Conventional soils surveys (CSS), generally based on the soil-landscape paradigm, are often expensive and time-consuming for updating available soil information at the precision required for modern environmental modeling. Soil-landscape segmentation using digital elevation models (DEM) and digital terrain analysis software have been proposed as a new approach to achieve this task. The objective of this study was to compare the stratification efficiency of the LandMapR software (LandMapper Environmental Solutions Inc.) and a 1995 CSS map used to delineate homogeneous soil-landscape units (SLU) in the Appalachian area (Beauce County, Québec, Canada). A 1:50 000 DEM and its derived topographic attributes were used as inputs in the segmentation process. The Beauce County was subdivided into six physiographic units using a 1:1 000 000 soil-landscape map of Canada. The most representative unit among these, the Gayhurst catena, which consists of five fine-loamy soil series developed on glacial till overlaying calcareous sandstones, siltstones and slates, was analysed in this study. Stream-burning and means filters were applied to improve the plausibility of the DEM. The LandMapR software approach to segmentation is a three-step process: 1) generation of flow topology and extraction of a stream network, 2) derivation of various topographic attributes, 3) and classification of landform units using a fuzzy logic algorithm. Twelve landform units were initially delineated and later were generalized into four major soil-landscape positions. A variance reduction analysis of two meaningful topographical attributes, i.e. slope and wetness index, was used to compare the stratification efficiency of both the LandMapR segmentation results and the CSS method. LandMapR-derived units were found to have an advantage over the CSS method since they respectively allow a variance reduction of 50% and 35% for slope and 37% and 18% for wetness index. An approach is proposed to describe the soil series composition of each landform unit.