A Semiautomatic Method for History Matching Using Sequential Monte Carlo

The aim of the history matching method is to locate nonimplausible regions parameter space complex deterministic or stochastic models by model outputs with data. It does this via a series waves where at each wave an emulator fitted small number training samples. An implausibility measure defined which takes into account closeness simulated and observed as well uncertainty. As progress, becomes more accurate so that samples are concentrated on promising poorer parts rejected confidence. While has proved be useful, existing implementations not fully automated, some ad hoc choices made during process, involves user intervention time consuming. This occurs especially when region it difficult sample uniformly generate new points. In article we develop sequential Monte Carlo (SMC) algorithm for implementing semiautomated. Our novel SMC approach reveals yields can multimodal, highly irregular, very uniformly. offers much reliable sampling space, requires additional computation compared other approaches used in literature.