Magnesium (Mg) and its alloys are studied increasingly as temporarily implant materials due to their non-toxicity, biocompatibility, and biodegradability properties. The main challenge is to tailor the degradation rates of these implant to be suitable for the human body, which needs a large number of experiments. However, reliable computational models can enhance the prediction process. In our group, we developed and calibrated an in vitro degradation of pure Mg under physiological conditions considering uncertainty aspects [1]. Surrogate modelling methodologies based on uncertainty quantification (UQ) reduce the complexity and the high computational costs of degradation models. We evaluated the validity and accuracy of three surrogate models in this study, specifically polynomial chaos expansion, Kriging, and polynomial chaos Kriging (PCK). Surrogate models are compared with the optimized model's response and in-house experimental data [1]. The three surrogate models are found to capture the behavior of the optimized model and they agree with the experimental data, as shown in Fig.1a. Nevertheless, further UQ analyses such as sensitivity analyses highlight potential deviations when different surrogates' models are used, as presented in fig. 1b. Consequently, it is important to select the surrogate models appropriately based on the problem.