This essay discusses the challenges in identifying and predicting hydrological models using calibration data. It argues that traditional calibration methods are incomplete and that the equifinality thesis—suggesting multiple acceptable models for environmental systems—should be more seriously considered. It proposes extending the GLUE methodology to make it more rigorous and outlines unresolved research issues. The equifinality thesis highlights that many acceptable models can fit observational data, and that model error is complex, involving input, structural, and measurement errors. It challenges the assumption that a single optimal model exists, emphasizing the need for multiple models and the importance of model evaluation. The GLUE methodology, which uses subjective likelihood measures, is criticized for not properly accounting for model error. However, it can be made more rigorous by using formal likelihood measures. The paper discusses the philosophical implications of equifinality, noting that science aims to find a single correct description of reality, but equifinality suggests multiple feasible descriptions. This is not incompatible with scientific research, as shown by Beven (2002a). The paper also addresses the problem of model evaluation and the representation of model error, emphasizing the need for a more comprehensive approach. The equifinality thesis is linked to the concept of ill-posedness in environmental models, where the information available is insufficient to determine a unique solution. This is common in hydrological models due to their complexity and the limited information from observational data. The paper discusses the implications of this for model evaluation and the need for robust methods to assess model performance. The paper also addresses the issue of model error, noting that it can arise from various sources, including input errors, model structural errors, and measurement errors. It argues that the additive error assumption is not always valid, especially in nonlinear models, and that more complex error structures need to be considered. The paper concludes that equifinality is a fundamental aspect of environmental modelling, and that model evaluation should consider multiple models and their performance. It emphasizes the need for a more comprehensive approach to model calibration and evaluation, including the use of fuzzy set theory and other methods to account for uncertainty. The paper also highlights the importance of considering input and boundary condition errors, as well as the potential for model failure, in the evaluation of hydrological models.This essay discusses the challenges in identifying and predicting hydrological models using calibration data. It argues that traditional calibration methods are incomplete and that the equifinality thesis—suggesting multiple acceptable models for environmental systems—should be more seriously considered. It proposes extending the GLUE methodology to make it more rigorous and outlines unresolved research issues. The equifinality thesis highlights that many acceptable models can fit observational data, and that model error is complex, involving input, structural, and measurement errors. It challenges the assumption that a single optimal model exists, emphasizing the need for multiple models and the importance of model evaluation. The GLUE methodology, which uses subjective likelihood measures, is criticized for not properly accounting for model error. However, it can be made more rigorous by using formal likelihood measures. The paper discusses the philosophical implications of equifinality, noting that science aims to find a single correct description of reality, but equifinality suggests multiple feasible descriptions. This is not incompatible with scientific research, as shown by Beven (2002a). The paper also addresses the problem of model evaluation and the representation of model error, emphasizing the need for a more comprehensive approach. The equifinality thesis is linked to the concept of ill-posedness in environmental models, where the information available is insufficient to determine a unique solution. This is common in hydrological models due to their complexity and the limited information from observational data. The paper discusses the implications of this for model evaluation and the need for robust methods to assess model performance. The paper also addresses the issue of model error, noting that it can arise from various sources, including input errors, model structural errors, and measurement errors. It argues that the additive error assumption is not always valid, especially in nonlinear models, and that more complex error structures need to be considered. The paper concludes that equifinality is a fundamental aspect of environmental modelling, and that model evaluation should consider multiple models and their performance. It emphasizes the need for a more comprehensive approach to model calibration and evaluation, including the use of fuzzy set theory and other methods to account for uncertainty. The paper also highlights the importance of considering input and boundary condition errors, as well as the potential for model failure, in the evaluation of hydrological models.