||For visual interpretation, mapping or empirical modelling purposes, the amount of information contained in a full spatio-temporal description of the groundwater table dynamics is simply too large. For such purposes, the data has to be compressed without loosing too much information. Methods have been developed to visualise the groundwater regime in overall graphs, or statistically characterise the dynamics with a limited set of parameters. More recently, methods have been sought to identify the properties that determine the dynamics of a groundwater system. In such approaches, it is believed that the spatial differences in the groundwater dynamics are determined by the system properties, while its temporal variation is driven by the dynamics of the input into the system. In this paper, a method is presented that links the dynamics of the input to the spatially variable system properties, and results in a new set of parameters that characterise the groundwater dynamics (GD). While the dynamics of the input are characterised by its mean level and annual amplitude, the functioning of the groundwater system is characterised by its impulse response (IR) function. The IR function can for instance be estimated empirically using a time series model. Subsequently, the input and system characteristics are combined into a set of parameters that describe the output, or GD, using simple analytical expressions. It is shown that these so-called GD characteristics (the mean depth, convexity, annual amplitude and phase shift), can describe the GD in detail (for as far as the time series model can). In the example application, the GD characteristics are compared to other methods for characterising the groundwater regime, using two example series of groundwater level observations. It is shown that the so-called MxGL statistics (Mean Highest, Lowest or Spring Groundwater Level) that are often used have some important drawbacks, as they filter out the low-frequency dynamics of a system and mix-up annual with higher frequencies. Consequently, it is concluded that the capability of MxGL statistics in characterising the GD at different locations is less than that of GD characteristics.