The use of dynamic mathematical models to simulate the behaviour of environmental systems is common practice. However, the output of these models remains uncertain, despite their widespread use and long history of application. This uncertainty arises, amongst other factors, from errors in the data, randomness in natural processes, incorrect assumptions in the model structure with respect to the processes taking place in the natural system, and the inability of calibration procedures to unambiguously identify an optimal parameter set to represent the system under investigation. The latter two problems may be caused by the inability of the calibration procedure to retrieve sufficient information from the model residuals. In this paper, a new approach called Dynamic Identifiability Analysis is presented in order to partly overcome this limitation. A case study shows how the proposed methodology can be applied to increase the identifiability of parameters of a river solute transport model.