Understanding Soil Spatial Variability and Uncertainty: The Role of Borehole Spatial Distribution in Single Pile Design

This study presents a Bayesian framework to assess how borehole spacing and soil variability influence single-pile design. Deterministic approaches, such as the Laboratoire Centrale des Ponts et Chausées (LCPC) method, estimate pile resistance from Cone Penetration Test data but typically ignore spatial heterogeneity. Data from the Bolivian Experimental Site for Testing, including piezocone profiles and axial load tests for bored piles, were used to define soil-type priors for cone resistance. Bayesian updating and Metropolis-Hastings algorithm yielded posteriors for end bearing (Qp), side friction (Qf), and ultimate capacity (Qult). Compared with LCPC predictions, the Bayesian estimates were more conservative for Qp and Qult, while Qf estimates were similar to the observed. Spatial effects were assessed using interpolated profiles; borehole spacing beyond 35 m increased prediction error and Root Mean Squared Errors. Results improved the reliability of single-pile capacity, particularly with sparse site data, and provided a robust framework for uncertainty incorporation.