Effects of Weather on Traffic Collisions in Edmonton, Canada

Abstract

Three years of hourly traffic collision data and associated weather data for the city of Edmonton, Alberta, Canada, were analyzed. Using a matched-pair analysis to compare periods of adverse weather with similar periods without adverse weather, the effect of weather on traffic collisions in Edmonton was assessed. Adverse weather took the form of strong winds (>35 km h−1), precipitation (>0.10 mm h−1), or limited visibility (<1.0 km). Risk ratio (RR) values were calculated to determine the relative increase in accidents resulting from each type of weather, as well as for multiple types of concurrent weather. Analysis was performed for collisions that resulted in property damage only (PDO) and severe collisions that resulted in injuries. To determine whether specific types of collisions were more likely to occur during adverse weather, RR values were determined for multiple types of collisions, such as following too closely or failing to stop at a traffic signal. Our results show that adverse weather affects traffic collisions in a variety of ways. Horizontal visibility less than 200 m and strong winds above 45 km h−1, while relatively infrequent in Edmonton, typically resulted in an increase in overall collisions (RR = 3.40 and 1.17, respectively). Precipitation, which is far more common to Edmonton, posed an increased risk to drivers. The type of precipitation was also clearly a factor. The RR value for PDO collisions was higher during periods of snowfall (1.90) than during periods of rainfall (1.28), but the RR value for collisions resulting in injuries (1.37) was higher than the value for PDO collisions during periods of rainfall. Collisions during snowfall had a higher RR value (1.90) for PDO collisions than for collisions resulting in injuries (1.55), and both were higher than the RR values for either type of collision during rainfall, suggesting that snowfall is a greater threat to driving safety than rainfall. For both precipitation types, a higher rate of precipitation was associated with increased risk and a larger number of collisions. Some collision types were clearly affected by the type of precipitation. Vehicles left the road resulting in property damage at far higher RR values during snowfall (3.36) than during periods of rainfall (1.09). Furthermore, analysis of collision types identify that two collisions with nearly identical actions, stopping at a stop sign and stopping at a traffic signal, have very different risks during adverse weather. Stopping at stop signs appears to be far more difficult than stopping at traffic signals during precipitation events and has a higher associated risk. Lastly, we identify a puzzling concern: our results show that Edmonton drivers do not perceive periods of rainfall as presenting a hazard, resulting in a disproportionately high number of injuries during rainfall.