A full-Bayesian approach to the inverse problem for steady-state groundwater flow and heat transport

The full (hierarchal) Bayesian approach proposed by Woodbury & Ulrych and Jiang et al. is extended to the inverse problem for 2-D steady-state groundwater flow and heat transport. A stochastic conceptual framework for the heat flow and groundwater flow is adopted. A perturbation of both the groundwater flow and the advection-conduction heat transport equations leads to a linear formulation between heads, temperature and logarithm transmissivity [denoted as ln (T)]. A Bayesian updating procedure similar to that of Woodbury & Ulrych can then be performed. This new algorithm is examined against a generic example through simulations. The prior mean, variance and integral scales of ln (T) (hyperparameters) are treated as random variables and their pdfs are determined from maximum entropy considerations. It is also assumed that the statistical properties of the noise in the hydraulic head and temperature measurements are also uncertain. Uncertainties in all pertinent hyperparameters are removed by marginalization. It is found that the use of temperature measurements is showed to further improve the ln (T) estimates for the test case in comparison to the updated ln (T) field conditioned on ln (T) and head data; the addition of temperature data without hydraulic head data to the update also aids refinement of the ln (T) field compared to simply interpolating ln (T) data alone these results suggest that temperature measurements are a promising data source for site characterization for heterogeneous aquifer, which can be accomplished through the full-Bayesian methodology.