Windbreaks are important measures for reducing wind erosion in arid and semi-arid areas. A series of experiments were conducted in a wind tunnel to assess the effectiveness of different vegetation patterns (uniform, random, and cluster) of simulated Caragana korshinskii stands on wind speed. The uniform pattern provided a better shelter effect and was optimal at a density of 50%. The protected area and the wind reduction ratio increased and the downwind minimum wind speed decreased with an increase in the number of rows and a reduction in the belt's spacing. The locations of minimum wind speed (X_min) were similar in arrangements with single-, two- and three-row belts. At the leeward distance close to vegetation (≤5 H, where H is the height of the plants), the efficiencies of vegetation patterns with high densities were greater than those of multiple-row belts and similar to those of multiple-belt shelterbelts; but at the leeward distance >5 H, the relationship was reversed. The single-row belt was the least effective, and the double-belt shelterbelt with belt spacing of 4 H was the most effective pattern. Multiple-belt shelterbelts had a lower downwind wind speed and a longer sheltering length than the other patterns, and so it is recommended windbreaks designed to reduce wind speed and control aeolian erosion in arid and semi-arid areas.