Probabilistic Modeling of Structural Wind Loads and Reliability Implications Considering Wind Direction Effects in Northeast China

Wind speed and wind direction are key random variables governing structural wind-load effects and reliability levels. Conventional probabilistic wind-load analyses usually derive return wind speeds from nondirectional marginal wind-speed distributions, which may obscure directional extreme-wind characteristics and their influence on structural safety reserves. This study develops a wind speed–direction joint probabilistic framework for Northeast China using annual maximum wind speeds and their corresponding wind directions from 98 meteorological stations in Liaoning, Jilin, and Heilongjiang Provinces. Candidate marginal wind-speed distributions, including Gumbel, Lognormal, generalized extreme value, two-parameter Weibull, three-parameter Weibull, and Frechet distributions, are evaluated using information criteria. Considering the circular nature of wind direction, the 16-sector recording scheme, and multimodality, von Mises mixture distributions are adopted to model marginal wind-direction probabilities. An extended Copula library is then introduced to identify station-specific dependence structures between wind speed and wind direction. Based on the established joint model, conditional directional return wind speeds are calculated for representative stations, design wind-speed biases are quantified relative to nondirectional return wind speeds, and regional statistics are conducted for 59 stations with valid directional return estimates. A simplified reliability analysis is further performed to examine the transmission of directional wind-speed differences to structural reliability indices. The results show significant station-to-station variability in both wind-speed and wind-direction distributions. Wind directions generally exhibit multimodal concentration, and wind speed–direction dependence varies spatially, indicating that a universal independence assumption is inappropriate. For the valid stations, the mean relative directional range (RDT ) is 18.58–21.63%, and the mean maximum positive bias is 10.36–11.66%, suggesting that nondirectional design wind speeds may underestimate conditional extreme wind speeds in adverse directions. The proposed framework provides a probabilistic basis for directional design wind-speed assessment and reliability-oriented wind-load evaluation in regions with pronounced wind-direction effects.