The general ecological situation of Loess Plateau is kept deteriorating, despite great efforts have been made and some achievements were gained after 50 years' endeavor to control soil erosion and renovate ecological landscape. Investigations on the past 20 years results from 11 experimental sites showed that the extension of prototypes and models derived from these sites, which were all based on one scale of small watershed, were unsatisfactory. The possible reasons for such a situation could be attributed to the too coarse scale of map, on which, the regional planning were set up and the locations of these experiment sites were decided, the too fine scale of the small watershed which were usually 3-10 km2 in size, the shortage of studies on various scales particularly on the meso-scale, and the lack of methods of scaling on up-scaling. Based on the analysis of restraints in management of small watershed, the specific concept and relevant proposed methods of resemblance-scaling and multi-scale complementarity were put forward and tested.