Abstract:

Using the daily maximum temperature during June and July from 1961 to 2020 at four national meteorological stations in Changsha tobacco district of Hunan province, the characteristics of high temperature heat damage during the maturation period of tobacco in Changsha tobacco district were analyzed by using M−K test, ROC curve and kernel density estimation methods. High temperature heat damage forecast models in Changsha tobacco district were constructed to provide early warning for decision−making service. The results showed that: the average high temperature days were 15.2 days per year, with an overall increasing trend from 1961 to 2020, averaging an increase of 1.0 days per decade. High temperature days start as early as June 1, with an average start date of June 13 at the 80% guarantee rate. On average, heat−damaged processes occured 0.9 times per year. The earliest occurrence date of high temperature heat damage was June 14, and the occurrence dates of high temperature heat damage increased significantly after July 1 (P<0.05). The average starting date for the 80% guarantee rate was July 5. The heat damage of high temperature was mainly mild, accounting for 86.3% of the entire mature period. The high temperature days during the ripening period of flue−cured tobacco in the eastern tobacco region of Changsha were 3.9 days longer than in the western tobacco region, and the average start date of the 80% guarantee rate was 7 days earlier than in the western tobacco region. The high temperature damage process occurred 0.3 times more often than in the western tobacco region, with the average start date of the 80% guarantee rate being 5 days earlier. The high temperature heat damage threshold model for Changsha tobacco district based on accumulated heat temperature had a good classification effect, with an AUC value of the ROC curve reaching 0.94. The classification threshold of model heat accumulation temperature was 4.6℃·d. The accuracy of predicting high temperature heat damage was close to 90%, and the start and end time of the prediction process, as well as the impact range, were consistent with the actual situation. The models were well−suited and could be applied to the monitoring and early warning services for flue−cured tobacco high temperature disasters, providing technical support for disaster prevention and mitigation. The probabilistic model of high temperature damage was established as an effective complement to the threshold model.

Key words: High temperature day, High temperature heat damage, Accumulated heat temperature, Forecast threshold, Flue?cured tobacco