Between september 10-15 of 2003, we focused on the grasslands in Inner Mongolia that are heavily populated with rodents, to understand the strategy adopted by the upland buzzard (Buteo hemilasius) to prey on the Mongolian gerbil (Meriones unguiculatus) in a typical steppe environment. Autumn is the season when M. unguiculatus are most active, therefore, the number of rodent holes in this season was used as a measure to determine the abundance of M. unguiculatus. Based on a previous study on M. unguiculatus community behavior, we found that when the rodents cluster in autumn, on an average, there is only 1 M. unguiculatus per 6.3 rat holes. Therefore, to reduce the interference of the temporary M. unguiculatus holes in autumn, we defined the groups that had fewer than five rodent holes as temporary, ineffective groups. This study only analyzed the groups with more than five rodent holes as a base group to assess the predation risk. The predation preference of B. hemilasius on different-sized M. unguiculatus clusters was analyzed. This study focused on 3 square meter of grassland, and the number of M. unguiculatus groups, based on the statistics, was 87. To determine the predation strategy of B. hemilasius, two researchers recorded the location and frequency of B. hemilasius within the sample area at different locations using telescopes. For 5 d B. hemilasius was observed swooping down 29 times within the sample area, which was inhabited 23 M. unguiculatus groups. The number of holes that B. hemilasius visited, and the frequency of B. hemilasius visits to each of the groups, was analyzed using the Mann-Whitney test for nonparametric data. From the results, we determined that B. hemilasius visited 2,569.5 holes, while 1,258.5 holes were not visited by B. hemilasius (U value=489.5, Adjusted Z Value=-2.37459), indicating a significant difference between the two (P=0.017574). This indicated that B. hemilasius was obviously biased towards the survey groups with more rodent holes, possibly for increased predation success rate. From the analysis that the correlation between the number of B. hemilasius visits (F) and the group size of another rodent, Brandt's vole (Lasiopodomys brandtii, (S)), was also determined as follows:F=-0.0559+0.023S (r=0.2707, P < 0.05). From these results, we concluded that the upland buzzard first chose the M. unguiculatus groups that have a higher number of rodent holes as potential prey before it chose the L. brandtii groups. During the winter, the number of M. unguiculatus individuals in clusters will become restricted, and may form an optimal number in the cluster model. The results of the present study may support the theory that prey species form an optimal cluster to reduce predation risk. M. unguiculatus may use rodent holes to escape predation, or autumn swarming behavior to reduce the density of individuals in clusters, to reduce predation risk.