Abstract: Typhoons frequently wreak havoc on China's eastern coast. The influence of typhoons with different cyclone structures and landing locations on sea-crossing bridges is quite different. The wind speed probability model at the bridge site is critical for studying the wind field characteristics and for evaluating the wind resistance capability of the bridge. The real-time monitoring data of wind field during typhoon periods obtained by the structural health monitoring system installed on the sea-crossing bridge provides the data basis for studying the wind field characteristics of bridge site and the analysis of a wind speed probability model. Three typical typhoons which have affected Xihoumen Bridge were investigated. Combined with the real-time monitoring data of the wind field at the bridge site, the parameters of the wind speed probability model were dynamically identified and continually updated using the Bayesian method. A time-varying model of wind speed probability distribution was developed, which can depict the wind speed probability distribution at any given time and its progression through the whole typhoon life cycle. Three typhoons' wind speed distribution and evolution were compared. As for the landing typhoon close to the bridge, the wind speed distribution at the bridge site fluctuates. When the typhoon's center is far away from the bridge, the wind speed distribution is relatively stable within the influence radius of the typhoon, and fluctuates strongly outside the influence radius.