1: \begin{abstract}

2: In this work, we propose a framework that combines the approximation-theory-based

3: multifidelity method and Gaussian-process-regression-based multifidelity

4: method to achieve data-model convergence when stochastic simulation models and

5: sparse accurate observation data are available. Specifically, the two types of 

6: multifidelity methods we use are the \emph{bifidelity} and \emph{CoKriging} 

7: methods. The new approach uses the bifidelity method to efficiently

8: estimate the empirical mean and covariance of the stochastic simulation outputs,

9: then it uses these statistics to construct a Gaussian process (GP) representing 

10: low-fidelity in CoKriging. We also combine the bifidelity method with Kriging,

11: where the approximated empirical statistics are used to construct the GP as 

12: well. We prove that the resulting posterior mean by the new physics-informed 

13: approach preserves linear physical constraints up to an error bound. 

14: By using this method, we can obtain an accurate construction of a state of

15: interest based on a partially correct physical model and a few accurate

16: observations. We present numerical examples to demonstrate performance of the method. 

17: 

18: \noindent \textbf{Keywords}: physics-informed, Gaussian process regression,

19: CoKriging, multifidelity, bifidelity, error bound.

20: \end{abstract}

21: