3. Technical difficulties were experienced resulting in the loss of encroachment data.

After discussions with VicRoads and the Victoria Police, a more reliable method was developed based on video recording traffic from an unmarked private vehicle. This procedure proved to be successful although it required considerably more staff time, thereby limiting the design of the study from that originally proposed.

Observations of red light running behaviour were taken at three carefully selected red light camera approaches and six matched non-camera control approaches around Melbourne. Red light running behaviour was measured as a function of speed zone, road cross section (divided or undivided), time of day, day of week and vehicle type.

The results of the study, in relation to the four main objectives were as follows:

Objective 1 - Extent of Red Light Running
Red light encroachments were relatively rare events and, for the most part, those observed were not all that dangerous. However, there was a relatively small number (8 in over 38,000 vehicle movements) that were dangerous manoeuvres during the full-red interval. If this rate is extrapolated to all traffic signal approaches in Victoria, it represents more than 500,000 dangerous encroachments per year. Though not able to be rigorously investigated in this study, there was also a suggestion in the data that trucks may be over-represented in red light running during the all-red interval. The results suggested that a simple, inexpensive countermeasure to reduce crashes resulting from red light running may be to increase the all red periods at signalised intersections by one or two seconds, although the behavioural consequences of this need careful consideration.

Objective 2 - Road and Environment Effects
The speed limit and cross-section of an approach appeared to influence red light encroachments only for right-turners. Arterial roads with 60 km/h limits and undivided roads had higher rates of illegal right-tum encroachments than did 80 km/h and divided roads.

Red light running rates were generally higher in right-turn than in through lanes, but were particularly so during the evening peak period. Further, there were fewer right-turn encroachments at the site where all approaches had fully-controlled right-tum phases.

Interestingly, there were no statistically reliable differences in red light encroachments observed between camera and non-camera approaches. It is unlikely that this could be attributed to traffic flow differences or other environmental influences as these were relatively constant across test and control approaches for any given site. The results suggest some investigation of the current mode of operation of an automated red light enforcement system. the red light camera program may be warranted.

Differences between week days and weekend days could not be analysed thoroughly because football traffic had an extreme influence on the weekend data, preventing a proper analysis of the red light running behaviour of drivers on weekends. This finding may be highlighting a substantial safety problem associated with football traffic (or sporting event traffic generally) and warrants further investigation.

Objective 3 - Red Light Running & Crashes
Andreassen (1995) found that the installation of RLC's at 41 sites studied did not result in any reduction in accidents at those sites. Further, he found no significant differences between crashes at RLC sites compared with signalised intersections in Melbourne generally. Thus, there was no need to do an extensive analysis of the correlation between red light running behaviour observed in this study and crashes. Nevertheless, a simple correlation analysis was undertaken for red light running data in the current study and revealed no significant relationship between the frequency of crashes at RLC and non-RLC sites and differences in red light running behaviour.
Objective 4 - Extension of the RLC Program

While the need for more RLC installations was not justified by this, or the recent accident study of Andreassen (1995), nevertheless there are grounds for improving the operation of an automated red light enforcement system. existing RLC locations. Highest priority should be given to increasing the perceived risk of detection at these sites. This could be done by overcoming existing technical difficulties at some sites, by increasing the level of activity and by supportive publicity. Subsequent testing might then indicate whether there is a need for more RLC installations.
Recommendations

The three basic recommendations from the findings of this study were that:

* An action program be undertaken to improve the effectiveness of existing Red Light Camera operations in Melbourne as outlined above;
* Further research be undertaken to test red light running behaviour more extensively than was possible in this study. A focus on red light running behaviour associated with weekend sporting events and/or heavy vehicles may be especially worthwhile;
* Consideration be given to increasing the duration of all-red periods by one or two seconds, as a simple, inexpensive countermeasure for crashes caused by red light running, taking into account the traffic capacity and behavioural consequences of such a change.