USING PARALLEL MARKOV CHAIN MONTE CARLO TO QUANTIFY UNCERTAINTIES IN GEOTHERMAL RESERVOIR CALIBRATION

We introduce a parallel rejection scheme to give a simple but reliable way to parallelize the Metropolis-Hastings algorithm. This method can be particularly useful when the target density is computationally expensive to evaluate and the acceptance rate of the Metropolis-Hastings is low. We apply the resulting method to quantify uncertainties of inverse problems, in which we aim to calibrate a challenging nonlinear geothermal reservoir model...