In this paper the problem of modelling partial mixing phenomena is considered. The ultimate modelling goal is to find identifiable, finite-dimensional state-space models, which are physically interpretable, realisable and which describe partial mixing. Hence, realization theory will be linked to prior physical systems knowledge to answer the question which mixing models are good candidates in environmental/reactor systems modelling. It appears however that only a limited set of low-dimensional structures is identifiable. From the real world example it appears that an appropriate, physically interpretable and realisable model within this class of models cannot be easily found.