Modeling Seasonal and Weather Impacts on Cycling Count

Abstract

Cycling has been proven to contribute to not only cyclists' heath but also a sustainable transportation system in urban cities. Policy makers and urban planners all over the world have been promoting bicycling. City of Seattle has been implementing new policies and programs to create a bike friendly environment and aims to quadruple ridership by 2030. Therefore, empirically confirming such growth in Seattle will help to justify current and future investment in bicycle infrastructure and programs in Seattle. This study uses Seattle cyclist count data to quantify cycling trend and examine their relationship between seasonal and weather factors. First, a systematic approach is taken to identify the explanatory variables and their appropriate forms of transformation. Then different models are investigated and compared to best capture the relationship between bike counts and factors. Specifically, non-linearity, discontinuity and interaction items are taken into account. Results are interpreted with intuitive visualization using counterfactual simulations. Furthermore, a predictive model is proposed to estimate daily count in the future. Autoregressive Integrated Moving Average (ARIMA) model is used to account for autocorrelation. Its predictive performance is evaluated using cross validation. Finally, proposed methodology is applied to multiple locations in Seattle and identify their unique bicycle travel patterns. This research will help policy makers and transportation planners to better understand the factors that could drive the bike demand and influence bike travel behavior.