Fiabilité d'une étude pédologique détaillée réalisée en terrain plat

In order to estimate the reliability of soil information presented in a detailed soil survey (1:20 000 scale) conducted in a flat terrain (Vercheres county, Quebec), 35 randomly selected delineations were inspected with 10 sites per delineation. Reliability was evaluated according to parent material (marine, estuarine and fluvial) and kind of map units (simple and compound). As a whole, mineral soils have been mapped and defined reliably for most criteria commonly used in soil interpretation systems. Reliability of soil information differed significatively with parent material. Taxonomic purity of marine, fluvial and estuarine soils delineations were 70, 58 and 28%, respectively. Although less precise, compound map units were more reliable than simple map units. Among the 13 soil variables studied, A and B horizons texture, B horizon designation and depth to carbonates were the more frequently in error. Results finally indicated that reliability is also a function of the delineation area and inspection density. In order to improve reliability of soil information and accelerate soil survey in future projects, it is suggested to adjust inspection density according to parent material, to reduce the number of classes for a few variables and to avoid excessive use of soil series' variants and delineations of small size (< 4 ha).