It is important to study how secondary species responds to topography-induced environmental changes and to identify dominant topographic factors that affect the spatial distribution of secondary species. The basic information is essential for prioritizing ecological restoration area. By using the GRASP utilities of SPLUS2000, we developed the GAM model for each secondary species of Caijiachuan watershed on the basis of the topographic variables that was calculated by using both spatial analysis packages of ArcGIS9.2 and the TauDEM. The topographic indexes included elevation, aspect, slope, plan curvature, slope position index (SPI), wetness index (TWI), and specific catchment area (SCA).The vegetation information for thirteen selected secondary species was extracted from Quickbird imageries. To establish the GAM models, we used 1 to represent the presence of each specific species, and 0 to represent the absence. Generally, a total of 3927 sample points recording both topographic information and the corresponding vegetation information were employed to build the GAM models of the specific species. The results show that most of the established models performed well, with the AUC values being higher than 0.7. Moreover, the models were stable as the AUC values were similar between the validation and the cross-validation test. In spite of the small difference in the form of GAM models between the herbaceous species and shrub species, both elevation and aspect were influential in controlling spatial distributions of the secondary species.The spatial distribution of the secondary species generally follows the spatial pattern of the water availability, showing a higher probability of the occurrence of specific species on the shady slopes and lower elevation area. Due to the effect of large-scale topographic variables such as the elevation, both specific catchment area (SCA) and wetness index (TWI) that reflects micro-scale had less influence on the spatial distribution of the species. Slope was considered as the least importance. We conclude that vegetation restoration in the Loess area should follow the spatial distributions of secondary species and take into account the spatial variability of water availability. According to the potential of the topographic variables in affecting the spatial distribution of vegetation, both elevation and aspect should be considered as the principle factors in identifying the target area for vegetation construction and in site suitability assessment. Both specific catchment area (SCA) and wetness index (TWI) should be considered as the secondary factors, whereas slope should be considered as the last factor.