A numerical model of a 5 km long curved floating bridge, planned for the Bjørnafjord, in Norway, is subjected to strong wind events with stationary mean properties that vary along the bridge axis. A Weather Research and Forecasting model (WRF) is used to estimate the mean wind speeds and directions in the fjord, with a 500 m resolution, during a 20-year period. The wind turbulence intensities are estimated as functions of the position along the bridge and the wind direction, using an artificial neural network trained on nearby sonic anemometer data and formulations given in the national annex of the Eurocode. A skew wind buffeting formulation is used to estimate the linear static response and the linear quasi-steady buffeting response in the frequency domain. The response under inhomogeneous winds is compared with the response under equivalent homogeneous winds. The inhomogeneous wind response is, on average, 1.5% to 47% larger, depending on the type of analysis and response component, but a high variability is observed, with much larger differences for some wind cases. These findings motivate case-specific investigations of inhomogeneous wind effects to improve fatigue and extreme response predictions of long wind-sensitive structures.
