Non-recurrent congestion caused by incidents has become an increasingly serious problem in traffic systems. In this research, the characteristics of vehicle breakdown incident durations on M4 motorway in the UK were analysed. The goodness-of-fit test showed that the distribution of incident durations on motorways conforms to a Weibull distribution. Algorithms to predict the duration of incidents gave results at a level of confidence of about 70%. This result was encouraging because it was an improvement on the current method that relies solely on engineering judgement. This research illustrated that the limitation to the development of forecasting algorithms for incident duration is the lack of incident data with an appropriate level of detail. Although traffic management and information centres are well established, there is no national standard method to collect incident data in the UK. It is clear that standardisation of incident data collection is urgently needed to improve the accuracy of the prediction. Attempts were made to apply the theory developed on the motorway to urban areas. Therecords of incidents available from Leicester City were analysed. The results identified that routinely recorded data was inappropriate for assessing the transferability of the theory from the motorway to the urban environment. Therefore, an incident database with a user friendly data entry interface was designed to collect the appropriate data. For the covering abstract see ITRD E134653.
Abstract