Surveying traffic delays at a roundabout near Plymouth

Traffic delay models can be no more accurate than the data on which they are based. This paper attempts to highlight some of the problems of site data collection experienced in a survey of traffic delays at a roundabout near Plymouth. Data quality of the level required for high level definition modelling of roundabouts may not always be available at reasonable costs, placing constraints on the practical range of modelling techniques which may legitimately be attempted. Surveys of observed vehicle delays are discussed concentrating on problems of estimating peak flows, vehicle composition, occupancy and queue length. Physical overloading of enumerators involved a reduction in the number of vehicle categories which could be recorded, and a need to sample only one minute in each five minute period. Problems of observing 4 lanes of vehicles, often in low sun, rain or spray conditions, often in waves are noted. The inadequacy of alternative procedures such as video or tape recording of number plates at exits up to 300 meters distant is also highlighted. The need to be able to model the effects of traffic at roundabouts, and possibly present results including the moments of delay distributions is one which may challenge modellers in the future.