Abstract

The paper analyses dwell time delays for commuter trains in Stockholm andTokyo. In both cities, small dwell time delays of at most five minutes makeup around 90% of the total delays. Therefore, it is valuable to understandand deal with these disturbances. To this end, we use high resolution data ondwell times and passenger counts from both countries over the last severalyears. We find that these data alone can explain about 40% of the variationin dwell time delays and produce simple models which can be used inpractice to assign more appropriate dwell times. A change of 15 passengersper car, in Tokyo translates to a delay of about one second. For every 10remaining passengers per door in Stockholm, the delay increases by aboutone second, and one boarding or alighting passenger per door correspondsto about 0.4 seconds of delay. We also find that trains in Tokyo are muchmore congested than in Sweden, and that at most stations in the latter, theexchange of passengers is modest. In both cities, the range of dwell timedelays is quite narrow, with between 40 and 50 seconds separating the 5thand 95th percentiles. This indicates further that most delays, by far, are verysmall, and that even small adjustments to dwell times can make a bigdifference in the overall picture. To facilitate such improvements, keystakeholders and practitioners are closely involved with the research.