This work focuses on a multimodal transportation hub that attracts an average of 140,000 people on a typical week day (2019). Passenger flows tend to form transportation routes towards areas of interest. However, depending on the time of day, the period of the year, or local events, these flows do not necessarily go to or transit through the same places. Several sensing systems positioned at different count locations in the transport hub allow multivariate count data to be collected. Extracting information from these large sets of highly noisy count series is challenging since these series share common dynamics in response to specific events, but may also have their own dynamics due to more localized events. Our goal in this paper is to model these mobility multivariate time series through a set of mobility patterns, to obtain a synthetic vision which helps understanding how the transport hub operates, or to use as a basis for prediction work.