Modeling Traffic Flow Phenomena Femke van Wageningen-Kessels Civil Engineering & Geosciences, University of Technology Delft, The Netherlands Thomas Schreiter * Civil Engineering & Geosciences, University of Technology Delft, The Netherlands Hans van Lint Civil Engineering & Geosciences, University of Technology Delft, The Netherlands Serge Hoogendoorn Civil Engineering & Geosciences, University of Technology Delft, The Netherlands ABSTRACT The increasing amount of detailed traffic data has led to the identification and the empirical under- pinning of numerous traffic flow phenomena, such as the capacity drop, stop-and-go waves or traf- fic hysteresis. On the basis thereof, in the past decades many traffic flow theories and models have been proposed and developed that explain and re- produce some of these phenomena. However, there is no undisputed theory or model which is able to reproduce all these phenomena. In this extended abstract we present a new classification scheme for traffic flow models on the basis of which model is able to reproduce and predict which traffic flow phenomenon. The resulting model - phenomenon matrix can be used is a support tool for traffic flow engineers and as a reference for scientists. Keywords: Traffic flow models, Phenomena, Capacity drop, Stop-and-go waves INTRODUCTION When observing traffic data, e.g. from floating cars, loop detectors, cameras or any other kind of data, some typical traffic flow phenomena are observed. For exam- ple, congestion usually sets in near a bottleneck, such as a lane drop, an on ramp or an accident, and then spills back upstream. The downstream front of con- gestion then remains at the same location. We also observe that the flow in the bottleneck just before the onset of congestion is usually higher than the outflow of congestion at this particular bottleneck. This is often referred to as the capacity drop. Another important ob- servation is stop-and-go waves. In a stop-and-go wave vehicles have a very low, or even zero, speed and den- sities are usually high. These waves are only a few * Corresponding author - Stevinweg 1, 2628 CN Delft, The Netherlands, Tel: +31 15 2784977, Fax: +31 15 2781723, E- mail: t.schreiter@tudelft.nl (a) Speed Contour Plot in km /h (b) Flow Contour Plot in veh /h Fig. 1: Raw loop detector data showing two traffic phe- nomena vehicles long but can travel large distances upstream. In Figure 1, empirical data are visualized to illustrate these phenomena (loop detector data from the Dutch A15R, 16 Nov 2010, aggregated over one minute and over all lanes of the carriage way). A lane drop at Kilometer 56 acts as an active bottleneck, as the front of the congestion is fixed at that location. Upstream, stop-and-go waves are clearly visible. Furthermore, the flow immediately before the breakdown at 16.00 was significantly higher than during congestion. Many traffic flow models have been developed dur- ing the last decades. They all try to reproduce some of