Sustainable management and protection of spring waters must be based on reliable predictions of flow and transport phenomena. Process-based models are frequently applied to accomplish this task. The provision of adequate model parameters is a prerequisite for predictive modelling. Frequently, this involves the use of inverse modelling, i.e. model parameters are adjusted until model results match field observations. Thus, process-based models are employed not only for prediction (predictive models) but also to support the aquifer characterisation (interpretative models). This work provides an overview of modelling techniques that can be applied to spring catchments for the aquifer characterisation or predictive purposes. The selection of an appropriate model must account for the type of aquifer: Continuum models describing flow through porous media are based on effective macroscopic properties of a representative elementary volume. Yet spring catchments are often composed of fractured or karstified hard rocks. The length scale of the discontinuities in these rocks may be so large that continuum approaches are found to be inadequate. Therefore, other distributed parameter models, such as discrete fracture networks or hybrid models that combine continuum and discrete approaches have been developed. All these models have in common that they require spatial parameter distributions. From a practical point of view, the application of distributed parameter models poses great challenges to the hydrogeological characterisation of spring catchments, in particular, in alpine terrain. These practical issues suggest the use of parsimonious global modelling approaches (lumped parameter models), in which the spring catchment is represented by only few global parameters rather than by spatial parameter distributions. Some of these approaches are empirical, i.e. they do not account for the physics of flow. Others are based on solving the flow equations for certain scenarios. Recently, process-based global approaches have emerged that incorporate the spatial heterogeneity of the spring catchment in the model equation. It is suggested that this type of approach is more appropriate than empirical models if predictions are required in a changing environment.