Development of an Enhanced Active Modes Model for Transportation by Electric Bicycle in the Netherlands

Development of an Enhanced Active Modes Model for Transportation by Electric Bicycle in the Netherlands


Noortje Groot, Rijkswaterstaat, Frank Hofman, Rijkswaterstaat, Lucas Harms, Netherlands Institute for Transport Policy Analysis


This paper describes how the Dutch National Model System has been improved for modelling electric bicycles. The changes in model structure, level-of-service input data, scenario development and the results of the model improvements are discussed.


The Dutch National Model System (NMS) represents a strategic model that produces mid- to long-term forecasts of traffic flows on Dutch road and railway networks. The aim of this model is to assess transport policy scenarios and to evaluate national infrastructure plans.
Further, NMS is a multimodal model where a traveler’s mode choice set includes the transportation by car, either as driver or passenger, train, bus/tram/metro, and the active modes cycling and walking. While originally, the main focus of NMS has been on transportation by car and train – the modes that directly affect the use of national infrastructure – the potential substitution of short- to medium-length car trips by alternative modes like (electrical) bicycles is increasingly acknowledged. In the NMS base year 2014, a third of all trips in the Netherlands was made by bike, corresponding to about 10% of all kilometers traveled. Especially given the technical developments of electricity-supported bicycles, the need arose to properly incorporate (e-)cyclist behavior in NMS.

This paper will describe (1) how the modeling of cyclist choice behavior is enhanced, including a new model for incorporating e-bike behavior, (2) what scenarios are defined for the adoption of e-bikes in the Netherlands, and (3) what model results are obtained with respect to the active modes, including elasticities and model fit as well as forecast results. Assumptions made for the final model are also presented. The first part can be further separated into an improvement of the model input (level-of-service) and the model setup.

In particular, as a first step, the level-of-service of the active modes has been redefined by making use of a detailed network of cycle lanes and pedestrian paths, based on OpenStreetMaps. For the access and egress modes for train, stations are integrated in the network and travel times and distances have been derived accordingly. In order to differentiate between cycle speeds, height differences and urbanization are taken into account and empirical data from a national pilot to monitor bicycle trips was used as well. Subsequently, the mode-destination choice model was re-estimated with the new level-of-service data.
For the evaluation of the newly estimated model of bicycle choice behavior, a comparison of trip length distribution has been made, the model fit (loglikelihood) has been analyzed, and elasticities have been derived. By the best knowledge of the authors, no elasticities for cyclists in a multimodal framework such as NMS are published yet. The evaluation results showed that the model fit improved when adopting the new level-of-service values, and that the resulting trip-length-distribution closely follows the data of the National Travel Surveys (MON, OViN).

A second novelty described in this contribution is the distinction made between cyclists using traditional versus electrical bicycles as their main mode of transport. That is, for travelers of various age groups and travel purposes, a comparison has been made of average travel speeds and distances travelled when using either traditional or electrical bicycles. For this analysis, data of OViN was used. Given this data, the bicycle model has been calibrated in order to mix the behavior of users of traditional and electrical bicycles for a chosen share of e-bike utilization. In this manner, it is accounted for that cyclists with an e-bike not only travel faster, but also travel longer distances. This so-called ‘comfort-factor’ is explicitly modeled.

The third and last part of this paper includes the derivation of different scenarios for the possession and utilization of e-bikes in the Netherlands. Results of these scenarios are described, including the effect of e-bike penetration on all modes of transportation in NMS.
Further improvements in the cycle model could be targeted at a similar modeling of e-bike-adoption in access and egress modes for train. Another improvement would be the inclusion of separate models for traditional cyclists and e-cyclists, that is, by treating both bicycle-types as strictly separate modes of transport. For both cases the main bottleneck is the availability of reliable large-scale data distinguishing between traditional and e-cyclists, as well as between main mode and access/egress mode behavior. Finally, case studies will be performed to analyze the scope of effects when improving bicycle infrastructure by creating fast cycle routes.


Association for European Transport