Transforming Catalonia’s Mobility System for Net Zero

Annex C. Methodological note: data analysis on travel time per mode

The method described below uses Google Maps Directions API to obtain and compare travel times per mode (taking account of public transport schedules and traffic conditions) for randomly selected sets of same-distance trips, via 2 steps.

The first step is to generate origin-destination pairs (OD-pairs) to feed into the Google Maps Directions API. OD-pairs between building locations1 are generated using Python libraries OSMnx and Shapely, based on data from OpenStreetMaps for different trip lengths (e.g. 1.5, 4, and 6 kilometers) for selected administrative boundaries (also based on OpenStreetMaps data). Trip origins are randomly-selected buildings at a 3.8 – 4.2km Euclidian distance from trip destinations. Trip destinations are randomly selected buildings within 200m of the administrative city center. Figure A C.1 shows the result of 100 4-km OD-pairs for the city of Tarragona.

Figure A C.1. 100 4km-trips OD-pairs to the city centre of Tarragona

Note: Trip origins are randomly-selected buildings at 3.8 – 4.2km Euclidian distance from trip destinations. Trip destinations are randomly selected buildings within 200m of the administrative city centre.

Source: Authors.

The second step involves computing travel times per mode using the Google Maps Directions API. Google Cloud Services API uses the following inputs to calculate travel times per mode:

• Origin: any coordinate

• Destination : any coordinate

• Mode: 'driving', 'walking', 'cycling' or 'public transport'

• Departure time: any time in the future

The API provides recommended routes to the destination and travel times, rounded to seconds, including time spent in traffic for driving (based on expected congestion) and frequencies for public transport. The Google Maps API does not calculate time spent for getting to the vehicle and searching for parking at arrival. To display an as-realistic-as-possible comparison, an estimate of parking searching time can be added to the displayed travel time by car provided by the API, based on (ITF, 2019[1]), as follows:

- Up to five minutes are added for car trips to compensate for the time spent accessing the vehicle at the trip’s origin. One minute is added if the population density is at the bottom 20%, five minutes if the population density is at the top 20%.

- Up to ten minutes are added for time spent looking for parking and the time required to reach the destination. One minute is added if the population density is at the bottom 10%, ten minutes if the population density is at the top 10%.

Figure 3.8 in Chapter 3 shows how travel time is distributed for each mode, on average, for a sample of 100 randomly generated 4-km trips from the periphery to the city centre of Tarragona. Travel times were simulated for Wednesday 25 September 2024 at 8:30 am, with the aim to simulate a travel to work during the week (obliged mobility). To indicate parking searching time, Figure 3.18 in Chapter 3 uses an estimate made by the local administration, rather than the ITF estimate above.

The method described in this Annex could be used to analyze specific destinations such as industrial areas, travel times in other cities in the region, for different trip lengths, and for other days of the week and times of the day. The method precision could be improved by doing a sensitivity analysis of simulations run with varying underlying assumptions, and by using network distance rather than Euclidian distance to generate OD-pairs. Methodological literature on measuring accessibility, such as (Miller, 2020[2]), describe in more detail possible approaches to evaluating travel time and travel distance.

Note

← 1.  Building locations are proxies for trip origins and destinations.