Abstract
In Sapporo, Japan, the effective number of lanes in winter was decreased by some snow removal operations (plowing) and increased by others (hauling); the average speed of vehicles in February 2013 (winter) was about 10 km/h slower than in October 2013 (autumn). In other words, the roads are more prone to traffic congestion in winter than in the other seasons. The objective of this study was to investigate the effects of weather conditions and snow removal operations on travel speed to determine the causes of winter traffic congestion in Sapporo. Four steps were performed to identify the relationships between traffic, weather, and snow removal operation factors. The first step was to establish a data set for analysis by combining traffic, weather, and snow removal operation data. The second step was to develop multiple linear regression (MLR) models with all the variables. The third step was to investigate the periodicity of the residuals to apply autoregressive integrated moving average (ARIMA) models between the actual and predicted values of the MLR models. Then, ARIMA models or seasonal ARIMA (SARIMA) models were developed, depending on the periodicity. The fourth step was to combine the R-squares of the MLR and ARIMA models. It was found that the explanatory power of the speed prediction model was improved by combining the two models. As a result of applying the ARIMA and SARIMA models at the residuals, the R-squared values increased by .60.
Get full access to this article
View all access options for this article.