is likely to occur were obtained (Table 10). Through these thresholds, it is possible to intuitively judge whether a disaster is likely to occur before using the prediction model.
[0157] Table 10 Thresholds of key climate factors
[0158]
[0159] 1.3.2 Camellia oil tree damage model
[0160] 1.3.2.1 Oil-tea tree damage probability model
[0161] Based on the 24 key climate factors that affect the probability of oil-tea damage during the main growth period (Table 9), the XGBoost algorithm was used to build the oil-tea damage probability model. 80% of the total samples were selected to train the model, and 20% of the samples were used to verify the model.
[0162] There are many hyperparameters in the XGBoost algorithm, including regularization terms, learning rates, the number of decision trees, and tree structure properties. Hyperparameters define specific basic properties of the model, such as complexity or learning ability. They are parameters that need to be determined before starting the learning process. The significance of adjusting hyperparameters is to minimize the expected risk, balance the model optimization and model complexity, and avoid underfitting and overfitting as much as possible. Grid search is the most widely used hyperparameter search algorithm based on cross-validation. This exhaustive parameter adjustment algorithm can find the best combination by looping through the possibilities of each parameter combination and finding the global maximum or minimum value. In the process, 5-fold cross-validation is used, that is, the training set is divided into 5 equal parts, one of which is taken as the validation set, and the other 4 are new training sets. After 5 tests on different validation sets, the hyperparameter combination corresponding to the best result is taken. It has been verified that the default parameter combination has the highest prediction accuracy of 76.5%.
[0163] By outputting the factor importance of XGBoost, see Figure 7According to the order of importance of the 24 factors from large to small, 1 factor, 2 factors, and 24 factors were selected to re-model the model. The accuracy of each model is shown in Table 11. As can be seen from the table, the model constructed by selecting the first 16 factors according to the importance of the factors has the highest accuracy, which can reach 88.2%. Based on the principle of selecting as few factors as possible to construct the model without losing the accuracy of the model, the first 16 factors (Table 11) were finally selected to construct the oil-tea climate disaster probability model.
[0164] Table 11 Accuracy of model construction with different numbers of influencing factors
[0165]
[0166]
[0167] Table 12 Influencing factors of constructing oil-tea disaster damage model
[0168]
[0169] The oil-tea tree disaster loss probability model was reconstructed using 16 factors. Through parameter adjustment and verification, it was found that the model prediction accuracy obtained by the default parameter combination was still the highest, reaching 88.2%. The ROC curve and AUC value were also used to evaluate the model ( Figure 8 ), the solid line in the figure is the ROC curve. The closer the ROC curve is to the (0, 1) point and the more it deviates from the 45-degree diagonal line (gray dotted line), the better the model prediction effect. AUC is the area under the ROC curve. The closer the AUC is to 1, the better the model prediction effect. According to the evaluation rules, the ROC curve of the constructed oil-tea disaster probability model deviates far from the 45-degree diagonal line and is closer to the (0, 1) point. The AUC value also reaches 0.87, indicating that the model has a good prediction effect.
[0170] Among the 16 factors selected to build the model, there are 3 factors related to high temperature and drought, and the rest are factors related to low temperature and rain. It can be seen that low temperature and rain are the main disaster types that cause oil-tea production reduction. From the SHAP values of each factor ( Fig.10 ), the daily average temperature <5℃ negative accumulated temperature during the spring shoot budding period contributed the most to the model, and its absolute value of SHAP value exceeded 1.0, followed by the average minimum temperature during the flowering period, the number of days with little sunlight during the flowering period, the average maximum temperature during the young fruit period, the number of days with daily precipitation ≥1.0mm during the peak period of fruit expansion, and the daily average temperature <10℃ negative accumulated temperature during the flowering period. The absolute values of the SHAP values of these factors all exceeded 0.5, and had the most obvious impact on the reduction of oil-tea production. From the absolute value of the cumulative SHAP values of the influencing factors of each growth period, the flowering period contributed the most to the model, followed by the spring shoot budding period, the young fruit period, the fruit ripening period, and the peak period of fruit expansion, and the oil accumulation period contributed the least.
[0171] 1.3.2.2 Camellia oleifera disaster loss regression model
[0172] Using 24 factors, the XGBoost algorithm was used to construct a regression model for oil-tea damage. Through parameter adjustment and verification, the best key hyperparameter combination is shown in Table 13, and MAE (mean absolute error), RMSE (root mean square error) and R-Squared (R square value) were used to evaluate the model (Table 14). From the evaluation parameters in the table, it can be seen that the prediction effect of the constructed model is not very ideal. The SHAP values of each factor in the constructed model show that ( Fig.10 ), among the 24 factors, only five factors, namely, daily average temperature during flowering period <10℃ negative accumulated temperature, daily average temperature during spring shoot budding period <5℃ negative accumulated temperature, average maximum temperature during young fruit stage, daily precipitation ≥37℃ effective accumulated temperature during fruit expansion peak period, and number of days with daily precipitation ≥0.1mm during maturity period, contributed to the model. These five factors were selected to reconstruct the model.
[0173] Through parameter adjustment verification, the forecast effect of the reconstructed model is the best with the default parameters. Compared with the 24-factor construction model, the forecast effect is improved. From the comparison chart of the measured value and the predicted value of relative meteorological output ( Fig.10 ), the predicted value and the measured value are basically consistent in the trend of increasing or decreasing production, and are also relatively close in magnitude.
[0174] Table 13 XGBoost optimal hyperparameter configuration
[0175]
[0176] Table 14 MAE, RMSE and R-Squared values of various models
[0177]
[0178] 1.3.2.3 Result analysis:
[0179] Judging from the constructed oil-tea disaster loss probability model and regression model, the probability model has a better forecasting effect and can accurately determine whether a disaster has occurred. Although the effect of the regression model needs to be further improved, it is a good supplement to the probability forecasting model. When it is determined that a disaster has occurred, it can determine the extent of the disaster from a specific magnitude and meet the needs of precise services.
[0180] Judging from the influencing factors used in the final model construction of the existing samples, low temperature and rainy weather are the main causes of the reduction in oil-tung oil production, while high temperature and drought are not prominent. Judging from the contribution of various influencing factors when modeling, the influencing factor of the flowering period contributes the most, followed by the spring shoot budding period, then the young fruit period, the fruit ripening period, and the peak period of fruit expansion. The contribution of the oil accumulation period is relatively small.
[0181] 1.4 Climate risk assessment for oil-tea plants
[0182] 2.4.1 2021 Camellia oleifera climate risk assessment
[0183] Based on the probability model of disaster losses in the main growing periods of oil tea, disaster loss assessment work can be carried out for the main meteorological disasters in each growing period in 2021. The following climate risk assessment of oil tea is carried out using the drought disaster loss assessment during the peak fruit swelling period in 2021 as an example.
[0184] During the peak period of fruit enlargement in 2021, the average temperature was 1-2°C higher than usual in most parts of the province; the precipitation was lower than usual in most parts of the province, with most parts of Shaoyang experiencing more than 50% less precipitation; the number of days with daily precipitation ≥1 mm was lower in most parts of the province, with most parts of southern Hunan and western Hunan experiencing 5-10 days less precipitation.
[0185] Through drought monitoring during the peak period of fruit expansion, it is known that the number of days with daily precipitation ≥ 1 mm in northwestern Hunan, northern Hunan and southeastern Hunan mostly exceeds 15 days, while other places are mostly 11-14 days. In some places in Hengyang, Shaoyang, Yongzhou and Huaihua, it is ≤ 10 days, reaching the standard of drought disaster.
[0186] According to the disaster loss probability model during the peak period of fruit expansion, the probability of drought losses in most parts of the province is over 50%, especially in southern Zhuzhou, Hengyang, most of Shaoyang, and most of Huaihua, where the probability of disaster losses is over 80%.
[0187] To sum up, the temperature was relatively high, the precipitation was relatively low, and the rainy days were relatively few during the period of fruit expansion of Camellia oleifera in Hunan. Drought occurred in some parts of Hengshao region. Combined with the results of the disaster loss probability model, it is believed that the risk of drought disaster losses in the Camellia oleifera planting areas of Hengyang, Shaoyang and Huaihua is relatively high.
[0188] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification, equivalent change and modification made to the above embodiment by any technician familiar with this patent without departing from the scope of the technical solution of the present invention according to the technical essence of the present invention still falls within the scope of the solution of the present invention.
Claims
1. A method for predicting oil-tea damage, characterized in that: The steps include: Step 1: Collect the climate influencing factors of oil-tea during the main yield-influencing periods and the corresponding relative meteorological yield of oil-tea; the main yield-influencing periods include spring shoot sprouting period, flowering period, young fruit period, fruit expansion peak period, oil accumulation period and fruit maturity period; Step 2: Analyze and obtain the key climate influencing factors of each major yield influencing period and combine them with the corresponding oil-tea relative meteorological yield to form a dataset of the major yield influencing period; The key climate factors affecting the spring shoot budding period include daily average temperature <5°C accumulated temperature, average maximum temperature, average minimum temperature, and accumulated precipitation; The key climate factors affecting the flowering period include daily average temperature <10℃ accumulated temperature, daily average temperature ≥10℃ active accumulated temperature, average minimum temperature, and few days of sunshine; The key climate factors affecting the young fruit stage include daily average temperature ≥5℃ active accumulated temperature, average temperature, average maximum temperature and extreme minimum temperature; The key climate factors affecting the peak period of fruit expansion include the maximum temperature ≥35℃ effective accumulated temperature, the daily maximum temperature ≥37℃ effective accumulated temperature, the cumulative sunshine and the number of days with daily precipitation ≥1.0mm; The key climate influencing factors during the oil accumulation period include effective accumulated temperature with daily maximum temperature ≥35℃, number of days with daily maximum temperature ≥35℃, the longest consecutive days with daily maximum temperature ≥35℃, and the longest consecutive days without precipitation; The key factors affecting the fruit maturity period include the number of days with daily precipitation ≥ 0.1 mm, cumulative precipitation, the longest continuous precipitation days with daily precipitation ≥ 0.1 mm, and the number of days with daily precipitation ≥ 10 mm; Step 3: Establish a disaster loss prediction model for each major production influencing period based on the relative meteorological yield of oil-tea and the key climate influencing factors of each major production influencing period, and then obtain the final disaster loss prediction model for each major production influencing period through training of the main production influencing period data set; Step 4, selecting the climate influencing factors of the whole growth period of Camellia oleifera from the key climate influencing factors of each major yield influencing period and combining them with the corresponding relative meteorological yield of Camellia oleifera to form a prediction data set; establishing a total disaster loss prediction model for Camellia oleifera according to the selected climate influencing factors and the corresponding relative meteorological yield of Camellia oleifera, and then obtaining the final total disaster loss prediction model for Camellia oleifera through training of the prediction data set; the climate influencing factors of the whole growth period include the daily average temperature of the flowering period <10℃ accumulated temperature, the daily average temperature of the spring shoot budding period <5℃ accumulated temperature, the average maximum temperature of the young fruit period, the daily precipitation of the fruit expansion peak period ≥37℃ effective accumulated temperature and the number of days of daily precipitation of ≥0.1mm in the maturity period; Step 5. When the tea oil tree is actually growing, the key climate influencing factors of the corresponding growth period of the tea oil tree are input into the disaster loss prediction model of the corresponding yield influence period to predict the probability of disaster losses in the tea oil tree during the yield influence period, thereby evaluating the disaster loss situation; when the tea oil tree fruits are ripe, the climate influencing factors of the entire growth period of the tea oil tree are input into the total disaster loss prediction model of the tea oil tree to predict the probability and loss rate of disaster losses in that year, thereby evaluating the disaster loss situation.
2. The method for predicting oil-tea damage according to claim 1, characterized in that: The disaster loss prediction models for each major yield-influencing period and the total disaster loss prediction model for oil-tea were both established using the XGBoost algorithm.
3. The method for predicting oil-tea damage according to claim 1, characterized in that: The main yield impact period data set and the total prediction data set both use 80% of the data as a training set and 20% of the data as a test set.
Citation Information
Patent Citations
Method for establishing camellia climate quality grades on basis of fuzzy evaluation method
CN108596448A
Tea frost disaster loss dynamic evaluation method
CN112907383A
Dynamic evaluation method for tea leaf loss caused by drought disasters in autumn and winter
CN113487127A
