Network Analysis, Travel Time, and Accessibility

Wu Wenhao
8 min readDec 17, 2018

--

“Everything is related to everything else, but near things are more related than distant things.”

— — Tobler’s First Law of Geography

There are many types of geospatial analysis that can reveal useful insights in city planning and urban design, especially regarding topics of a spatial nature, such as measuring transportation accessibility and walkability for the existing condition and proposed scenario studies. Actually, network analysis is one of the most frequently used toolkit that can be valuable in such a process.

Network analysis, as indicated by the word “network”, models the mobility or transportation network and performs a variety of analysis on them. This network, in the context of urban planning, can be formed by streets grids and blocks, points of intersections, highways, metro lines, railroads, ferry and air routes etc. The big idea behind is trying to understand how different locations within the given network can be reached by certain types of travel mode, how easy, and at what costs. (e.g. see here an introduction about functions of network analysis by ArcGIS product. (What is the ArcGIS Network Analyst extension? — Help | ArcGIS Desktop)

What’s good about Network Analysis?

Take walkability as an example. Do you often see the 5 or 10 min walk circle in planning? Following is a comparison of new urbanism style neighborhood design and planning across different times. Clarence Perry designed in 1930 the “ideal” urban neighborhood with a compact nature — at the center of the map sits a community center, from which a 1/4 mile (400 meters) or 5 min walk circle expands out and covers the majority of the neighborhood. In 1980 Douglas Farrproposed a new suburban typology of the neighborhood also with a compact mix pattern. In 2008, Andres Duany’s plan inherits Perry’s proposal and builds upon it with denser and more mixed-use programs proposed. However, the walking circle never changes — it is still a perfectly shaped circle with a 1/4 mile radius, indicating that most of the neighborhood can be reached by foot within 5 mins — what a wonderful walkable community!

But we don’t walk exactly like that……

The idea behind a perfectly-shaped walk circle is that it shows the possible reachable area by foot — it is possible to walk to the edge of the circle through and only through a straight line street from the original center within 5 minutes. So in each direction from the center, it would need a straight line street to make the area represented by the circle fully walkable. The problem is that some areas in the circle could be not walkable at all — buildings, derelict land, fenced-off private property land, etc. Some other areas are connected only by a zigzag road or a very curvilinear route, through which we need to walk way more time to get there. The perfectly-shaped circle is not showing this layer of information.

Walk circle as in New Urbanism style neighborhood planning

Network analysis can help improve this situation. Using the City of San Francisco as an example, this map below shows the location of the SF City Hall in the center. Using the street grid of San Francisco as the network, a network dataset is created and the toolkit calculates the 5min, 10min, 15min and 30min walking distance (or 400m, 800m, 1200m, and 2400m). The different colors, from dark green, light green, to yellow and red, represent the reachable area by walking along the streets in different amount of time. You can specify the type of the turns, and include one-way street specification in building up the base network, to simulate a more actual situation.

Walk shed from the SF City Hall based on street network (made by author)
Walk shed from BART stations in San Francisco — Oakland Area (made by author)

It can also work on larger scale accessibility problems. The map above shows the walksheds from the BART stations across San Francisco and Oakland. The green buffer areas around the stations are the immediate 200-meter proximity zone measured by walking distance, while the read ones are more than 2 kilometers away — usually beyond what is considered as “walkable” areas. Another map below shows the different travel time by interstate highways from several selected cities in the east and mid-west of U.S. on a regional scale. This type of maps could reveal useful insights in issues such as siting of business or logistics centers, public facilities or even planning a trip get-away.

Looking at our area of interest (AOI) or interesting sites based on these walksheds or travel time sheds, we can get a better understanding of the location’s connectivity through specified transportation mode.

Travel time from selected American cities by highway

To create better simulation results, you can also specify the different types of “streets” in the network your calculation is based upon: for Federal Highway, the speed can be the highest, and thus the impedance of the travel may be set as the smallest; the State Highway then may have a lower speed and a little higher impedance. Further applies to city level expressway and local roads — All these will be quantified by a number that corresponds to the level of impedance traveling along the route.

Illustrative plan of proposed Stanboroughbury town, DFZ

Image how this plan diagram above can be drawn differently? Showing the different walksheds with different time highlighted above the buildings and landscapes, etc. It could be a more accurate representation.

However, executing such spatial analysis would require: 1) acquiring data to build a network database; 2) performing analysis with the appropriate tools. Understanding the flow of pedestrians or vehicles, and what types of infrastructures they use is important in building a network database that makes sense. Implementing the analysis can take some time, depending on the size of the network and its ultimate resolution or how detail the dataset is defined.

On the prospect of improving the analysis speed, the more recent trend, as things are moving toward open source nowadays, is that functions of network analysis can be completed on the cloud and on-the-fly. Below are a few example web tools to calculate transit accessibility and travel time.

Realtime travel time mapping by WebCAT
Travel time by public transit by Mapnificent
Geotrellis transit by Azavea

To consider impenetrable zones, such as buildings, rivers, or forbidden areas, a way to go is to rasterize, meaning that to categorize those areas not welcoming walking with a high impedance level or super slow to penetrate or cross. This involves a different set of tools that are built for raster-based calculation and thus more sophisticated yet more flexible.

As big data infrastructure and cloud-computing becoming increasingly available and accessible, one noticeable trend is to apply data-heavy computing power on planning and design problems at an unprecedented scale, with flexibility and interactivity.

Google Earth Engine, a global scale raster imagery computing and analysis platform, is part of this trend empowering on-the-fly geospatial analysis. The snapshot below shows the 15 years trend of night light change, thus indicating urbanization or de-urbanization in the Chile — Argentina — Brazil region in South America. Via a built-in predictive model regressing on the intensity of night light from 1999 to 2013, this analysis identifies the potential areas with an increasing nigh light intensity trend in green (thus new urbanized area or areas that have been growing), the areas relative in stable (blue) or decreasing (pink) trend. See all the green transportation corridors? The metropolitan of Buenos Aires and all the satellite towns around it and linked towards it through the transportation network. Chilean capital city of San Tiago and one of the largest Brazilian cities São Paulo are also showed in this snapshot.

Figure. Tracking global urban growth trend with time series night light data (made by the author with GEE)

In addition, back to urban scale computing, the CityForm Lab also developed a powerful toolkit called the Urban Network Analysis, with built-in functionalities studying the physical pattern of cities through a network approach, by measuring a series of quantitative indicators. Here is a pretty cool intro video for the toolkit.

Other functionalities

Other questions it can answer include where is the best route from A to B? How to optimize your logistics network? What is a good city and town network in a metropolitan region? Spatial problems like these could be an important factor in many decision-making processes, and thus the resolution of them can bring value to the process and justify the planning and design decisions we make. Tools available such as ArcGIS network analyst extension and QGIS can provide great convenience in performing this type of study.

At last, a few interesting related posts on network analysis’ application:

- Travel time and public health

https://medium.com/@abertozz/mapping-travel-times-with-malariaatlas-and-friction-surfaces-f4960f584f08

- Travel time and trip planning

https://blog.conveyal.com/visualizing-urban-accessibility-with-opentripplanner-analyst-60ff58b97a31

Also, several useful links related:

- Performing network analysis with QGIS (given that ArcGIS is an enterprise license based while QGIS is open source and free while giving many similar capabilities.):

https://docs.qgis.org/2.8/en/docs/training_manual/vector_analysis/network_analysis.html

Bibliography:

1. DPZ, Approach to Planning and Urban Design, 2012.

https://www.dpz.com/Services/UrbanPlanning

2. ESRI Help section, What is the ArcGIS Network Analyst extension? 2018.

http://desktop.arcgis.com/en/arcmap/latest/extensions/network-analyst/what-is-network-analyst-.htm

3. John Elledge, TfL has a tool that lets you map travel times to anywhere in London, and it is brilliant | CityMetric https://www.citymetric.com/transport/tfl-has-tool-lets-you-map-travel-times-anywhere-london-and-it-brilliant-1150, June 2015.

4. Mapnificent, Mapnificent — Dynamic Public Transport Travel Time Maps, 2018.

5. Azavea, Travelshed Demo — GeoTrellis Transit, 2017.

6. Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., & Moore, R. (2017). Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of Environment.

--

--

Wu Wenhao

designer by day, developer by night / Design / GIS / Programming @SOM San Francisco