Quantitative modelling of spatial variability of piezocone data from Venice lagoon silty soils

As geotechnical research and design codes rely increasingly on probabilistic approaches, site characterization should also be conducted in the light of the explicit quantification of the uncertainty and spatial variability in soil properties. Several techniques are available to model spatial variability quantitively for characterization properties. Geostatistical interpolation methods such as kriging provide a cost-efficient means to obtain estimates of soil properties at unsampled locations and to parameterize the associated uncertainty. The characterization of spatial variability through the calculation of empirical semivariograms and the subsequent fitting of theoretical semivariogram models is a fundamental preliminary step to geostatistical interpolation. Such step provides important information regarding the spatial variability of data from in-situ testing data and from derived geotechnical parameters. This paper provides a practical case-study application of spatial variability analysis of piezocone data obtained at a test site in the Lagoon surrounding the historic city of Venice, in North-Eastern Italy, where an ambitious research project was carried out and enabled to have unusually dense and regularly spaced CPT test data. Empirical semivariograms are calculated for cone resistance, sleeve friction, and porewater pressure, along with the soil behavior classification index obtained from these measurements at a set of reference measurement depths. Piezocone data are pre-processed to ensure adequate preliminary interpretation of vertically contiguous measurement in the light of the volumetric phenomena involved in cone penetration. A number of theoretical semivariogram models are fitted comparatively, and best-fit models are selected based on objective criteria and subjective judgment. Characteristic semivariogram parameters, providing information on correlation distance and small-scale variability, are retrieved. Modeling options are explained and results are analyzed and assessed critically.