Number of Lane Changes Determined by Splashover Effects in Loop Detector Counts

Lane changes are important in quantifying traffic for both operational and planning purposes. Traditional in-lane loop detectors do not count lane changes; hence, historically, traffic engineers have estimated them using other data sources. This paper provides a method for estimating the number of lane changes based on observations of “straddling” vehicles that are simultaneously detected by the loops in adjacent lanes. In the data considered here, such “straddles” almost always correspond to vehicles that are in the process of changing lanes. However, many lane changes take place between detector sites and, hence, do not result in straddles. The methods developed here estimate probability distribution for the number of lane changes given an observed number of straddles. The efficacy of this approach depends on calibration issues and on the size of the aggregation period. In the evaluation study presented here, the results are good: The proposed method gives the number of lane changes with approximately 10% error, even though the number of lane changes per aggregation period varies by a factor of 10 over time.