The purpose of this study is to classify drainage basins that have similar landforms into the same clus-ters. These similarities are obtained from topographic measurements based on digital elevation mod-els (DEMs). To exclude the effects due to geological differences from topographic characteristics, the geological structure of the study field should be homogeneous and should not include any Quater-nary volcanoes. Therefore, the Abukuma Mountains in northeast Japan were selected as the study field. As the basic dataset, 250-m grid-interval DEMs provided by the Geographical Survey Institute of Japan were used. To prevent scale factor effects, the areas of study basins were determined to be ap-proximately 32 km² (512 grid-points) to 64 km² (1, 024 grid-points), and 54 basins met this condition. Then five hypsometric and six planimetric characteristics of topography were measured at all basins. Hypsometric characteristics are drainage average dissection height (H_d), drainage average undissec-tion height (H_u), drainage maximum relief (H_mr), drainage average relief (H_ar), and average relative elevation (H_re). Planimetric topographic characteristics are drainage frequency (F_D), bifurcation ratio between first- and second-order streams (R_b2¹), bifurcation ratio between second- and third-order streams (Rb₃²), kurtosis of path length frequency distribution (g₂), basin form ratio (R_bf), and drain-age form ratio (R^df) F_D, R_b2¹, and R_b3² indicate the development of stream networks. R_df is a ratio of the drainage area versus the square of the longest diameter of the drainage basin and R_df is that, of the drainage area versus the square of the longest path of the drainage network, respectively.<br> In order to classify the study basins, two kinds of variables were applied to cluster analyses (Ward method). As the first method, all drainage basins were classified into four clusters using the obtained topographical characteristics as variables. As the second method, the composite properties of basin landforms were examined by a principal component analysis (PCA), resulting in three major components. All drainage basins were classified into three clusters by cluster analysis in which these major components were used as variables. The standard deviations of the topographical character-istics of each cluster were small in the second method and thus the second method was adopted in the <br>following discussion.<br> Basins belonging to cluster 1 showed a large first PC and a large third PC. Factor loadings of first PC were large on vertical topographic characteristics (lid, H_d, H_u, H_mr, and H_re ). The basins of cluster 1 are located in the center of the Abukuma Mountains. These basins include isolated peaks standing near the center ridge. The steep landforms of cluster 1 are caused by these isolated peaks. Thus the first PC explains the steepness of landf orms. The factor loadings of the third PC were large at the drain-age frequency (F_D) and bifurcation ratio between first- and second-order streams (R_b2¹). These two cha-racteristics indicate the development of stream networks. Thus the third PC explains the develop-ment of stream networks. The drainage basins of cluster 1 are characterized as the steepest landform in the Abukuma Mountains and have well-developed stream networks.<br> The basins of cluster 2 were large in the second PC. The second PC is determined by kurtosis of path length frequency distribution (g₂), basin form ratio (R_bf), and drainage form ratio (R_df).