JOINT INVERSION FOR THERMAL CONDUCTIVITY AND BASAL HEAT FLUX

Abstract

In geothermal modelling, thermal data collected at the earth’s surface and at well locations can help in the characterization of the subsurface. In the steady-state conductive heat flow scenario, there are two primary parameters which dictate the temperature and the heat flow distribution: the thermal conductivity within the domain of interest and the boundary conditions placed on the domain.
In the geothermal community, the problem of estimating the thermal conductivity given temperature and/or heat flow measurements is often dealt with in a deterministic framework for efficiency. Furthermore, boundary conditions are often taken to be known precisely or are calibrated manually.
Instead, we propose carrying out simultaneous inversion for both the thermal conductivity and the basal heat flux in the statistical, in particular Bayesian, framework. We show that this can be done for essentially the same computational cost as carrying out inversions in the deterministic framework. The added benefits to using the Bayesian framework are twofold: prior uncertainty in the parameters can be systematically incorporated into inversions, and the resulting uncertainty in the estimated parameters can be quantified approximately, at the cost of one matrix inversion.
We also compare the local sensitivity of temperature measurements to changes in the conductivity and the basal heat flux.