A Comparison of Alternative Approaches to Constraining Parking Demand to Car Park Capacity in Hierarchical Logit Choice Models: Do They Work and Which Method is Best?
P Davidson, P Clarke, A Thomas, Peter Davidson Consultancy, UK; M Shahin, University of Alexandria, EG
The cost and availability of a parking space is a key issue in travellers decision making yet many models ignore this. This paper puts forward alternate methods of incorporating parking into multi mode models, drawing conclusions on each of them.
In many urban areas, a key determinant of travel choice is the availability of a parking space at the destination - yet we seldom put this into our models. If a person has convenient free parking at his destination he will certainly not consider bus nor park and ride. Even if the bus alternative is both faster and cheaper, many drivers would still not consider using the bus. Only if there is no convenient parking, does bus start being considered by these drivers - and even then its consideration just gets it into the choice set - it is still unlikely to be chosen. Parking becomes even more important when we forecast into the future as traffic levels (and parking demand) are set to increase fast - much faster than we can build parking spaces. Our utility functions for mode, destination, time period, peak spreading, station, and several other choices, are more likely to include time and money then the availability of a parking space - yet the unavailability of a parking space is perhaps the single most important determinant of whether a person takes the car or considers alternatives.
So why do we ignore parking when it is such a fundamental determinant of travel choice? Is it that we don't know much about how to model it? Yet there are several methods which the authors have used in a whole series of study models - studies for which real forecasts were required - studies where we would be laughed out of court if we produced absurdities when it comes to matching the car arrivals in a town centre to the parking spaces available. The paper considers these methods under three main categories as follows:
A simple method of dealing with parking is by ad-hoc adjustments to match the forecast car attractions with the parking stock. Adjustment methods range from changing the trip productions (or trip attractions if doubly constrained) to making the town centre less attractive in the model - perhaps with extra parking segmentation (eg free/ pay).
A more dynamic method is to constrain the number of car trips arriving into the controlled parking zone, to the number of parking spaces, by including a zone-specific penalty perhaps representing the extended time spent searching for a parking space. The issue then is should these penalties be passed up the choice nest through the logsum or not? What is their effect on the other choices? what is their effect on the trip frequency model? Parking can be considered as a demand model to chose which space to park in and a supply model of parking spaces, to which demand is constrained. We analyse these issues.
Possibly the most complete method is to micro simulate every single car park space and every single tour. This is best done under an activity model framework for which we draw out the comparisons.
This paper explores these methods, analyses the mathematics of including them within the logit choice nest structure while maintaining the integrity of the hierarchy and compliance to the UK Department for Transport's (DfT) WebTag modelling guidance It analyses alternative utility functional forms, analyses their convergence and how to jump-start them. It then describes a practical implementation of each method in a set of real-world case studies and draws comparisons of their associated advantages and disadvantages and then puts forward conclusions about how parking should be incorporated into our models in the future.
Association for European Transport