With the increase in computational capabilities in recent decades, the possibility of modeling complex phenomena has continued to grow, inevitably increasing the number of calibrated parameters. Determining these parameters for complex coupled models can be challenging due to inherent nonlinearities, couplings, and uncertainties. This paper aims to provide a reliable methodology for calibrating and validating large-scale transient models through a rigorous assessment of model uncertainties. The methodology consists of performing a “one variable at a time” (OVAT) sensitivity analysis to identify the main effects, followed by a variance-based (VB) sensitivity analysis to uncover the most relevant model cross-couplings over the entire parameter domain. The outcomes of the sensitivity analysis are then used to identify the most influential parameters for model calibration. The calibration process leverages deterministic gradient-based methods in conjunction with a probabilistic treatment of the data using a Bayesian approach to rigorously quantify modeling uncertainties. A pivotal aspect of the Bayesian approach lies in its incorporation of the prior parameter distribution, enabling the quantification of the impact of deviating from their respective best-known values. This results in the establishment of 95% model confidence bounds around the optimized solution, reflecting the model uncertainties.