Representing Weather in Travel Behaviour Models: A Case Study from Sydney, AUS

In the transportation literature in general and the travel behaviour literature specifically, there is no consensus on the representation of weather in behavioural models, ranging from studies that use direct meteorological measurements, such a continuous variable for inches of rainfall or temperature, to indicator variables (which themselves are often subjectively derived) for qualitative weather assessments. To explore the representation of weather, this paper develops constructs of weather based upon cluster analysis of various daily weather characteristics and test the validity of these weather representations against two dimensions of travel behaviour: mode share and trip generation. Using Sydney, AUS as a case study, ten years of hourly weather data for the region are reduced using a two-stage cluster analysis technique, resulting in 17 meteorologically distinct weather types. These weather types are then tested for their correlation with one way of data from the Sydney area Household Travel Survey. Initial results comparing mode shares with weather types indicate that the weather types can be further reduced into fewer types. The inclusion of these weather types into models of trip generation show improved model performance; however, the results are weak with respect to contributing to a final valid set of weather constructs. More work needs to be done to explore the representation of weather in travel behaviour models and the overall effect of weather on travel choices.