LAUSR

Background Requests for pedestrian crossings are evaluated using warrant criteria. Different properties of the location, such as vehicle and pedestrian volumes, are fed into a weighted matrix. If the location scores high enough, the location is placed on a waiting list to get the pedestrian crossing. Aim We performed a desktop statistical and simulation study to quantify the potential effects of variability in vehicle and pedestrian counts on warrant evaluations. Method Total pedestrian and total vehicle counts collected between 2013 and 2016 at controlled intersections were extracted from the City's traffic count database. The dataset was limited to standard 6 hr counts taken on weekdays, encompassing the periods between 07:00-09:00, 11:00-13:00, and 16:00-18:00. School and non-school periods were analyzed separately. Variability in vehicle and pedestrian volumes was defined using the relative percent difference (RPD) between subsequent pairs of dates at each location. The likelihood of a given vehicle or pedestrian volume meeting a warrant criterion was based on the empirical probability distributions of the RPDs extracted from the database. Results Overall variability was 9% for vehicle counts, and 20% for pedestrians. Variability in vehicle counts was not related to total volume, while variability in pedestrian counts was 10% larger on average (overall variability of 30%) when pedestrian volumes were low. Simulation results suggest that vehicle volumes of 187 and 468 were sufficiently close to the warrant criteria that, if a location were recounted, 20% of the time the observed volumes would meet the 200-500 and >500 vehicle warrant criteria, respectively. Similarly, pedestrian volumes of 13 and 26 could be observed 20% of the time as meeting the 15-30 and >30 pedestrian warrant criteria, respectively. Conclusions This study validates the use of the canonical 10% variability in vehicle volumes for Calgary. It also provides a default value for variability in pedestrian volumes. Evaluations based on fixed thresholds should account for variability in the input data. Simulations based on empirical probability distributions provide a low cost, low effort method of quantifying and predicting the effects of variability in vehicle and pedestrian volumes.