This study investigates the effects of wind direction on the airflow and pollutant dispersion inside a long street canyon using computational fluid dynamics (CFD). A 3D CFD model for predicting the flow and dispersion in a canyon is first developed using the FLUENT code and then validated against wind tunnel experiments. Then, the airflow and traffic pollutant dispersion in an isolated canyon with a street-length-to-building-height ratio of 10 are numerically simulated for seven wind directions (α = 0°, 15°, 30°, 45°, 60°, 75° and 90°, where α is the angle between the approaching flow and street axis). The results demonstrate that the mean () and turbulent (ACHʹ) air exchange rates (ACHs) for the canyon are close when α = 0°, 15°, 30°, 45° and 60°, whereas the magnitude of ACHʹ is significantly greater than that of when α = 75° and 90°. Additionally, the ACH reaches its maximum and minimum values when α = 30° and 90°, respectively. The computed velocity and concentration fields clearly reveal the variation in the in-canyon flow pattern and pollutant distributions on the canyon walls and footpaths due to the wind direction. Evaluating the maximum, minimum and average concentrations on the canyon walls and footpaths for each of the seven wind directions, we determine: On the leeward-oriented wall, the wall-averaged concentration increases greatly with α, and the values of the wall-maximum are highest when α = 75° and lowest when α = 0°. By contrast, on the windward-oriented wall, both the wall-averaged and wall-maximum concentrations are highest when α = 0°. Finally, at the human respiration height, the highest concentration on the footpath next to the leeward-oriented wall occurs when α = 75°, whereas the highest concentration on the footpath close to the windward-oriented wall is observed when α = 0°.