1. Introduction and Scope
 We thank Beven et al. , hereafter referred to as B12, for taking the time to comment on our opinion paper [Clark et al., 2011a]. We are pleased that our paper piqued their interest, and we are pleased that B12 agree with much of what we say. We also welcome the opportunity to elaborate on our critique of the Generalized Likelihood Uncertainty Estimation (GLUE) methodology.
 The B12 comment sets the stage for some interesting discussion and debate. While the B12 comment is primarily focused on expressing their opinions about the superiority of GLUE over Bayesian approaches, the comment provides a good summary of many important challenges in hydrological sciences, including the B12 perspective on hypothesis-testing. In doing so, the B12 comment brings to the forefront the fundamental issues that wemust address as we collectively seek to improve the fidelity of our models.
 Our response below highlights the need for a carefully controlled approach to model evaluation. This was one of the central aims of our opinion paper, where we present a methodology which entails both (1) isolating the constituent hypotheses in a model (e.g., experimenting with different options for specific processes and/or scaling behavior, while keeping all other components of the model fixed); and (2) using the available data and physical insights in creative ways to scrutinize different modeling alternatives. Elements of this methodology have been applied in several recent studies using a mix of qualitative and quantitative diagnostics [e.g., Clark et al., 2011b; Kavetski et al., 2011], and using the extended GLUE framework [e.g., Krueger et al., 2010]. The controlled approach to model evaluation we advocate in our opinion paper—and in the response below—requires a combination of multiple tools and strategies to pursue the several distinct aspects of this methodology, with Bayesian methods being one of these tools.
 Our response identifies a great deal of common ground between our opinions and the sentiments expressed in B12. Although B12 may disagree with our choice of methods, including standard probability theory, statistics, and Bayesian techniques, we do share several broader aims and perspectives. In particular we agree with B12 that “We want a tool that will be useful in simulation or prediction and that reflects our qualitative perceptual knowledge of real-world processes.” We are hence confident that our exchange adds to the ongoing constructive discussion on the suitability of different model analysis strategies, and helps define tractable ways forward for those interested in improving the process of model development and evaluation.
 Our response to the B12 comment is structured as follows. First, we review the B12 summary of the major science challenges in hydrological model analysis and consider how these challenges are pursued in our multiple hypothesis methodology. Second, we question the B12 defense of GLUE, pointing out that the aspects in which GLUE actually differs from Bayesian methods do not address the real challenges of an inference framework. Rather they simply weaken its descriptive, predictive and diagnostic capabilities, including the rigor with which model hypotheses can be tested. Third, we discuss the B12 perspectives on our modeling approach, emphasizing the importance of controlled approaches to model rejection and the need for subjectivity when stronger knowledge is not available. In responding to B12, our aim is to encourage hydrologists to think more critically of the fundamental premises, assumptions and limitations of different model analysis methodologies.