|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-RD-97-106
Rural Public Transportation Technologies: User Needs and Applications FR1-798
The first major activity in the research was to identify one or more existing data bases of geometric design, traffic volume, and accident data for interchange ramps and speed-change lanes that were suitable for testing the development of statistical models for accident prediction. In order to be useful in the planned statistical analyses, the various data files of geometric and accident data needed to be linked together by a common location identification system, such as the milepost systems used by many State highway agencies.
The candidate data bases that were considered included the data bases available in the FHWA Highway Safety Information System (HSIS) and data bases from non-HSIS States. The two existing State data bases that were found to be best suited to this effort were those maintained by the California Department of Transportation (Caltrans) and the Washington State Department of Transportation. A detailed review of these data bases was made and trips were made to each agency to discuss the collection and coding of their data. At the time of this review, FHWA had reached a decision to include both California and Washington as new States in the HSIS, so it was apparent that data from both States would be readily available in HSIS formats.
The Caltrans highway data base, as part of the Traffic Accident Surveillance and Analysis System (TASAS), was found to include a file containing data on each individual ramp on the California State highway system. This file identifies the configuration of each ramp (e.g., diamond, loop) and includes an estimate of the annual average daily traffic (AADT) volume for each ramp. Each ramp is identified by the county, route number, and milepost of the gore area at which the ramp entered or exited from the mainline freeway. The locations of accidents on each ramp are identified by a code indicating whether the accident occurred at the beginning of the ramp, in the middle of the ramp, at the end of the ramp, or on the arterial crossroad (when the arterial crossroad was not a State highway with its own mileposting system). However, no data on the geometrics of individual ramps were included in the data base. If geometric data were to be considered in statistical analyses, they would have to be obtained from aerial photographs, photologs, or field visits.
The Washington data base also included data on each ramp in the State highway system. While the Washington data base did not include data on the ramp configuration, it did include a number of key geometric variables that were not available in the Caltrans data base. In particular, the Washington data base defined the cross section of each ramp (e.g., number of lanes, surface width, shoulder width), including variations in the dimensions of these cross-section elements along the length of the ramp. Accident locations were defined by mileposts along each individual ramp, which made it possible to link the accident data to particular geometric features of the ramp. For example, it was found that with some manual data reduction from existing interchange diagrams maintained by the Washington State DOT, it would be possible to distinguish ramp accidents from speed-change lane accidents. Although ramp configuration data were not available in the existing data file, it was also found that this could be determined directly from the interchange diagrams. Several other barriers to using the data (discussed in section 4) were also identified, but it was determined that these barriers could also be overcome through some manual data reduction.
The only major concern with the Washington data files was that traffic volume data were not available for every ramp. A separate ramp traffic volume file was available, but it included AADT data for only about 67 percent of the ramps. Previous research has shown that traffic volumes are the strongest predictor of ramp accident experience.
Based on the factors discussed above, it was decided that the advantages of the Washington data base were so great that they outweighed the lack of traffic volume data. The Washington data base includes geometric data for ramps at a level of detail that is not available for any other State. Thus, a decision was reached to obtain and use the existing Washington data, but to limit the analyses conducted to those ramps for which AADT data were available.
Topics: research, safety, design, intersection safety, speed management
Keywords: research, safety, accident modeling, traffic accidents, geometric design, interchange ramps, Poisson regression, speed-change lanes
TRT Terms: Traffic accidents--Statistical methods, Roads--Interchanges and intersections, Geometric design, Speed change lanes, Accident data, Regression analysis, Poisson distributions