Steady-state Reynolds-averaged Navier–Stokes (RANS) simulations of urban wind fields and linear interpolation between computed wind directions, are validated against experimental wind sensor measurements. In comparison with linear interpolation, the effectiveness of Proper Orthogonal Decomposition (POD)-based reduced-order modeling is also evaluated with the objective of reducing data storage requirements while maintaining interpolation accuracy. The study focuses on a complex urban area of approximately 3 km 2 surrounding the campus of the Technical University of Berlin. A total of 36 wind field simulations are conducted for inflow wind directions at 10°intervals. Validation against long-term experimental wind measurements at nine locations within the urban area yields average deviations of 21.7% for relative wind speed, 0.58 m/s for absolute wind speed, and 28°for wind direction. A standard linear interpolation approach is subsequently applied to estimate wind fields for various wind directions using an interpolation database of 12 simulated wind fields. The validation of the interpolated wind fields against experimental measurements results in average deviations of 29.7% for relative wind speed, 0.56 m/s for absolute wind speed, and 31°for wind direction. It is further demonstrated that increasing the number of wind fields in the interpolation database beyond 12 does not lead to an improvement in estimation accuracy. Finally, POD-based interpolation is introduced as a reduced-order modeling approach aimed at substantially reducing data storage requirements. The accuracy of the wind field estimates obtained using POD-based interpolation is evaluated in comparison with standard linear interpolation. Using six POD modes, an estimation error of 6.3% is obtained, compared to 12% for linear interpolation based on six wind directions, this demonstrates an improved accuracy for an equivalent level of data storage using POD-based interpolation.
Ebert et al. (Wed,) studied this question.