Abstract:

Based on the temperature and precipitation data from the late rice region of Hubei, spanning the period from 1970 to 2022, this paper delves into the characteristics of chilling dew wind in Hubei, encompassing its climate characteristics, atmospheric circulation pattern and sea surface temperature background. A simulation model for the beginning date of chilling dew wind was constructed using the Hyperparameters Tuning of Random Forest method and previous circulation indices, in order to provide a reference for preventing and mitigating the impact of the chilling dew wind. The results showed that: (1) over the past 53 years, the chilling dew wind had shown a trend of delayed beginning date and reduced total number of days, with a delay rate of nearly 0.4d·10y⁻¹ and a reduction rate of nearly 44d·10y⁻¹. The decade with highest cumulative number of stations was 1970−1980, the lowest in 2001−2010. Since 1990s, the proportion of severe station occurrence had been increasing, indicating a shift towards fewer but more intense occurrences of the chilling dew wind. (2) The typical circulation pattern associated with chilling dew wind in Hubei was characterized by a "− + −" anomalous distribution of geopotential height from west to east over the mid-to-high latitudes of Eurasia, accompanied by the northward transport of warm and humid air from the southern regions. (3) The beginning date of chilling dew wind was significantly correlated with the large−scale meridional circulation anomaly in Eurasia and the anomalous sea surface temperature (SST) in the western Pacific warm pool and the north of the South Atlantic. The beginning date of chilling dew wind was earlier (later) when the geopothetic height near the Ural mountains increases (decreases). When SST over western Pacific warm pool was abnormally cold (warm), and the north of the South Atlantic SST showed cold to warm (warm to cold) distribution from north to south, the beginning date of chilling dew wind tend to be earlier (later). (4)The simulation error using hyperparameters tuning of random forest method was minimal. The historical fitting rate from 1970 to 2007 was 91%, and average absolute error of sample test from 2008 to 2020 was 2.9d, indicating that the model had a good ability to simulate the beginning date of chilling dew wind in Hubei. 

Key words: Chilling dew wind, Climatic characteristics, Circulation analysis, Random forest