Hail Forecasting and Key Feature Analysis in the Qinghai Plateau Using Decision Tree Algorithms 

Expand
  • 1. School of Computer Technology and ApplicationsQinghai UniversityXining 810016QinghaiChina
    2. Qinghai Provincial Laboratory of Intelligent Computing and ApplicationsQinghai UniversityXining 810016QinghaiChina
    3. School of Computer and Information ScienceQinghai Institute of TechnologyXining 810018QinghaiChina
    4. Qinghai Institute of Meteorological ScienceXining 810001QinghaiChina

Online published: 2025-04-29

Abstract

Due to its unique geographical environmentQinghai Province is highly susceptible to frequent hail events. Considering the complex topography of high-altitude regionsparticularly the Qinghai Plateauthis study constructs a hail forecasting dataset by integrating hail observations from 52 meteorological stations in Qinghai from 2009 to 2023corresponding hail disaster recordsand the ERA5 atmospheric reanalysis dataset. Based on this datasetthree ensemble decision tree models-Random ForestXGBoostand LightGBM-are employed to develop a hail forecasting modelwith separate analyses conducted on hail samples with diameters of ≥2 mm and ≥5 mm. Experimental results demonstrate that the LightGBM model consistently outperforms both Random Forest and XGBoostwith particularly superior performance in forecasting large hail eventsdiameter ≥5 mm. Specificallyfor small hail samplesdiameter ≥2 mm),the LightGBM model achieves a hit rate of 0. 923a false alarm rate of 0. 041a Critical Success IndexCSIof 0. 858an accuracy of 0. 946and a recall rate of 0. 924while for large hail samplesdiameter ≥5 mm),it attains a hit rate of 0. 938a false alarm rate of 0. 038a CSI of 0. 908an accuracy of 0. 960and a recall rate of 0. 964. Further analysis of the hail forecasting model in the complex terrain of the plateau reveals that the most influential meteorological factors for hail fore‐ casting in Qinghai Province include thermodynamic conditionsvertically integrated temperature p54. 162vertically integrated thermal energy p60. 162and 2-meter dew point temperature d2m),characteristic height layer conditions100 hPa temperature t100400 hPa temperature t400and 20 hPa geopotential height z20),and dynamic conditions500 hPa zonal wind component u500200 hPa meridional wind component v200and 200 hPa zonal wind component u200. Kernel density estimation analysis indicates that most feature variables exhibit limited separabilitysuggesting that no single factor alone can determine the occurrence of hail events. A case study demonstrates that the LightGBM-based hail forecasting model exhibits strong spatial forecasting capabilities. Analysis of the 24-hour evolution of key meteorological variables preceding a large-scale hail event at the Chaka station identifies several crucial atmospheric indicators:(1significant fluctuations in vertically integrated temperaturep54. 162),indicating intense convective activity;(2persistently high 2-meter dew point temperatured2m),reflecting abundant near-surface moisture;(3strong 500 hPa zonal wind speedu500),suggesting enhanced mid-level atmospheric dynamicsand4low 100 hPa temperaturet100),capturing upper-at‐mosphere characteristics. The coordinated evolution of these atmospheric variables not only reveals key stages in the development of severe convective weather systems but also provides a scientific foundation for improving hail potential forecasting methods in Qinghai Province.

Cite this article

Liu Jie, Zhang Guojing, Wang Xiaoying, Guan Qin . Hail Forecasting and Key Feature Analysis in the Qinghai Plateau Using Decision Tree Algorithms [J]. Plateau Meteorology, 0 : 1 . DOI: 10.7522/j.issn.1000-0534.2025.00044

Outlines

/