A Bayesian framework is developed to quantify predictive uncertainty in environmental modeling caused by uncertainty in modeling scenarios, model structures, model parameters, and data. This framework can be used to consider multiple models and to evaluate them for model selection and/or model averaging. Based on the variance decomposition, new Sobol’global sensitivity indices are defined for multiple models and multiple scenarios so that biased identification of important parameters can be avoided, given that parameter importance may vary substantially between alternative model structures and/or alternative modeling scenarios. An example of using the framework to quantify model uncertainty is presented to simulate soil microbial respiration pulses in response to episodic rainfall pulses (the “Birch effect”). The Bayesian framework is mathematically general and can be applied to a wide range of environmental problems.