This chapter proposed a new method for landform classification of the Loess Plateau based on grid DEM data and the GIS method. The slope spectrum, as well as corresponding indices, was first extracted from 5 m resolution DEMs. Then, 13 slope spectrum index layers were integrated into a comprehensive layer, on which were based a series of unsupervised classifications applied to landform classification. Experimental results show that the slope spectrum is appropriate for describing the spatial distribution of the loess landform and the automatic classification of loess landform types is realizable through an integrated analysis of the slope spectrum. Furthermore, the method pays more attention to the spatial distribution of the terrain at a global level and is more fitted to loess landform classification at the regional scale. It is also of great significance in extending the application of DEMs and it improves the methodology of digital terrain analysis.