Soils in a 4.5 ha Napa valley vineyard were investigated to determine the smallest number of sampling sites required to obtain an accurate large scale map of soil variability. Natural soil variability and soil profile changes resulting from cultivation have an important impact on crop ripening and yield. The study site was on a complex convexo-concave slope cutting across a thick series of estuarine river alluvium with contrasted granulometry. Soil types included Haploxeralfs, Haploxererts and Xerofluvents, depending on their position on the slope, parent material, and recent cultivation practices. A preliminary survey included 44 auger holes regularly spaced on the slope and divided into 6 toposequences. Subsequently, 12 soil pits, forming 3 toposequences were described and sampled, based on the preliminary observations. Each sampling point was precisely georeferenced using a GPS and a digital elevation model was constructed using the elevation data. The soil samples were analyzed for texture, CEC, organic matter content and exchangeable cations. This information was combined in a geodatabase with field observations such as coarse fragment content, soil color, horizon thickness, and analyzed using a geostatistical package. The lateral continuity of the horizons, the presence of textural gradients and abrupt thresholds were studied. The optimal sampling density to conserve the greatest amount of information on soil characteristics was examined by varying the number of observations included in subsamples. The study site showed great variability, primarily due to parent material heterogeneity, that kept the required number of sampling points higher than expected. Cultivation somewhat reduced variability in the upper part of the profile. This study demonstrates the feasibility of assessing soil variability in vineyards using traditional methods of soil analysis combined with GIS and geostatistical approaches.