These days the city of Utrecht is the epicentre of roadrace cycling as the city hosts Le Grand Depart of the Tour de France. With Utrecht being the navel of the cycling world, it is amazing to see how much efforts are made to make everything going smooth. Parts of the city were renovated, like the Mariaplaats, where car space was removed to give space for terraces and promenading people. Other parts will have their complete make-over after the Tour, the parcours was a good excuse to temporary remove obstacles and rethink the design of important junctions. Also the Tour itself is a masterpiece of technique and knowledge. When the 198 racers will have their individual time trial over 13,8 kilometers, the differences will be measured on the spot in milliseconds and broadcasted for 3,5 billion spectators in 190 countries all over the world. What a difference when the daily peloton takes over the streets of Utrecht again when the professionals leave the city!
Until last year the city had very limited data of the cyclist. No fancy bicycle counters like in Copenhagen, or even Helmond, no tracking like we have in the Netherlands for public transport travellers, by the OV-Chipcard. The city counted the number of cyclists one week a year on two cordons, one in the inner city and one in the first ring. Althoug we have high cycling rates, this doesn’t mean we have a sound policy process of analysis, measures and evaluation. By tradition we have limited research and most pro-cycling decision making is done on incidental observation in the streets, educated guesses of experts and the convincing power of Fietsersbond. For me this is bothering, as a policy maker I want to rely on facts to prove the necessary measures and to allocate budget. And besides the city of Utrecht I have to serve other municipalities, who have even less insights in the daily pattern of their cycling population.
These municipalities apply for grants at the Province (in the past at the Region of Utrecht) to improve the bicycle network, but in times of limited budgets it is hard to weigh the different measures in different municipalities. And, as the bicycle network in the Netherlands and specially in the Utrecht region is rather mature, network improvements are costly, like the building of bridges and tunnels. So, there is demand for better insights in the travel pattern of our population and the possibilities to improve their quality of journey. It was by coincidence of visiting a Velo-city conference abroad I encountered the transport model skills of a Finnish company. Their approach is different of the traditional models, who have cycling included, but on a very vestigial way. In the Finnish approach all individuals are modelled, based on a sample of diary information, and placed on a very detailed network. This approach serves better to the dense pattern of cyclists in cities.
Test model with a surprise
Luckily I got the chance to try the modeling for a sample for the campus area in Utrecht. Most readers will know there is massive cycling every day to and from the University, and as the number of institutions was rather limited, I could get acces to origin data of the employees and students. For the network data we make use of the network gathered by the volunteers of Fietsersbond for their unbeaten journey planner. The very accurate geographical data is enriched by hundreds of volunteers for the kind and quality of the surface, and also with less objective qualities like pleasantness and attractiveness of the surroundings. A very good source to base the route choice algorithm on, although the model still surprises me. In the test model i found an oddity and to find out the truth, I went out in the rush hour to see it myself. My Velo-city presentation of this year was about this surprise, it turned out the model found a route that is used by thousands of cyclists, but is not on the map of main cycle routes!
Full model under construction
Last week we presented the full model for the province of Utrecht for a selection of experts. The model is still under construction, we must validate it with countings, route tracking apps and other nifty location based tools. And we still have some issues to solve. The share of bike & train trips (a huge share of traffic in some parts of the cities!) is inaccurate, due to the limited number of registered trips in the diary data. We also face the inaccuracy of national diary data for a lot of bicycle trips to cafes and bars in the centres of cities. And we have the so-called “lunchbox problem” for the primary schools. These schools generate a lot of bicycle traffic for trips on a very short distance, less than one kilometer. Important to have them in the model too. Some schools still have a long lunch break to give children the option to cycle home and have lunch there. Logically this doubles the number of bicycle trips to those schools. I was told this typical Dutch habit is in use at 25% of the primary schools, but it is unclear which have this long lunch break, and which not. Any suggestion for an automated solution is welcome.
Preview of the model
The model itself is currently working as an webbased map and will be public available in the future for consultation. Until now I leave you with some screenshots. The numbers represent an average day, keep in mind we have large differences during the day, in the week and even in the year. The color differences represent the “bike ability index” calculated for the network by attributes of the links, blue is poor, purple better and red (not on this map excellent.
As a bonus the model also delivers an outcome for pedestrians, unique for the Netherlands, but very, very unreliable. Specially this weekend, when hundreds of thousands of spectators are expected for watching the Tour de France and celebrating the best bicycle city of the world.