Safety Study of I-35W Improvements Done Under Minnesota’s Urban Partnership Agreements (UPA) Project

Abstract

Minnesota’s Urban Partnership Agreements (UPA), of which the majority were completed in November, 2010, consisted of a series of improvements addressing the congestion on Interstate highway I-35W corridor and in Downtown Minneapolis. MnDOT Problem Statement NS-329 noted that there was interest in extending some or all of these interventions to other corridors and called for an estimation of their safety effects to assist in making these decisions. Following the UPA, the frequency of rear-end crashes increased substantially in certain regions on I-35W. The objective of this study was to determine if the increase in crash frequency was due to changes in traffic conditions or was a direct effect of the UPA interventions. A preliminary analysis was done to determine priority crash type and study regions. I-35W, from its start to its junction with I-94, was divided into 17 one-mile sections, and bi-directional (northbound and southbound) crash frequencies in Before-UPA (2006-2008) and After-UPA periods (2011-2013) were compiled for each one-mile section. Rear-end crash turned out to be the most prevalent crash type, but the changing trend of bi-direction rear-end crash frequencies from the Before to After period varied among those one-mile sections. Our interest lay in those regions where there was an outstanding increase in the rear-end crash frequency in the After period, which were approximately the I-35W HOT region (from TH-13 to I-494) and the I-35W PDSL region (from 37th Street to 26th Street). Both the I-35W HOT and PDSL regions were divided into analysis sections based on constant flow and geometry criteria as well as the loop detector availability. Crash, loop detector, weather condition, and PDSL activation (only for PDSL sections) were compiled for Before and After periods for each analysis section. Rear-end crash records were extracted using MNCMAT, and hard copies of the original crash reports were then reviewed to verify the crash type, location, direction and time for each crash. Traffic conditions came from loop detector data retrieved using MnDOT’s DataExtract tool. The source of weather conditions during non-crash hours was MnDOT’s RWIS, while that of weather conditions during crash hours were taken from original crash reports. The PDSL historical operation data came from MnDOT’s log for the Intelligent Lane Control Signal (ILCS) located at 37th Street. Logistic regression models were established to estimate the change in rear-end crash risk in a given hour before and after the UPA project controlling for changes in traffic conditions and weather conditions. The analysis results showed that: (1) Most analyzed sections in the I-35W HOT region showed no significant change in rear-end crash risk associated with the UPA project except for Section S9 (southbound, just north of Minnesota River). Section N17 actually experienced fewer crashes after the UPA project, but the reduction was not as great as the change in lane occupancy would predict. (2) The PDSL region experienced substantial increase in traffic congestion following the completion of UPA interventions. This was due to the removal of the old TH 62 & I-35W bottleneck, causing the bottleneck move northward to the I-35W & I-94 junction. The observed increase in rear-end crash risk was not associated with the operation of PDSL when controlling for the changes in traffic conditions. (3) An “inverted U” relationship between rear-end crash risk and a proxy for traffic density, lane occupancy, when controlling for other factors, were seen in most of the analysis sections. Rear-end crashes were most likely when lane occupancies were approximately 20%-30%. This study demonstrated a methodology that could be used to evaluate the safety effects of freeway-related projects. To be more specific, this study worked out a way to estimate changes in hourly crash risk while controlling for variations in traffic conditions.