Mobility data has increasingly grown in volume over the past decade as loc- alisation technologies for capturing mobility ows have become ubiquitous. Novel analytical approaches for understanding and structuring mobility data are now required to support the back end of a new generation of space-time GIS systems. This data has become increasingly important as GIS is now an essen- tial decision support platform in many domains that use mobility data, such as eet management, accessibility analysis and urban transportation planning. This thesis applies the machine learning method of probabilistic topic mod- elling to decompose and semantically enrich mobility ow data. This process annotates mobility ows with semantic meaning by fusing them with geograph- ically referenced social media data. This thesis also explores the relationship between causality and correlation, as well as the predictability of semantic decompositions obtained during a case study using a real mobility dataset.