Migrating 4-step Models to an Activity Based Modelling Framework in Practice

Migrating 4-step Models to an Activity Based Modelling Framework in Practice


P Clarke, P Davidson, A Thomas, Peter Davidson Consultancy, UK


This paper analyses different migration sequences for migrating 4-step models to Activity Based Models drawing on our experiences of converting models for Cheshire, Ashford and others.


Our 4-step models are unable to meet the requirements of policymakers to forecast the effect of the many of the new policy measures, which they wish to implement (eg road pricing, peak spreading, parking). On the other hand there is intense interest in Activity Based Models because they are policy-responsive. However most legacy models are 4-step and to convert them to Activity Based Models is expensive on resources. This paper explores the methods which can be used, the intermediate stages in the process and the advantages and disadvantages of each approach against a background of making the best use of the existing data. Examples are drawn from recent 4-step conversions which have been undertaken by the authors.

The paper examines examples of policy responsiveness with a view to illustrating the order for converting the most needed mechanisms first. It shows how activity based models provide a suitable framework for structuring a set of travel choice models in such a way which is extensible and scaleable and which provides much more scope for market segmentation. It shows how activity based models are able to meet the forecasting requirements of policymakers.

Most of the models in Europe are 4-step models and if they are to be migrated to an activity based modelling framework then this can either be undertaken by building a new activity based model from scratch or by migrating the existing model to an activity based model framework. Most of these models are in use, so migration is often the preferred option, with a requirement to complete the migration in stages so that the model can use the enhancements introduced at each stage in the migration process.

To take full advantage of the activity based framework requires new data to be collected. However calibrating an activity based model with 4-step model data provides a better model than the 4-step model calibrated with the same data.

The paper shows how to configure distribution and mode choice as a hierarchy of choices which are modelled with a set of nested choice models which can be calibrated with the same source data as was used for the 4-step model. It shows how this can be extended to include other travel choices including time period choice, time interval choice and parking choice, interfaced with the same highway and public transport assignment models as used for the 4-step model. Complex elements such as activity duration, which is dependant upon a range of characteristics such as the makeup of the travelling party can also be encapsulated within the model. If necessary a trip frequency model can be applied at the beginning of the choice hierarchy to provide sensitivity of the model to trip suppression and induced trip making in response to improved transport provision or worsened congestion. The paper shows how the nested choice models can be dumbed-down to mimic the 4-step model if that is necessary. This could be useful so as to show consistency between the models and demonstrate the extra capabilities of the new model.

However the full power of the activity based approach is realised when two other features are implemented: people?s activities (the activity synthesiser) and micro-simulating every individual in the study area (the population synthesiser). The paper describes these processes and possible migration paths and discusses the relative merits of different migration sequences.

There are a number of significant problems that need to be overcome during the migration process. Not least of these is the validation of the assignment models. The activity based model will compose much more detail regarding the trip including departure time, preferred arrival time etc. These have implications for assignment model validation which need to be overcome. The methodologies used on recent studies are presented here along with a discussion of their relative advantages and disadvantages.

These methodologies can provide an effective migration path from 4-step models to activity based models.

The paper goes on to describe the authors experience in migrating several models from 4-step to activity based models including Cheshire Sub Regional Strategic Transport Model, Ashford Area Transport Model and other models which were successfully migrated from 4-step models to activity based models using existing data sources.


Association for European Transport