Improving Manual Counts of Turning Traffic Volumes at Road Junctions

Abstract

Turning traffic volume data gathered manually at four intersections have been studied to determine if either Fratar, Furness, or Kruithof models could be used to design labor‐saving strategies for manual counts of traffic flows at road junctions and to detect observation errors in a given data set. Each model was used to recompute the observed turning volumes by using different abstractions of the original data as model inputs. The standard errors obtained correlate partly with the level of traffic flow and the magnitude of error‐inducing features at the observation sites. The results indicate that the models could be applied to obtain an indication of the magnitudes of observation errors in given data sets. It is also demonstrated that the models can be used to design a labor‐saving observation strategy for T‐intersections in all cases and for fourlegged intersections when reliable historical data exist. The Kruithof model was found to give the least errors, and it generally entails the fewest calculations.