A6.2 - Choosing to undertake an accident analysis

Introduction

Several factors affect the decision of whether or not to undertake an accident analysis and the choice of method used for that analysis including:

• the nature of the site (eg, AADT, length)
• the availability of a reliable accident history for at least five years
• the availability of suitable accident prediction models or exposure-based accident prediction equations
• if the project options will result in a fundamental change in the site

The following sections discuss these factors and their effect on the accident analysis.

Site

A site is the specific road infrastructure for which an evaluation is carried out. A site can be a bridge, intersection, mid-block, curve, S-bend, etc, or any combination of these, eg, a mid-block and an intersection. In the case of combinations, a site may have to be broken into parts for the purpose of evaluation.

Accident history

For the purpose of accident analysis, generally a minimum of the past five years (sixty months) of reported accident history is used. This reduces the error caused by regression to the mean.

The principle of regression to the mean states that when an earlier measurement is either extremely high or extremely low, then the expected value of later measurements will be closer to the true mean than the observed value of the first.

The effect of regression to the mean can be reduced by using a longer accident history when investigating accidents at a site, and by ensuring that there is a commonality amongst accidents at the site.

Completeness of accident history data

The latest data available in the Crash Analysis System (CAS) should be used for accident analysis. As there is typically a lag between the time when an accident occurs and when it is entered into CAS, care should be taken to ensure that the data being used is complete.

When establishing the accident history, it is considered good practice to check all the Traffic Crash Reports (TCR) along the length of the site and up to one kilometre either side. Where possible, the location of serious and fatal accidents should be discussed with the loc al Police to confirm the location, particularly along roads where it is suspected that accidents may have incorrect locations noted in the TCR. At sites with low accident occurrence, the impact of an incorrectly coded accident in the TCR, particularly a serious or fatal accident, can have a major impact on accident benefits (both positive and negative).

Local accident data

Transit NZ and local authorities have set up systems that involve the collection of local contact accident data (also called 'contractor reported' or 'unreported to Police' accidents) from contractors, local residents and network management personnel. The quality of this data varies and caution should be taken when using it in accident analysis.

Local contact accident data can be used in an accident-by-accident analysis (Method A) where the data is supported by sufficient evidence to be audited and a reasoned justification provided as to why it should be used to supplement information from CAS. Evidence might include a second independent report of the accident, confirmation of accidents by the local Police or by local network contractors or consultants.

If local contact accident information is used for an analysis then under-reporting factors must not be included in the calculations of injury or non-injury accident costs.

Site characteristics

There are four site characteristics which have an impact on the time-span of accident history required and the method used for analysis:

• the traffic volume through the site
• whether or not there has been a major change at the site
• whether or not it is a new site (eg, new road or intersection)
• when there is no accident history

The table below illustrates the adjustment to the accident history requirements or the choice of accident analysis methods resulting from these characteristics.

Minimum number of accidents required for Method A

The use of Method A for accident analysis requires that a minimum number of accidents have occurred at the site, depending on the length of the site as follows:

at intersections or sites less than 1 kilometre in length, within the last 5 years there have been:

• 5 or more injury accidents; and/or
• 2 or more serious or fatal accidents;

at sites longer than 1 kilometre in length, within the last 5 years there have been:

• 3 or more injury accidents per kilometre; and/or
• 1 or more serious or fatal accident per kilometre;

Generally, there should be some commonality amongst the accidents that have occurred.

Where a site does not meet these minimal requirements, then Method C (weighted accident procedure) may be used.

Fundamental change in a site

A project option results in a fundamental change in a site when the types of accident or the level of accident severity is expected to change significantly. The following list gives examples of site changes that would result in a fundamental change:

• a completely new site is being provided (such as a new road or intersection)
• realignment of a road (other than an isolated curve)
• removal or significant modification of road elements (eg, grade separation of a railway crossing and conversion of a single lane bridge to a two-lane bridge)
• change in intersection form of control
• flush median
• adding lanes, including passing lanes.

Project options that are not normally regarded as resulting in fundamental changes include:

• upgrade of a single or s-bend to a higher design speed curve or s-bend
• shoulder widening on rural roads (in the absence of road realignment)
• signage and delineation improvements, including lighting
• traffic volume changes (in the absence of other improvements)
• road resurfacing and shape corrections
• minor improvement works.

Method A (accident-by-accident) is normally applied to project options that do not result in a fundamental change in a site. When there is a fundamental change, Method B is generally used for analysis of the project option, while Method C or A can be used for the do minimum. In some cases, Method C or Method B may be used for both.

Where there is a fundamental change in a site but no accident prediction models or exposure-based accident prediction equations are available for the do minimum, Method A can be used for the do minimum while Method B is used for the project options, providing that models are available for the project options.

Area-wide changes in traffic networks

When considering projects of an area-wide nature, such as the evaluation of an urban traffic network, eg, for transport planning or traffic management studies, it is insufficient to calculate accident costs from changes in global totals of vehicle-kilometres of travel.

Where a new road link is being added to a network, or a network change will result in major redistributions of traffic, analysis is required of the incidence of accidents on the links to which the traffic is being diverted.

For a new link, use Method B accident prediction models or exposure-based accident prediction equations appropriate to its intended design, speed limit and intersections along it. On major links that experience significant changes in traffic volumes, accident prediction models are preferred (where available) over exposure-based accident prediction equations.

Availability of models and equations

In the absence of an adequate accident history for the site, Method B or C may be used, provided there is a suitable accident prediction model or exposure-based accident prediction equation available. A summary of the available models and equations is found in appendix A6.3 while appendix A6.5 provides the details about them.

Accident prediction models or exposure-based accident prediction equations other than those specified may be used if the robustness of these models or equations can be demonstrated to Land Transport NZ and a peer reviewer.

Guidance

The procedure below gives step-by-step guidance as to when an accident analysis may be required and what method(s) should be applied.

Selecting the accident analysis method

Follow the steps below to determine the need for an accident analysis and the appropriate accident analysis method(s).