Physics based or mechanistic simulation models are frequently used in current engineering design to generate, evaluate and validate designs in order to shorten the design cycle time and reduce the cost. It often involves solving a set of none linear partial differential equations, which usually requires a lot of computing time to search for a solution. It can not satisfy the need to explore more design alternatives in the early design stage. Metamodeling is an active academic field and is attracting increased attention from the industry. Surrogate models built with the metamodeling techniques provide faster analysis alternatives to the mechanistic models. Using these surrogate models in the early design stage allows engineers to explore more design alternatives with less development time and therefore has the potential to enable better engineering decision to produce quality products for reduced cost. In this paper, we review the existing metamodeling techniques including design of experiments, response surface methodologies, machine learning, and kriging. The paper also documents the results of the performance evaluation of several different metamodeling tools using a vehicle greenhouse example.