In the Major Land Resource Area (MLRA) 107 pilot project, a multi-phase probability sampling design for updating soil surveys was implemented in western Iowa. In general, multi-phase designs are used when a variable of interest is expensive to measure, but is strongly related to another (auxiliary) variable which is inexpensive to observe. In a multi-phase design, the auxiliary variable is observed for a sample and the study variable is observed for a relatively small sub-sample. In the estimation stage, the auxiliary information is used to improve estimators of distributional quantities relating to the study variable. In particular, we consider estimation of quantiles in this context;Chambers and Dunstan (1986) (CD) presented an estimator for a finite population distribution function which incorporates auxiliary information. A linear relationship between the study variable and the auxiliary information is assumed. The residuals in the linear model are assumed to be homoskedastic. We derive a Bahadur-like representation for the quantile estimator corresponding to the CD distribution function estimator. This expression is used to derive an expression for the asymptotic variance of the quantile estimator;We consider estimation of quantiles for soil texture profiles using data from the MLRA 107 pilot project. The laboratory determination of soil texture is the variable of interest. Auxiliary information is available in the form of field determinations of soil texture. Due to the multi-phase sampling design used for data collection, field determinations are available at more sites than laboratory determinations. The CD quantile estimator is modified to incorporate sampling weights and to allow heteroskedasticity in the assumed linear model;A Bayesian approach to this estimation problem is also considered. A hierarchical model is used to describe the relationships between observed data and unknown parameters. Soil horizon profiles are modeled as realizations of Markov chains. Transformed textures are modeled with Gaussian mixtures. The posterior distribution of soil texture profiles is numerically approximated using a Gibbs sampler. The hierarchical model provides a comprehensive framework which may be useful for analyzing other variables collected in the pilot project. The two approaches are compared using simulated and real data.