A Seasonal Autoregressive Model of Vancouver Bicycle Traffic Using Weather Variables

TitleA Seasonal Autoregressive Model of Vancouver Bicycle Traffic Using Weather Variables
Publication TypeConference Paper
Year of Publication2012
AuthorsChristopher Gallop, Cindy Tse, Jinhua Zhao
Conference NameTransportation Research Board 91st Annual Meeting
PublisherTransportation Research Board
Conference LocationWashington, D.C.
KeywordsAutoregressive models, Behavior, Bicycle travel, Cyclists, Mode choice, Time Series Analysis, Urban transportation, Weather conditions

This paper uses hourly bicycle counts and weather data that are continuous and year-round to model bicycle traffic in Vancouver, Canada. The study uses seasonal autoregressive integrated moving average (ARIMA) analysis to account for complex serial correlation patterns in the error terms and tests the model against actual bicycle traffic counts. Temperature, rain, rain in the previous 3 hours and humidity are all found to be significant, with clearness found to be marginally significant at the 10% level. The combined effect of rain and its lags is close to 24% of the average hourly bicycle traffic counts, which is larger than the impact of it being a holiday or a Saturday, although the impact of it being a Sunday is still larger. An increase of one degree Celsius from the mean is generally found to increase bicycle traffic counts by 1.65%, so an increase of 10 degrees would increase bicycle traffic by 16.5%. The coefficients on humidity and clearness are small. A decrease in bicycle traffic of only 0.08% is observed per unit change in relative humidity and 0.62% at each of the four transitions between categories of cloudy to perfectly clear skies.