In the current paradigm shift towards digitisation in almost every industrial sector, vast amounts of data are becoming available. The mobility domain is one of the key sources of spatiotemporal datasets. The potential of such datasets is far from being fully exploited so far since it is quite challenging to make sense of the complex spatiotemporal dependencies available in the data. In this paper, we propose a mining methodology tailored to spatiotemporal data. The multi-stage mining approach allows to uncover insightful spatial patterns and dependencies while taking full advantage of the temporal dimension. Initially, the time series data is segmented into appropriate time windows, which are subsequently converted into thoughtfully designed feature vectors. Characteristic temporal traffic states are derived by pooling the feature vectors across all locations and identifying clusters of homogeneous traffic behaviour. The so derived traffic states are labelled and further subjected to a semantic interpretation. Subsequently, the temporal states are migrated to the spatial dimension by using them to represent spatial trajectories of interest. In this way, the considered spatial trajectories are represented as labelled sequences (strings of traffic states) for each time segment. Those sequences are subjected to further examination by exploiting techniques from text and pattern mining domains, allowing to discover interesting spatial dependencies in time. The proposed methods are validated on a real-world ANPR dataset.