The National Model System (Landelijk Model Systeem [LMS]) is used to model and predict travel behaviour for the whole of the Netherlands. The LMS is a multimodal model that makes predictions for the main road and rail network of the Netherlands and is an important tool for policy making. One of the inputs of the LMS is data based on yearly travel surveys (Onderzoek Verplaatsingen in Nederland [OViN]).

The work in this thesis shows that the LMS is currently unable to accurately capture the effect of the differences in spatial environment effectively. The LMS uses the Degree of Urbanisation (DU) to model the spatial environment, which proves insufficient to correctly model travel behaviour.

This conclusion was reached by using a cluster analysis and propensity score matching and by comparing the ‘real’ modal split (OViN) with the ‘predicted’ modal split (LMS). All zones in the Netherlands were clustered based on characteristics of the spatial environment. This was done using the so-called D-variables.

The propensity score matching showed that the differences in modal split between the different DUs and the different clusters are caused by both demographic characteristics and differences in the spatial environment. However, the effect of the spatial environment is larger.

Policy makers and other users of the LMS should be aware that testing policies or future scenarios in the LMS that change aspects of the spatial environment might introduce additional uncertainties in the forecasts of some regions.