Variabilité des sols et stratégie d'échantillonnage dans les études pédologiques détaillées de la plaine de Montréal
In order to estimate variability of soil information presented in detailed soil surveys of the Montreal lowlands area, more than 2000 samples were collected along stratified random transects in 133 mapping units. The coefficient of variation (CV) was used as variability criteria. Variability of selected soil characteristics of the plow layer was studied according to phases of surface texture. The CV of clay, silt, pH, CEC and exchangeable cations decrease with an increase in clay content. The CV of sand content follows the opposite direction. Four modes of deposition (marine, fluvial, till and estuarine) were compared according to the variability within taxon (soil series or variant) of soil characteristics of the A, B and C horizons. More samples are needed to estimate the population mean value with the same relative error and the same confidence level for soils developed in fluvial deposits than in those formed in marine sediments. At the 90% confidence level, the allocation of 15 samples per map unit allows one to estimate the population mean with a relative error ranging from 5 to 30% for most soils and variables studied. Quality standards are proposed for soil sampling in future soil survey projects. Key words: Soil variability, accuracy, sampling, taxon, mapping unit