What Works and What Backfires: Mode Shift, Rerouting and Departure-Time Policy on a Saturated Kathmandu Corridor

Abstract

Travel-demand management offers several instruments,
and which of them a city should use is an empirical
question about its demand and its network. This paper tests four
on the Tripureshwor–Koteshwor corridor in Kathmandu, using
a SUMO testbed, run mesoscopically, whose demand matches
2019 classified counts at 42 locations (GEH<5 at 95.2%), at 24
treatment-assignment seeds per grid point. Two work and two
backfire. Removing 15% of motorcycle trips reduces network
delay by 55.1 ± 2.8% and diverting 20% of corridor trips onto
alternative routes reduces it by 42.6 ± 2.5%; both raise cordon
throughput in the mean. The first is a ceiling rather than an
option: the buses needed to carry those riders reclaim as much
carriageway as the motorcycles vacate. Retiming departures
fifteen minutes earlier raises delay at every share tested, and
a school-hours shift raises it at both shares tested. The failures
have a measured cause: hourly counts from three Department
of Roads stations put the morning peak at 6.8% of daily vehicle
traffic against the 10–12% national design practice assumes, so
the profile is a plateau and retiming moves departures into hours
already as busy. A third result concerns method. The calibrated
network’s critical accumulation falls at about half the counted
demand, so it has no stable congested regime, and effects read
near that boundary depend on the loading they are read at. We
report the diagnostic sequence, including a network-build defect
in which unmarked six-metre carriageways are encoded as single
lanes.