On snow hazard mapping over the Chinese mainland by using observational and reanalysis datasets

Abstract

Assessment of snow hazard, more specifically the hazard due to snow load, is inherently complex, partly due to insufficient direct measurements of ground snow load and sparse distribution of meteorological stations across regions of interest. In the present study, the extreme ground snow load across the Chinese mainland was assessed using up-to-date surface observational data (SOD) (mostly ground snow depth) and six reanalysis datasets spanning 1979 to 2023. A comparative analysis of the statistics of the annual maximum ground snow load, L_A, and the T-year return period value of L_A, l_T, was conducted across all datasets. The results revealed that the spatial trends of L_A derived from two reanalysis datasets, JRA_3Q and GLDAS_2, closely align with those from SOD. Among the Gumbel, lognormal, and generalized extreme value distributions, the lognormal distribution was preferable for approximately two-thirds of the 1684 stations analyzed when using SOD, JRA_3Q, and GLDAS_2. Furthermore, L_A for 90.8% and 80.1% of the stations was stationary when using SOD and JRA_3Q, respectively, while L_A for 70.2% of the stations increased slightly with GLDAS_2. In general, l_T values obtained from reanalysis datasets differ substantially from those derived using SOD for stations in the Tibetan Plateau and Xinjiang region. However, l_T values for other regions are similar when using SOD, JRA_3Q, and GLDAS_2. Most notably, since the mapped l₅₀ values derived from SOD_SD, JRA_3Q, and GLDAS_2 differ from those specified in the current structural design code, the code-recommended l₅₀ values should be carefully scrutinized and potentially updated.