A number of geophysical methods, such as ground-penetrating radar
(GPR), have the potential to provide valuable information on hydrological
properties in the unsaturated zone. In particular, the stochastic
inversion of such data within a coupled geophysical-hydrological
framework may allow for the effective estimation of vadose zone hydraulic
parameters and their corresponding uncertainties. A critical issue
in stochastic inversion is choosing prior parameter probability distributions
from which potential model configurations are drawn and tested against
observed data. A well chosen prior should reflect as honestly as
possible the initial state of knowledge regarding the parameters
and be neither overly specific nor too conservative. In a Bayesian
context, combining the prior with available data yields a posterior
state of knowledge about the parameters, which can then be used statistically
for predictions and risk assessment. Here we investigate the influence
of prior information regarding the van Genuchten-Mualem (VGM) parameters,
which describe vadose zone hydraulic properties, on the stochastic
inversion of crosshole GPR data collected under steady state, natural-loading
conditions. We do this using a Bayesian Markov chain Monte Carlo
(MCMC) inversion approach, considering first noninformative uniform
prior distributions and then more informative priors derived from
soil property databases. For the informative priors, we further explore
the effect of including information regarding parameter correlation.
Analysis of both synthetic and field data indicates that the geophysical
data alone contain valuable information regarding the VGM parameters.
However, significantly better results are obtained when we combine
these data with a realistic, informative prior.