Modelling and analysing effects of complex seasonality and weather on an area's daily transit ridership rate: Fid-Bau Portal

Modelling and analysing effects of complex seasonality and weather on an area's daily transit ridership rate

Kashfi, Syeed Anta / Bunker, Jonathan M. / Yigitcanlar, Tan

Abstract Adverse weather is generally perceived as deterrent for public transit uses. This has also been highlighted in previous literatures. In contrary, our previous study found no association between weather and transit ridership while investigating the underlying temporal influences behind variation in daily ridership across the sub-tropical city of Brisbane, Australia. This contraindication led to the primary focus of this research. This research acknowledged that Inclusion of weather variables in conjunction with other relatively strong independent variables might result in washout of the weather effects on ridership. Variables such as rainfall do not recur on a daily basis throughout the year. Thus, generalising their effect on ridership with other independent variables that consistently influence ridership may create a similar problem. Hence, weather variables were converted into their normalised factors and combined with other independent variables while formulated the optimised the daily ridership rate estimation model. Several models were developed concerning various combinations of weather variables and through rigorous analysis it was identified that only the rain variable has noticeable effect on daily ridership. Evidently, this study functions as an update of our former study by directing towards a new approach to the analysis of the relationship between weather and transit ridership.

Highlights • A new research method for weather variables analysis • Converted weather variables into their nominal factors • Influence of weather variables were harmonised with the dominating seasonality factors. • Optimisation of previously developed daily ridership rate estimation model • Reduced ridership estimation error in the optimised model.