Study on model of single leaf weight of flue-cured tobacco in western Guizhou province based on multiple meteorological elements
[Objective]The present study aimed to explore the impact of meteorological factors on the single leaf weight of flue-cured tobacco and to construct a model for single leaf weight of flue-cured tobacco in the western region of Guizhou province based on multiple meteorologi-cal factors.[Method]Based on meteorological elements and data about single leaf weight of field flue-cured tobacco in the western tobacco planting area of Guizhou province from 2010 to 2021,four artificial intelligence and statistical algorithms,including BP neural network,ran-dom forest,linear regression and stepwise regression,were used to analyze the trend of change in single leaf weight of tobacco in the western tobacco planting area of Guizhou province in the past 12 years.Based on this,the relationship between meteorological elements and single leaf weight of tobacco was studied,and a model of multiple meteorological elements and single leaf weight of tobacco was constructed and valida-ted.[Result]In the western tobacco growing areas of Guizhou province in the past 12 years,the average single leaf weight of the lower leaves was 6.25 g,the average single leaf weight of the middle leaves was 9.95 g,and the average single leaf weight of the upper leaves was 10.81 g;The single leaf weight of the middle leaves showed little change,while the lower and upper leaves showed a trend of increasing weight.U-sing four artificial intelligence and statistical algorithms,a model for single leaf weight was established based on four meteorological elements(temperature in mature period,precipitation in fast growing period,sunshine hours in growing period and available time in growing period,hereinafter referred to as the four elements)and meteorological elements(multiple elements)from March to September during the critical pe-riods of tobacco growth.The model was simulated year by year from 2010 to 2021,and it found that although the BP neural network algorithm had the highest simulation accuracy when establishing the model,however,when simulating actual single leaf weight,the stepwise regression algorithm model based on multiple factors had the best simulation effect.Compared with other models,it could simulate the peak and valley values of single leaf weight for each leaf position year by year,followed by the random forest algorithm.[Conclusion]The stepwise regression and random forest algorithm have good simulation effects on the single leaf weight of flue-cured tobacco in the western region of Guizhou prov-ince.Therefore,when establishing a prediction model for single leaf weight of flue-cured tobacco,the meteorological factors prediction can be considered to provide scientific basis for tobacco production decision-making.
Single leaf weight of flue-cured tobaccoMeteorological factorsModel for single leaf weightMachine learning
NETL
NSTL
万方数据
