Temporary speed limit reductions are a common countermeasure aimed at improving work zone safety, particularly when the work is occurring on or near the roadway. In theory, reduced speed limits may serve at least three important functions: reduce variability in travel speeds and the potential for work zone crashes, reduce average travel speeds and the severity of crashes when they do occur, and enhance worker safety. This project involved the investigation of driver speed selection in work zones throughout the state of Iowa. The sites for analysis were chosen based on input provided by the technical advisory committee (TAC). Although the approximate location of each work zone was known, there was no single reliable source to precisely determine the location, time, and type of work zone activity. Four different databases were mined to estimate the specific physical limits of the work zones as well as the time during which each work zone was in place. The four datasets included the following: Construction contract documents; Iowa 511 archive database; Advanced Traffic Management System (ATMS) database; and Iowa Department of Transportation (DOT) crash database; and As noted above, there were significant inconsistencies among these databases in regard to the work zone activities and locations. The work zone daily work reports (DWRs) were not available. In the absence of the DWRs, to obtain reliable work zone activity and location information, the researchers utilized the subset of data for which the available data sources (i.e., contract documents, 511, ATMS, and crash information) were consistent in terms of reporting whether work activity was ongoing at a specific time and location. Data for which one of the data sources indicated that a work zone was present but another source could not corroborate this information were discarded from the subsequent analyses. This significantly reduced the data set available for statistical analysis. The primary consequence of this data quality issue was that a sufficient amount of data was not available to conduct a reliable crash analysis. Consequently, the research focused on assessing the impact of work zone speed limit reductions on drivers’ speed choices. It is presumed these speed changes have a related impact on safety performance, though additional research is warranted to understand the nature of the relationship between speed and safety. Data were collected from nine construction work zones in Iowa during 2014 and 2015. Similar data were obtained for the preceding years during the same time period when work zones were not in place at these same nine locations. Given the large amount of high-resolution historical data, extensive data reduction procedures were utilized in this big data application. Quantile regression was employed to examine the impacts of work zone speed limits on speed distribution percentiles before and during construction activities. The results show that speeds were consistently reduced when work zone speed limits were in place. Overall, work zone speed limit reductions of 10 mph in locations where the normal statutory speed limit was 65 mph showed the largest reduction in all speed percentiles. It should be noted the present research couldn't include the exact type of work activity due to data inconsistencies. This affects the transferability of results to other sites, and it is highly recommended that work zone activity type should be included in future research to further corroborate the above results. Some of the innovative aspects of this research are as follows: * This research reused data being collected with the Iowa DOT’s Intelligent Work Zone systems. This ensured that the experimental setup was not influencing the driver behaviour. This also ensured that the seasonal variation can be accounted for and drivers were given enough time to adjust to the presence of a work zone. * Since real-world data was being used, several filters were designed to eliminate any noisy data, which is a commonly occurring phenomenon with permanent sensors. * High-resolution speed data were collected at multiple sites (constituting more than 7 gigabytes of raw data) and high performance cluster analysis was utilized for data visualization and discovery. * This study also highlights the inconsistencies that exist in the state of the practice of recording the work activities in real-world work zones (in contrast to pre-designed experimental setups). (Author/publisher)
Samenvatting