The authors have proposed a new type of greenhouse with a cross section similar to a wing. Such a cross section reduces the wind loads significantly compared with conventional shapes and allows a long-span structure. Simultaneous pressure measurements were made with a scale model in a simulated turbulent boundary layer Focus was on the external pressures. First, the effects ofrwind speed (Reynolds number) and surface roughness of the model on the pressure distribution are investigated; several kinds of commercial sand paper were used for roughening the surface. Then, the peak pressure coefficients for the design of cladding are discussed. Finally, the design wind load for the main wind-force resisting systems is discussed, using three kinds of evaluation methods, i.e. the time-history analysis, the LRC method and a gust effect factor approach. The results indicate that the gust effect factor approach provides reasonable load estimation for winds roughly normal to the ridgeline, while it results in a wind pressure distribution different from those estimated by the other two methods for oblique winds. The LRC method generally provides results similar to those of the time-history analysis. However, the pressure distribution depends on the load effect under consideration. The peak factor is also dependent on the load effect; it is generally larger than 3.5. This feature is related to the non-normality of wind pressure fluctuations acting on the surface.