Flue-cured tobacco waterlogging disaster grade evaluation method based on meteorological indexes

Abstract

The invention discloses a flue-cured tobacco waterlogging disaster grade evaluation method based on meteorological indexes. According to the method, core indexes are calculated through meteorological data, soil humidity and water accumulation day number related items needing field observation are replaced, the meteorological data are easily obtained from an existing meteorological station or automatic monitoring equipment, the threshold of data collection is greatly reduced, and waterlogging disaster assessment can be popularized and applied to more flue-cured tobacco planting areas. Meanwhile, through a threshold value optimization step, an accurate corresponding relation is established between the meteorological indexes and the core parameters required by the national standard, the authority of national standard evaluation is reserved, the evaluation process is more simplified and efficient, and the convenience, timeliness and space coverage capability of flue-cured tobacco waterlogging disaster monitoring and early warning are greatly improved. The method does not need to deploy a soil moisture sensor and manually inspect accumulated water in each tobacco field, can achieve large-range and automatic disaster assessment only by means of a weather station, is beneficial to business popularization, and has definite economic and social values for guiding agricultural production and reducing disaster loss.

Description

Technical Field

[0001] This invention belongs to the field of flue-cured tobacco production technology, specifically a method for assessing the level of waterlogging disasters in flue-cured tobacco based on meteorological indicators. Background Technology

[0002] Tobacco waterlogging disaster assessment is a professional undertaking that systematically investigates and comprehensively evaluates the extent of damage caused by excessive soil moisture and root hypoxia during the tobacco planting process due to continuous rainfall or poor drainage. Its core basis includes the symptoms of damage to tobacco plants, such as leaf wilting, yellowing, lower leaf rot, and black spots at the stem base, as well as environmental indicators such as the duration of waterlogging in the field, soil water holding capacity, and meteorological precipitation data. By establishing control fields and affected fields for yield comparison and quality testing, and combining satellite remote sensing monitoring and UAV low-altitude remote sensing technology, the assessment team can accurately quantify the impact of waterlogging disasters on tobacco yield, chemical composition compatibility, and industrial availability. The assessment results provide a scientific basis for developing post-disaster remedial measures, optimizing drainage engineering design, conducting agricultural insurance loss assessment, and guiding disaster prevention and mitigation in tobacco-growing areas. However, traditional assessment of waterlogging disasters in flue-cured tobacco relies on field observation data such as soil moisture and number of days with water accumulation in the field. This type of data collection is costly and difficult to obtain continuously in real time in large-scale planting areas, resulting in limited assessment coverage. At the same time, the process of manually calculating and judging according to national standards is cumbersome and time-consuming, which is not only inefficient but also prone to errors due to human operation. It cannot quickly respond to the assessment needs when disasters occur and is difficult to flexibly adapt to the different situations in different planting areas. Summary of the Invention

[0003] The purpose of this invention is to provide a method for assessing the level of waterlogging disasters in flue-cured tobacco based on meteorological indicators in order to solve the problems mentioned above.

[0004] The technical solution adopted in this invention is as follows: a method for assessing the level of waterlogging disaster in flue-cured tobacco based on meteorological indicators, characterized in that the method includes the following steps:

[0005] A method for assessing the severity of waterlogging disasters in flue-cured tobacco based on meteorological indicators, characterized by the following steps:

[0006] S1: Collect daily meteorological data and field observation data of flue-cured tobacco planting areas; calculate the daily natural precipitation deficit rate W using Formula 1. 20 This result will serve as the basis for S2 threshold optimization; at the same time, the national standard for cumulative precipitation R over 5 consecutive days will be retained. s5 This provides core indicators for S4 level determination;

[0007] S2-A: W obtained based on S1 20 Test candidate threshold W0, calculate W in the sample 20 >W0 consecutive days D W ;Statistical D WCompared with field observation d 20 The correlation coefficient is used to select W0, which corresponds to the maximum value, as the final threshold.

[0008] S3-A: Use Formula 2 to determine the effective precipitation standard; test candidate P. th Calculate R in the sample s1 >P th consecutive days T w ;Statistical T w With the actual site w The correlation coefficient is selected, and the P value corresponding to the maximum value is chosen. th As the final threshold, based on the intervals in Table 1, new statistical indicators are used to derive the corresponding waterlogging disaster level;

[0009] The corresponding waterlogging disaster level also includes:

[0010] S2-B: Combined with measured R s5 d 20 d w Refer to Table 1 to determine the disaster level Y; this label will be used as the training output of the S5 model;

[0011] S3-B: Calculate the new indicator characteristics X:R s5 Similar to the national standard, the daily natural precipitation deficit rate W 20 Effective precipitation duration days T w Value: Depends on the daily precipitation threshold P th T w Rainfall R on that day s1 >P th The duration of the disaster is determined by X and Y as targets, and an algorithm is selected to train the model. After training, meteorological data is input, and the model directly outputs the disaster level, achieving automated assessment.

[0012] In a preferred embodiment, in step S1, the natural precipitation deficit rate W 20 The calculation formula is as follows:

[0013] (1)

[0014] In the formula: R s1 Daily precipitation, potential evapotranspiration of flue-cured tobacco Calculated using the FAO Penman-Monteith method.

[0015] In a preferred embodiment, step S1 requires collecting daily meteorological data from the tobacco-growing area for at least five consecutive years, covering factors such as daily precipitation, average temperature, maximum temperature, minimum temperature, sunshine duration, average wind speed, and relative humidity. Simultaneously, field observation data from the same period is also collected, including relative humidity at a depth of 20 cm, days of field waterlogging, and disaster level records. After data collection, the data is first cleaned to remove missing and outlier values. Then, the core indicator is calculated: the Penman-Monteith method is used, combining factors such as temperature, sunshine duration, wind speed, and humidity to obtain the potential evapotranspiration E. T0 ; through daily precipitation and E T0 The difference divided by E T0 The daily natural precipitation deficit rate W was obtained. 20 In addition, the cumulative daily rainfall over the past 5 days has resulted in a cumulative rainfall of R over 5 consecutive days. s5 This indicator will be used for subsequent level determination.

[0016] In a preferred embodiment, in step S2, the daily natural precipitation deficit rate W obtained in S1 is... 20 For each candidate threshold, count the number of consecutive days in all samples where the daily natural precipitation deficit rate is greater than that threshold, denoted as D. W ; then D W Pearson correlation analysis was performed between the value and the number of consecutive days with the relative humidity exceeding the standard at a depth of 20 cm in the soil observed in the field, and the correlation coefficient between the two was calculated. Finally, the candidate threshold with the largest correlation coefficient was selected as the final W0.

[0017] In a preferred embodiment, in step S3, formula 2 is:

[0018]

[0019] In the formula:

[0020] This is the possible evaporation rate.

[0021] Net radiation,

[0022] G represents soil heat flux, which is equivalent to small soil heat flux in the reference grassland over a timescale of one to ten days.

[0023] This is the constant of the wet / dry meter;

[0024] T mean The average daily temperature;

[0025] The wind speed at a height of 2 meters;

[0026] For saturated water vapor pressure, based on Tmax and T min calculate;

[0027] This is the actual water vapor pressure;

[0028] This represents the slope of the saturated vapor pressure-temperature curve. From this, we can see that W... 20 >0 indicates excessive rainfall and is positively correlated with waterlogging.

[0029] In a preferred embodiment, for the effective precipitation duration threshold P th The candidate range was set to 0.1 mm to 10 cm, and multiple candidate values ​​were determined at 0.1 mm intervals. For each candidate threshold, the number of consecutive days with daily precipitation greater than that threshold in all samples was counted, denoted as T. w ; T w Pearson correlation analysis was performed between the value and the number of days of field waterlogging observed in the field, and the correlation coefficient was calculated; the candidate threshold with the largest correlation coefficient was selected as the final P-value. th This threshold will be used for subsequent level determination and feature extraction for machine learning model training.

[0030] In a preferred embodiment, the cumulative precipitation R obtained from S1 over 5 consecutive days is combined with... s5 The number of consecutive days with excessive relative humidity at a depth of 20 cm observed in the field. 20 and the number of days of water accumulation in the fields (d) w The disaster level is determined according to the criteria in Table 1. Specific criteria are based on R... s5 The numerical range of d 20 The duration of days, d w The combination of the duration of the disaster, the number of days, and the disaster level is used to determine the disaster level for each sample; this level is used as the truth label Y, and compared with the R value of S1. s5 W 20 and S3's T w Together they form the input-output pair of a machine learning model, providing the data foundation for model training.

[0031] In a preferred embodiment, a random forest algorithm is selected for model training, with 100 trees, a maximum depth of 10 layers, and a minimum number of splits of 5. The dataset is divided into training and testing sets in a 7:3 ratio, with S1's R... s5 W 20 and S3's T w As input features X, the true label Y of S4 is the output target.

[0032] In a preferred embodiment, after training the model using the training set, performance is evaluated using the test set, with accuracy, precision, and recall as evaluation metrics. After model training is complete, W is first calculated using the S1 method on the input meteorological data. 20 With R s5 Then, based on S3's P th Calculate T w After inputting the model, the disaster level is directly output, enabling automated assessment.

[0033] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0034] 1. In this invention, core indicators are calculated using meteorological data, replacing soil moisture and waterlogging days that require on-site observation. Meteorological data is readily available from existing weather stations or automatic monitoring equipment, significantly lowering the data collection threshold and enabling the assessment of waterlogging disasters to be widely applied in more flue-cured tobacco growing areas. Simultaneously, through threshold optimization, a precise correspondence is established between meteorological indicators and the core parameters required by national standards. This maintains the authority of national standard assessments while simplifying and increasing efficiency, greatly improving the convenience, timeliness, and spatial coverage of flue-cured tobacco waterlogging disaster monitoring and early warning. There is no need to deploy soil moisture sensors in every tobacco field or conduct manual inspections for waterlogging; large-scale, automated disaster assessment can be achieved solely using weather stations (or gridded meteorological data), facilitating operational promotion and possessing clear economic and social value in guiding agricultural production and reducing disaster losses.

[0035] 2. In this invention, the introduction of a machine learning model transforms the assessment process from manual calculation and judgment to intelligent output. Only meteorological data needs to be input, and the model can quickly provide the disaster level, significantly improving assessment efficiency and making it suitable for rapid response during disasters. Furthermore, the model training uses national standard levels as truth labels, combined with a closed-loop adjustment mechanism, to continuously optimize thresholds and model parameters, ensuring that the assessment results are highly consistent with national standards. In addition, the combination of statistical threshold optimization and machine learning retains the interpretability of threshold methods while leveraging the advantages of machine learning in handling complex relationships. This allows the assessment method to meet the transparency requirements of operational applications while also addressing the nonlinear correlation between meteorological indicators and disaster levels, improving the accuracy and adaptability of the assessment. Attached Figure Description

[0036] Figure 1 This is a schematic diagram illustrating the principle of the threshold optimization method of the present invention;

[0037] Figure 2 This is a schematic diagram illustrating the principle of the machine learning method of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0039] Reference Figure 1-2 A method for assessing the severity of waterlogging disasters in flue-cured tobacco based on meteorological indicators, the method includes the following steps:

[0040] S1: Data Preparation and Calculation of Key Indicators

[0041] Collect daily meteorological data (daily precipitation R) in flue-cured tobacco growing areas s1 (Including temperature, sunshine, wind speed, humidity, etc.) and field observation data (relative humidity of 20cm soil, number of days of waterlogging in the field). w (Disaster level records); calculate the daily natural precipitation deficit rate W using Formula 1. 20 This result will serve as the basis for S2 threshold optimization; at the same time, the national standard for cumulative precipitation R over 5 consecutive days will be retained. s5 (Σ R in the last 5 days) s1 This provides core indicators for S4 level determination.

[0042] S2-A: Optimize the natural precipitation deficit rate threshold W0:

[0043] W obtained based on S1 20 Test candidate thresholds W0 (e.g., 0, 0.2, 0.5, etc.), and calculate W in the sample. 20 >W0 consecutive days D w ;Statistical D W Compared with field observation d 20 The correlation coefficient is used to select W0, which corresponds to the maximum value, as the final threshold.

[0044] S3-A: Optimize the threshold P for the duration of effective precipitation. th :

[0045] Use Formula 2 to determine the effective precipitation standard (assuming Formula 2: daily precipitation R). s1 >P th (Time is considered as effective precipitation); test candidate P th (e.g., 0.1mm, 0.2mm, etc.), calculate R in the sample. s1 >P th consecutive days T w ;Statistical T w With the actual site w The correlation coefficient is selected, and the P value corresponding to the maximum value is chosen. th This result will be used as the final threshold for determining the grade in S4 and for training the model in S5.

[0046] S4: Generate ground truth labels Y (to provide training targets for machine learning)

[0047] R combined with field observations s5 d 20 d w Refer to the table below to determine the disaster level Y (light / moderate / severe); this label will be used as the training output of the S5 model.

[0048] Table 1 Classification Standards for Waterlogging Disasters in Flue-cured Tobacco

[0049]

[0050] The corresponding waterlogging disaster level also includes:

[0051] S2-B: Combined with measured R s5 d 20 d w Determine the disaster level Y; this label will be used as the training output of the S5 model;

[0052] S3-B: Calculate the new indicator characteristics X:R s5 Similar to the national standard, the daily natural precipitation deficit rate W 20 Effective precipitation duration days T w Value: Depends on the daily precipitation threshold P th T w Rainfall R on that day s1 >P th The duration of the disaster is determined by X and Y as targets, and an algorithm is selected to train the model. After training, meteorological data is input, and the model directly outputs the disaster level, achieving automated assessment.

[0053] Formulas 1 and 2 are common calculation methods for natural precipitation deficit rate and effective precipitation, and the steps form a closed loop through data transfer and feedback.

[0054] In step S1, the natural precipitation deficit rate W 20 The calculation formula is as follows:

[0055] (1);

[0056] In the formula: R s1 Daily precipitation, potential evapotranspiration of flue-cured tobacco Calculated using the FAO Penman-Monteith method.

[0057] In step S1, daily meteorological data for at least five consecutive years from the tobacco-growing area needs to be collected, covering factors such as daily precipitation, average temperature, maximum temperature, minimum temperature, sunshine duration, average wind speed, and relative humidity. Simultaneously, field observation data from the same period should be collected, including relative humidity at a depth of 20 cm, number of days with field waterlogging, and disaster level records. After data collection, the data is first cleaned to remove missing and outlier values. Then, the core indicator is calculated: the Penman-Monteith method is used, combining factors such as temperature, sunshine duration, wind speed, and humidity to obtain the potential evapotranspiration E. T0 ; through daily precipitation and E T0 The difference divided by E T0 The daily natural precipitation deficit rate W was obtained. 20 In addition, the cumulative daily rainfall over the past 5 days has resulted in a cumulative rainfall of R over 5 consecutive days. s5 This indicator will be used for subsequent level determination.

[0058] Based on the daily natural precipitation deficit rate W obtained from S1 20 For each candidate threshold, count the number of consecutive days in all samples where the daily natural precipitation deficit rate is greater than that threshold, denoted as D. w ; then D w Pearson correlation analysis was performed between the value and the number of consecutive days with the relative humidity exceeding the standard at a depth of 20 cm in the soil observed in the field, and the correlation coefficient between the two was calculated. Finally, the candidate threshold with the largest correlation coefficient was selected as the final W0.

[0059] Formula 2 is:

[0060]

[0061] In the formula:

[0062] The amount of potential evapotranspiration is expressed in mm / d.

[0063] Net radiation (MJ / (m²·d))

[0064] G represents soil heat flux. On a timescale of one to ten days, the soil heat flux of the reference grassland is relatively small. Therefore, G is approximately 0 in this paper, with units of (MJ / m² / d).

[0065] This is the wet / dry constant, expressed in kPa / ℃.

[0066] T mean The average daily temperature is expressed in degrees Celsius (°C).

[0067] The wind speed at a height of 2 meters is expressed in m / s.

[0068] For saturated water vapor pressure, based on T max and T min Calculations are performed in kPa.

[0069] This is the actual water vapor pressure, in kPa.

[0070] is the slope of the saturated vapor pressure-temperature curve, in units of (kPa / ℃). From this, we can see that W 20 >0 indicates excessive rainfall and is positively correlated with waterlogging.

[0071] For the effective precipitation duration threshold P th The candidate range was set to 0.1 mm to 10 cm, and multiple candidate values ​​were determined at 0.1 mm intervals. For each candidate threshold, the number of consecutive days with daily precipitation greater than that threshold in all samples was counted, denoted as T. w ; T w Pearson correlation analysis was performed between the value and the number of days of field waterlogging observed in the field, and the correlation coefficient was calculated; the candidate threshold with the largest correlation coefficient was selected as the final P-value. th This threshold will be used for subsequent level determination and feature extraction for machine learning model training.

[0072] Combined with the cumulative precipitation R over 5 consecutive days obtained from S1 s5 The number of consecutive days with excessive relative humidity at a depth of 20 cm observed in the field. 20 and the number of days of water accumulation in the fields (d) w The disaster level is determined according to the criteria in Table 1. Specific criteria are based on R... s5 The numerical range of d 20 The duration of days, d w The combination of the duration of the disaster, the number of days, and the other three factors determines the disaster level (mild, moderate, or severe) for each sample; this level is used as the truth label Y, and compared with the R value of S1. s5 W 20 and S3's T w Together they form the input-output pair of a machine learning model, providing the data foundation for model training.

[0073] The Random Forest algorithm was chosen for model training, with 100 trees, a maximum depth of 10 layers, and a minimum number of splits of 5. The dataset was divided into training and test sets in a 7:3 ratio, using S1's R... s5 W 20 and S3's T w As input features X, the true label Y of S4 is the output target.

[0074] After training the model using the training set, performance is evaluated using the test set, with accuracy, precision, and recall as evaluation metrics. After model training is complete, W is first calculated using the S1 method on the input meteorological data. 20 With R s5 Then, based on S3's P th Calculate T w After inputting the model, the disaster level is directly output, enabling automated assessment.

[0075] From the above, we can conclude that:

[0076] In this invention, core indicators are calculated using meteorological data, replacing soil moisture and waterlogging days that require on-site observation. Meteorological data is readily available from existing weather stations or automatic monitoring equipment, significantly lowering the data collection threshold and enabling the assessment of waterlogging disasters to be widely applied in more flue-cured tobacco growing areas. Simultaneously, through threshold optimization, a precise correspondence is established between meteorological indicators and the core parameters required by national standards. This maintains the authority of national standard assessments while simplifying and increasing efficiency, greatly improving the convenience, timeliness, and spatial coverage of flue-cured tobacco waterlogging disaster monitoring and early warning. Without the need to deploy soil moisture sensors in every tobacco field or conduct manual inspections for waterlogging, large-scale, automated disaster assessment can be achieved solely using weather station (or grid point meteorological data), facilitating operational promotion and possessing clear economic and social value in guiding agricultural production and reducing disaster losses.

[0077] In this invention, the introduction of a machine learning model transforms the assessment process from manual calculation and judgment to intelligent output. By simply inputting meteorological data, the model can quickly determine the disaster level, significantly improving assessment efficiency and making it suitable for rapid response during disasters. Furthermore, the model training uses national standard levels as truth labels, combined with a closed-loop adjustment mechanism, to continuously optimize thresholds and model parameters, ensuring a high degree of consistency between the assessment results and national standards. In addition, the combination of statistical threshold optimization and machine learning retains the interpretability of threshold methods while leveraging the advantages of machine learning in handling complex relationships. This allows the assessment method to meet the transparency requirements of operational applications while also addressing the non-linear correlation between meteorological indicators and disaster levels, thus improving the accuracy and adaptability of the assessment.

[0078] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0079] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the severity of waterlogging disasters in flue-cured tobacco based on meteorological indicators, characterized by: The method includes the following steps: S1: Collect daily meteorological data and field observation data of flue-cured tobacco planting areas; calculate the daily natural precipitation deficit rate W using Formula 1. 20 This result will serve as the basis for S2 threshold optimization; at the same time, the national standard for cumulative precipitation R over 5 consecutive days will be retained. s5 This provides core indicators for S4 level determination; S2-A: W obtained based on S1 20 Test candidate threshold W0, calculate W in the sample 20 >W0 consecutive days D W ;Statistical D W Compared with field observation d 20 The correlation coefficient is used to select W0 corresponding to the maximum value as the final threshold. 20 The number of consecutive days with a relative humidity of more than 90% in the 20cm layer of soil, as defined by the national standard. S3-A: Use Formula 2 to determine the effective precipitation standard; test candidate P. th Calculate R in the sample s1 >P th consecutive days T w ;Statistical T w With the actual site w The correlation coefficient is selected, and the P value corresponding to the maximum value is chosen. th As the final threshold, the new statistical indicators are used to determine the corresponding level of waterlogging disaster. The corresponding waterlogging disaster level also includes: S2-B: Combined with measured R s5 d 20 d w Determine the disaster level Y; this label will be used as the training output of the S5 model; S3-B: Calculate the new indicator characteristics X:R s5 Similar to the national standard, the daily natural precipitation deficit rate W 20 Effective precipitation duration days T w Value: Depends on the daily precipitation threshold P th T w Rainfall R on that day s1 >P th The duration of the disaster is determined by X and Y as targets, and an algorithm is selected to train the model. After training, meteorological data is input, and the model directly outputs the disaster level, achieving automated assessment.

2. The method for assessing the level of waterlogging disaster in flue-cured tobacco based on meteorological indicators as described in claim 1, characterized in that: In step S1, the natural precipitation deficit rate W 20 The calculation formula is as follows: （1）; In the formula: R s1 Daily precipitation, potential evapotranspiration of flue-cured tobacco Calculated using the FAO Penman-Monteith method.

3. The method for assessing the level of waterlogging disaster in flue-cured tobacco based on meteorological indicators as described in claim 1, characterized in that: In step S1, daily meteorological data for at least five consecutive years from the tobacco-growing area needs to be collected, covering daily precipitation, average temperature, maximum temperature, minimum temperature, sunshine duration, average wind speed, and relative humidity. Simultaneously, field observation data from the same period should be collected, including relative humidity at a depth of 20 cm, number of days with waterlogging in the field, and disaster level records. After data collection, the data is first cleaned to remove missing and outlier values. Then, the core indicator is calculated: the Penman-Monteith method is used, combining temperature, sunshine duration, wind speed, and humidity factors to obtain the potential evapotranspiration E. T0 ; through daily precipitation and E T0 The difference divided by E T0 The daily natural precipitation deficit rate W was obtained. 20 In addition, the cumulative daily rainfall over the past 5 days has resulted in a cumulative rainfall of R over 5 consecutive days. s5 This indicator will be used for subsequent level determination.

4. The method for assessing the level of waterlogging disaster in flue-cured tobacco based on meteorological indicators as described in claim 1, characterized in that: In step S2-A, the daily natural precipitation deficit rate W obtained in S1 is used as the basis. 20 For each candidate threshold, count the number of consecutive days in all samples where the daily natural precipitation deficit rate is greater than that threshold, denoted as D. W ; then D W Pearson correlation analysis was performed between the value and the number of consecutive days with the relative humidity exceeding the standard at a depth of 20 cm in the soil observed in the field, and the correlation coefficient between the two was calculated. Finally, the candidate threshold with the largest correlation coefficient was selected as the final W0.

5. The method for assessing the level of waterlogging disaster in flue-cured tobacco based on meteorological indicators as described in claim 1, characterized in that: Formula 2 is: In the formula: This is the possible evaporation rate. Net radiation, G represents soil heat flux, which is equivalent to small soil heat flux in the reference grassland over a timescale of one to ten days. This is the constant of the wet / dry meter; T mean The average daily temperature; The wind speed at a height of 2 meters; For saturated water vapor pressure, based on T max and T min calculate; This is the actual water vapor pressure; The slope of the saturated vapor pressure-temperature curve is expressed in kPa / ℃.

6. The method for assessing the level of waterlogging disaster in flue-cured tobacco based on meteorological indicators as described in claim 1, characterized in that: In step S3-A, the effective precipitation duration threshold P is... th The candidate range was set to 0.1 mm to 10 cm, and multiple candidate values ​​were determined at 0.1 mm intervals. For each candidate threshold, the number of consecutive days with daily precipitation greater than that threshold in all samples was counted and denoted as T. w ; T w Pearson correlation analysis was performed between the value and the number of days of field waterlogging observed in the field, and the correlation coefficient was calculated; the candidate threshold with the largest correlation coefficient was selected as the final P-value. th This threshold will be used for subsequent level determination and feature extraction for machine learning model training.

7. The method for assessing the level of waterlogging disaster in flue-cured tobacco based on meteorological indicators as described in claim 1, characterized in that: In step S2-B, the cumulative precipitation R over 5 consecutive days obtained in S1 is combined with... s5 The number of consecutive days with excessive relative humidity at a depth of 20 cm observed in the field. 20 The number of days of waterlogging in the fields (dw) is used to classify the disaster level according to the judgment rules; the specific basis is R. s5 The numerical range of d 20 The combination of the duration of the disaster, the duration of the disaster in dw, and the disaster level of each sample is determined; this level is used as the truth label Y, and compared with the R value of S1. s5 W 20 and S3's T w Together they form the input-output pair of a machine learning model, providing the data foundation for model training.

8. The method for assessing the level of waterlogging disaster in flue-cured tobacco based on meteorological indicators as described in claim 1, characterized in that: In step S3-B, the random forest algorithm is selected for model training, with 100 trees, a maximum depth of 10 layers, and a minimum number of sample splits of 5. The dataset is divided into a training set and a test set in a 7:3 ratio, using the R... s5 W 20 and S3's T w As input features X, the true label Y of S4 is the output target.

9. The method for assessing the level of waterlogging disaster in flue-cured tobacco based on meteorological indicators as described in claim 1, characterized in that: In step S5, after training the model using the training set, performance is evaluated using the test set, with accuracy, precision, and recall as evaluation metrics. After model training is complete, W is first calculated using method S1 on the input meteorological data. 20 With R s5 Then, based on S3's P th Calculate T w The disaster level is directly output after the model is input.

Images