The Relationship Between Yield And Climate Soil And Prediction Models Of Yield In Flue-cured Tobacco

Tobacco is the main source of economic sources of Tobacco farmers.Scientific and accurate predicting that the production of tobacco leaves is of great significance to optimize the tobacco industry,adjust planting plans,profit-oriented avoidance,and maintain the stable development of the industry.Climate factor is one of the main factors that affect the production of tobacco.This study uses the climatic and grilled cigarette production data from2009-2022 to analyze the relationship between climate soil and output.The random forest and BP neural network model predict the output of Qi Hou,and the combination of the two is an integrated model to achieve the prediction of the production of tobacco production.The main conclusions are as follows:(1)Climate and soil characteristics of Bijie Tobacco Area:The most climate type in Bijie tobacco growing season is normal years,followed by wet years,warm years,dry years,cold years,and warm dry years.Low.Most of Bijie is mainly for medium-granular reunion.Most soil p H is in the appropriate range of tobacco and a large number of elemental nutrients and organic matter content,which is suitable for planting tobacco.(2)Analysis of factors affecting yield:After analysis of gray correlation,the common climate factors of the association R>0.7 in Bijie are the average temperature in May,the average temperature in June,the average temperature in July,the average temperature in August,the accumulated temperature,the accumulated cumulative total Precipitation,June sunshine time and July sunshine.The correlation between the production and soil factors of the tobacco R>0.7 is based on the large to the small to the soil p H>medium-granular reunion>alkali nitrogen>full nitrogen.(3)Qualified analysis of yield:The overall trend of the output of tobacco in Bijie County is a sharp decline in 2011,and slowly fluctuating after 2012.Using the HP filtering method to separate the actual mass production of tobacco into climate output and trend output:the trend output of the trend production of golden sand and weaving gold has always shown a slow upward trend.The change is basically the same.Low-yield years have appeared in the warm-drying year and the year of dry years,and the high-yield years mostly appear in the warm and humid year type.(4)Quantitative prediction of yield:Linear regression model predicts the output of roasting tobacco.The decisive coefficient R2 of each area model is above 0.93,and the all-in-form root error RMSE is below 50kg/hm2.The trend yield simulation effect is good.The random forest(RF)and BP neural network models(BPNN)predict the production of grilled flue climate,the test set R~2is 0.95 and 0.84,respectively,the validation set R~2is 0.83 and 0.72,respectively,the prediction set RMSE is 129.32kg/hm~2and 209.26kg/hm~2,and the validation set RMSE is 216.66kg/hm~2and 358.27kg/hm~2,respectively.After the combination model,the relative error of the final linear regression+RF is small,and the fitting effect is the best.Its predictive values are related to the actual values of 0.97,which is extremely significant.The results show that the characteristic index can be extracted by the grey correlation analysis,and the tobacco yield and the actual yield are 0.28%to 2.70%,indicating that the HP filter and linear regression+RF model can predict the flue-cured tobacco yield.