Road Data Aggregation and Sectioning Considerations for Crash Analysis
Increasingly, roading authorities are collecting a variety of datasets related to their networks, including horizontal/vertical alignment, cross-section, traffic volumes, crashes, and the location of features such as intersections and passing lanes. This data may be a mixture of point locations, fixed-length records, and variable-length records. A critical question, both in terms of computational ease and practical usefulness, is how best to aggregate or section the available data into appropriate road segments for operational and safety analyses. This issue is becoming more pertinent with the development of tools such as Interactive Highway Safety Design Model and the Highway Safety Manual, which require a logical partition of roads based on many different attributes. The guidance on how to do this however is rather scant. Analysis of traffic exposure versus crash risk is also affected by using fixed or variable length segments. Research is nearing completion in New Zealand to combine road feature, geometry and crash data on the national rural State Highway network. The resulting database will enable better analysis of crash patterns against different types of road elements and be used to calibrate IHSDM for New Zealand use. This paper outlines the investigation done to determine a rational method for aggregating the available data into logical road segments. The resulting method uses horizontal alignment, significant cross-section changes, and changes in speed limit. It also attempts to minimize the number of very short segments. The resulting dataset contains approximately 83,400 segments generated from 20,900 lane-km (13,000 lane-miles) of highway.