The calibration of trip lengths is an important challenge for multi-regional MFD-based applications, as it can influence the dynamics of regional densities (and speeds). Existing research has not paid significant attention in the topic, giving opportunities for answering some fundamental questions. In this paper, we propose an original methodology to explicitly determine trip length distributions based on a subset of trips in the city network and its partitioning. Since the full description of all-realized trips is-difficult-to-estimate for a large city network, we propose to define a set of virtual trips corresponding to a full coverage of potential local origin-destination pairs and shortest-paths in distance. We investigate how different levels of information can influence the accuracy of the multi-regional dynamic MFD-based models, through the estimated trip length distributions. This information ranges from regional trip lengths without any information of origin, destination, previous or next regions up to the specific regional path associated to each trip. We test this methodology on a simulated environment with a significant amount of real information for a network with 757 links that corresponds to the 6th district of Lyon. We first provide guidance on how to properly define the set of virtual trips. We show that a single average trip length is not representative of all possible trip lengths estimated by the most detailed levels of information for one region. We also highlight that the first level of information assigns similar trip lengths for regional paths that are composed by the same regions, but in a reverse ordered sequence. Nevertheless, this is not the case for the most detailed levels of information since they capture the directed links city network topology. We propose a procedure to update the trip lengths when an update of the regional Origin Destination matrix changes, without a need for a full recalculation. We also investigate the influence of the trip length tuning on the traffic dynamics of the regional network. The traffic states are modeled by a trip-based MFD model. We show that the setting of the trip lengths influences the traffic dynamics in the regions. Moreover, the traffic conditions predicted by the two most detailed levels of information are close. (C) 2019 The Authors. Published by Elsevier Ltd.