Combining Spatial and Temporal Dimensions in Destination Choice Models
C Cirillo, E Cornélis, L Legrain, P L Toint, FUNDP-GRT, University of Namur, BE
The context. This paper deals with the question of including travel constraints in a modelled representation of spatial choices. The basic idea is that the individual?s daily activity pattern is inserted within a space-time bounded region and that the number of possible alternatives processed by a single individual is limited (Hägerstrand, 1970). We also want to point out that work trips and non-work trips are chained into complex travel patterns. Thus activity-based framework (Cirillo and Toint, 2001) accounting for complex interactions across trip chains can better represent travel behaviour under spatial and temporal constraints. The use of advanced discrete choice models will give good estimation of the parameters of the utility function and correct prediction of choices. The data sources. The data source used in this context is Mobel, the Belgian National Mobility Survey held in 1999. Information on socio-demographic characteristics of the household and each individual in the household are available, as well as a one-day travel diary to be filled out by all members of the household. For this paper, only the travel data related to households from the Antwerpen province were retained. To this core data set, level of service data and land-use data variables have been added using a GIS. The alternative generation process. Usually the choice set is constructed by adding to the chosen alternatives a number of alternatives randomly taken from the entire set of possible alternatives. In the literature there are several approaches to drawing the subset of alternatives from the universal choice set (Ben-Akiva et al., 1984). In our spatial?temporal context, the number of possible destinations is restricted by keeping only those alternatives that are included in the household individual action-space (Dijst and Vidakovic (1997)), i.e. the area containing all activity places which are reachable, subject to a set of temporal and spatial constraints. The types of variables. A large number of explanatory variables found significant by other authors will be estimated: level of service variables (in-vehicle travel time, out-of-vehicle travel time and travel cost), level of employment (retail and service employment), and area by land use type (industrial, parking ...) (Kockelman (1996); Simma, Schlich and Axhausen (2002) and Bhat, Govindarajan and Pulugurta, (1998)). The modelling issues. The assumption that the error terms are IIA distributed is no longer satisfactory for destination choice models. The use of mixed logit allows us to take into account heterogeneity, state dependency and heteroscedasticity across individuals (Revelt and Train, 1998) and across spatial-analytic issues (Bhat and Zhao, 2002).
Association for European Transport