YY's bike shed: Traffic evaporation, and the models that can't see it

When road space shrinks, some traffic simply disappears

I updated the wiki page on traffic evaporation: the well-documented phenomenon where closing a road or reallocating its space makes some car trips vanish rather than divert, although models predict the opposite. A review of more than 70 cases across 11 countries found that on average 21.9% of the vehicles previously using the treated road (median 10.6%) could not be found anywhere in the surrounding area afterward, and that pre-project congestion predictions tend to be “excessively pessimistic”. Some striking cases:

• Seoul removed a 5.8-kilometer elevated highway and half the surface lanes to restore the Cheonggyecheon stream. Most of the former traffic did not reappear on nearby streets; people shifted departure times and switched to the Metro.
• Pontevedra, Spain saw a ~90% drop in car traffic after pedestrianizing its historic center.
• Barcelona’s tactical-urbanism streets lost 14.8% of their traffic relative to the rest of the city, while immediately adjacent parallel streets gained only 0.7%.

Traffic demand is not weather

A rain forecast does not make it rain, but a traffic projection that dictates a road’s design can create the traffic it predicts. I updated the traffic projection page on why these models deserve far more skepticism than they get:

• A model that holds car demand fixed can only move the assumed cars around the network. It cannot show people changing route, departure time, destination, mode, or skipping the trip, which is exactly what the evaporation cases above document.
• The forecasts can be self-fulfilling: the model projects more driving, the city widens the road, easier driving induces more driving, and the resulting traffic is cited as proof the forecast was right.
• WSDOT’s successive forecasts for Seattle’s SR-520 bridge (1996, 2002, 2011) each projected substantial growth while observed traffic stayed flat or declined for more than a decade.
• US models are calibrated on decades of car-centric development and almost no bike or walking data, so their mode-shift predictions are a shot in the dark. If you don’t count it, it doesn’t count.
• The better question is person-throughput: NACTO estimates a motor-vehicle lane moves 600–1,600 people per hour, a two-way bikeway 7,500, and a transitway 10,000–25,000.

My comments on Charlottesville’s 5th Street design and West Main

Following the August 6 meeting of the city’s Bicycle and Pedestrian Advisory Committee, I wrote up comments on the 5th Street SW Road Diet design and the West Main study, applying the two pages above. Some highlights:

• West Main’s multi-year utility replacement is a once-in-a-generation opportunity: construction breaks everyone’s travel habits, and whatever the rebuilt street makes easiest is what new habits form around. We should decide what street we want before reconstruction locks the old one back in.
• The proposed 5th Street design sandwiches a bike lane between a through lane and a right-turn lane, creating a mandatory weaving conflict and the classic right-hook crash pattern that NACTO explicitly advises against.
• Congestion is not traffic calming. A street must be safe at midday and at night, when there is no queue to slow drivers down.
• Daylighting means physically clearing the sightline near a crossing. A parking sign is not daylighting.

If you live in Charlottesville, feel free to borrow any of it, and let me know if you spot problems.

yy