Bo Lan

Evaluation of Multivariate Poisson Log Normal Bayesian Methods for Before-After Road Safety Evaluations

The multivariate Poisson log normal (MVPLN) Bayesian method has recently been introduced in road safety analysis mainly for network screening and using different crash severity levels. However, there is little or no research applying MVPLN to different crash types. Besides, only one model structure for the expected crashes of a given severity was investigated in previous MVPLN studies. Another knowledge gap is that this method has not yet been evaluated for before–after treatment effect analysis. The objective of this study was to evaluate the application of MVPLN Bayesian method for before–after road safety evaluation studies. Two groups of unsignalized California intersections, for which a naive before–after comparison shows a significant change in the crash frequency after a hypothetical treatment was assigned, were used to conduct the study. It was found that the crash reduction rates are sensitive to the function form of expected crashes. For each model structure, MVPLN, univariate Poisson log normal (PLN) and Poisson-gamma models provided comparable results while PLN was seen to be superior. Finally, models that consider temporal effects of unobserved latent variables were found to be superior to those that don’t.

Development of Crash Modification Factors for Uncontrolled Pedestrian Crossing Treatments

The objective of this study was to develop crash modification factors for four treatment types: rectangular rapid-flashing beacon (RRFB), pedestrian hybrid beacon (PHB), pedestrian refuge island (RI), and advance yield or stop markings and signs (AS). From 14 cities throughout the United States, 975 treatment and comparison sites were selected. Most of the treatment sites were selected at intersections on urban, multilane streets, because these locations present a high risk for pedestrian crashes and are where countermeasures typically are needed most. For each treatment site, relevant data were collected on the treatment characteristics, traffic, geometric, and roadway variables, and the pedestrian crashes and other crash types that occurred at each site. Cross-sectional regression models and before–after empirical Bayesian analysis techniques were used to determine the crash effects of each treatment type. All four of the treatment types were found to be associated with reductions in pedestrian crash risk, compared with the reductions at untreated sites. PHBs were associated with the greatest reduction of pedestrian crash risk (55% reduction), followed by RRFBs (47% reduction), RIs (32% reduction), and AS (25% reduction). The results for RRFBs had their basis in a limited sample and must be used with caution.

Changing the Future? Development and Application of Pedestrian Safety Performance Functions to Prioritize Locations in Seattle, Washington

This study aimed to use robust analysis methods to identify and screen locations at risk for pedestrian crashes and injuries to help Seattle, Washington, a Vision Zero city, broaden treatment priorities beyond only high-crash locations. For this objective, data from the entire network were used to develop safety performance functions (SPFs) for two pedestrian crash types: total pedestrian crashes at intersections (a high frequency type) and a subset of intersection crashes involving through motorists striking crossing pedestrians (a high severity type). Many variables from roadway, built environment, census, and activity measures were tested. A similar but not identical set of variables, including measures of activity and intersection size and complexity, significantly contributed to crash prediction in both models. Pedestrian volume exhibited a curved relationship to crashes and demonstrated a tendency for expected crashes to begin to decline above a threshold value; however, the causes of this relationship were unknown. The SPFs were used in several ranking methods, including SPF-predicted crashes, empirical Bayes estimated crashes, and potential for safety improvement, to aid in prioritization of locations that might have been candidates for safety improvement but that had not necessarily experienced a high frequency of crashes. On the basis of this example, this approach is feasible for jurisdictions that wish to be more proactive in addressing potential crashes and injuries. Jurisdictions must, however, begin routinely collecting the data needed to implement the method efficiently.

Safety Evaluation of Discontinuing Late Nighttime Flash Operations at Signalized Intersections

Late-night flash (LNF) mode at signalized intersections consists of traffic signals that flash yellow for one road (requiring drivers to show caution but not stop) and flash red for the other road (requiring drivers to stop, then proceed after yielding to the main road). Although the intent of this mode has been to reduce energy consumption and delay during periods of low traffic demand, safety concerns have caused agencies to begin replacing LNF with normal phasing operations. Studies of the safety impacts of replacing LNF with normal phasing operations have indicated that doing so will reduce night crashes; however, a limitation of many studies is a potential bias due to regression to the mean (RTM). In this article, the authors examine the effect of eliminating LNF operations at signalized intersections using state-of-the-art methods and to address the noted limitations. The authors aimed to include an adequate sample of locations for which traffic volume data were available.This study examined the safety impacts of converting late nighttime flash operation to normal phasing operation at signalized intersections by using the empirical Bayes, the univariate full Bayes, and multivariate full Bayes before-and-after methods. Data were obtained from the North Carolina Department of Transportation for 61 treatment sites and 395 reference intersections that remained on late nighttime flash operation from 2000 to 2007. The results from the empirical Bayes method are almost identical to those of the univariate full Bayes. The full Bayes method offered more flexibility in selecting the functional form of expected crashes at similar sites (similar to the safety performance function in the empirical Bayes) and in addressing uncertainty in the data. Compared with the univariate full Bayes, the multivariate full Bayes with the multivariate Poisson lognormal (MVPLN) model provided better results based on much lower deviance information criterion values. The MVPLN model was favored and the recommended crash reduction factors are 48% (±6%), 53% (±8%), and 57% (±7%) for nighttime total, injury and fatal, and frontal impact crashes, respectively.

Safety Evaluation of Converting Traffic Signals from Incandescent to LED Bulbs

Data from 282 signalized intersections in Charlotte, North Carolina, were used to examine the safety effect of converting the signals to composite LED bulbs. An empirical Bayes before-and-after method was used for the evaluation. Since this was a blanket installation by the city of Charlotte, a comparison group of stop-controlled intersections was used to account for possible trends during the study period. Crash modification factors (CMFs) were estimated for three- and four-leg intersections for eight different crash types including crashes at dawn and dusk and in dark conditions. For three-leg intersections, all CMFs were higher than 1.0; this factor indicates a possible increase in crashes due to the LEDs. However, none of these CMFs were statistically different from 1.0 at the .05 significance level. For four-leg intersections, the CMFs associated with rear-end crashes were lower than 1.0 and statistically significant at the .05 level; this finding indicates a reduction in these crash types following the changeover to the LEDs. There was substantial difference between the sites in terms of the effect of the LEDs. The reasons for these differences are not known at this time. Future research should investigate whether LEDs are more or less beneficial depending on the characteristics of the intersection including type of area, sight distance, traffic volume, and phasing scheme.

Crash Modification Factors for the Flashing Yellow Arrow Treatment at Signalized Intersections

This paper presents the results of an evaluation of the flashing yellow arrow (FYA) treatment using data from signalized intersections in Nevada, North Carolina, Oklahoma, and Oregon. The evaluation method was an empirical Bayes before–after analysis. The treatments were divided into seven categories depending on the phasing system in the before period (permissive, protected–permissive, or protected), phasing system in the after period (FYA permissive or FYA protected–permissive), the number of roads where the FYA was implemented (one road or both roads), and the number of legs at the intersections (three or four). The first five treatment categories involved permissive or protected–permissive phasing in the before period. Intersections in these five treatment categories experienced a reduction in the primary target crashes under consideration: left turn crashes and left turn with opposing through crashes. The reduction ranged from 15% to 50%, depending on the treatment category. Intersections that had at least one protected left turn phase in the before period and had FYA protected–permissive left turn phase in the after period experienced an increase in left turn crashes and left turn with opposing through crashes, indicating that replacing a fully protected left turn with FYA will likely cause an increase in left turn crashes.