1. Milky Way

The Milky Way Transit Authority shows the spiral of the galaxy, with major constellations and nebula appearing as transit stations. Perhaps unsuitable for navigating an FTL-enabled spacecraft, but it serves well for orienting terrestrial astronomers.

The Milky Way Galaxy

2. People

We are surprised to see Leonardo da Vinci at the intersection of the Comedians and Italian Artists lines in Simon Patteron’s otherwise wonderful Great Bear map of famous people.

Famous People

3. Music

Different genres of music are represented by different transit lines in Dorian Lynskey’s map, with bands and musicians appearing as transit stations. Björk is well-placed at the intersection of Pop, British Folk, Jazz, and Avant-garde. (thanks to Paul Lamere)

Musicians

4. Publishing Trends

In this map of publishing trends, Soybits diagrams the connections of people, formats, and products involved in the publishing industry. The Kindle, iPhone, and Android platforms form major intersections.

Publishing Trends

5. O’Reilly Books on Open Source

The O’Reilly books are known for covering a wide range of technologies and programming languages. This map diagrams the relationships among books on open-source technology.

O'Reilly books about open-source technologies

6. Web trends

As we mentioned in an earlier post, this subway map of the web by Information Architects is a fascinating exploration of popular Internet sites.

Web Trends Startpage shows websites as if they were the Tokyo subway

7. CMS Vendors

CMSWatch has published a diagram of popular content management technologies.

Content technology vendors

8. Subway map of world subway maps

Penguin’s book on Transit Maps of the World features a map of subway systems around the world. While the connections between cities seem somewhat arbitrary, we appreciate the style and spirit of this map.

Transit map of world transit maps

9. Cancer pathways

From Jonathan Stott’s thesis, Automatic Layout of Transit Maps, which we discussed previously.

Visualizing cancer pathways (Hahn and Weisberg, designed by Claudia Bentley)

10. Project plan

From Jonathan Stott’s thesis, Automatic Layout of Transit Maps, which we discussed previously.

Project plan

Bonus: Gotham City

Gotham City Rail has published their transit map. Wayne Central is, as expected, at the heart of the city.

Gotham City transit map

London trains and metro map, 1874

Jonathan Stott’s thesis Automatic Layout of Metro Maps Using Multicriteria Optimisation (PDF) is a comprehensive look at the state-of-the-art of automatic transit map generation. The premise: start with a geographic map of subway or transit stations and lines, then convert this map into an abstract yet informative representation of the transit system akin to those standard transit maps found in the world’s major cities. My aim in this post will be to provide a brief summary of Stott’s thesis.

Aesthetics of the transit map

Transit maps have existed for centuries. From their genesis, these maps have expressed a tendency towards simplicity—perhaps encouraging customers to visions of quick and simple travel in an age when trains and ferries were still novelties and their complexity not yet understood. The 1874 London trains map removes much of the city clutter for the sake of a simple, almost playful, representation of rail lines and stations (see above).

But the classic example in this space is Harry Beck’s 1933 map of the London Underground. Previous maps of the London Underground had maintained high geographic accuracy for line and terminal position. Beck’s landmark map codified rules for an abstract map: building from an inflexible grid independent of geography, adjusting the map’s scale to account for discrepencies in station density, and aligning transit lines along regular horizontal, vertical, or 45-degree angles. This codification produced a highly-readable diagrammatic map whose influence is strongly felt to the present day.

London Underground Map by Fred Stingemore, 1928

Harry Beck's classic 1933 London Underground map

Altogether, we might create the following list of guidelines for generating a stylized, diagrammatic transit map:

• Angle/shape generalization – Line segments should be restricted to vertical, horizontal, and 45-degree angles, and shapes should be made regular where possible.
• Scale generalization – The scale of the map should be compressed in sparse areas and expanded in dense areas to maintain readability.
• Color-coded lines – Separate transit lines should be colored with separate, distinguishable colors.
• Intelligent labels – Sans-serif font, proper capitalization, and a predominately horizontal orientation.
• Topographic metadata – Large rivers, parks, or coastlines should appear as background elements to provide context for the map.
• Stations as symbols – Continuity of symbol should be used for individual stations. Dots, hashes, or circles are all suitable. Transfer points between lines should be clearly indicated, usually with a symbol that stretches to touch all the relevant lines.

Syott’s thesis is concerned with codifying these guidelines and writing software that will automatically generate transit maps given an arbitrary, geographic transit map.

Hill-climbing

A large portion of this thesis is dedicated to exploring and critiquing existing algorithms to automatically draw transit maps. Transit maps are but one example of a directed graph, that is, a set of nodes connected by a set of directed edges that each contain a weight (such as distance, or travel time). Since graphs are well-studied in the computer science literature, much of the research on visualizing graphs can also be used when visualizing transit maps.

The first step: Aligning nodes to the grid

Stott’s basic approach is to define a set of criteria for what might be considered a good transit map (as we have outlined above), and then use a hill-climbing algorithm to find an optimal visualization. He breaks the visualization strategy into multiple stages, starting with grid alignment and finishing with label placement.

At each phase, software “guesses” what a good map might be, and subsequently an evaluation function is run on the resulting map. If the current guess is better than the previous guess, then the software uses the current guess as the basis from which to generate its next guess. An optimal solution is reached when the algorithm has achieved a high score that other guesses are unable to best—that is to say, the algorithm will have reached the peak of the hill.

A full overview of this algorithm is beyond the scope of this review, but can be found in section 3 of the thesis.

Engaging tourists and travelers

Stott tests his computer-generated maps (a wide variety of city transit maps were used) by having individuals compare his maps with both their geographic and official counterparts. The results are mostly positive. Travelers find the diagrammatic maps on the whole to be better than their geographically-oriented counterparts. Computer generated maps tend to score against official maps, and many in fact preferred Stott’s maps to the official ones. The automatically-generated maps rate especially well when travelers were asked to estimate an optimal route from one place to another within a transit system.

Geographically accurate map of the DC Metro

Official DC Metro map

Stott's automatically-generated DC Metro map

Our evaluation: Stott’s DC Metro map is in many ways a better map than Metro’s official version. Stott’s map possesses greater resemblance to actual geography. The transfer stations and lines in central DC, especially around Metro Center and Federal Triangle are greatly distorted on the official map, whereas Stott’s version better captures the geographic relationships of these stations and their neighbors. Having lived in DC for some time, we are well acquainted with the scattering of downtown stations (many with multiple entrances), and Stott’s map neatly captures a lot of what we know from personal experience.

Stott’s maps are far from perfect, but are easily classified as good starting points that have massive head-starts on the traditional approach. In the case of the DC map, the adjustments might be simple: a green swath for Rock Creek Park, blue lines for the Potomac and Anacostia Rivers, adjustments to line weight, and perhaps label font and alignment.

Repurposing the transit map metaphor

As we briefly explored in our last post that described a stylized subway map of the web, a transit map metaphor is potentially useful in any number of network visualization or graph visualization tasks. Stott summarizes some examples of transit map visualizations.

Visualizing cancer pathways (Hahn and Weisberg, designed by Claudia Bentley)

In one example, a “trains of thought” metaphor demonstrates logical strands of argument woven through the author’s PhD thesis. Another example shows the route of a project plan from start to finish, highlighting intersections where key components need to be brought together. The most striking visualization is a map of cancer pathways (above).

Resources

• Download Jonathan Stott’s complete thesis (PDF, 366 pages. Oct 2008).
• Helpful distortion at NYC and London subway maps at 37Signals
• Edward Tufte on the London Underground map

Web Trends Startpage shows websites as if they were the Tokyo subway

Following up on our earlier post about the Tokyo subway map, the Information Architects Web Trends Startpage is a clever mashup of the subway map metaphor with the standard start-page. Authors of this map have manually categorized and located some of the web’s most popular sites into color-coded lines and stations. Especially interesting is the “inner loop” of sites that represent the authors’ favorite sites.

We could compare this site to other start pages such as My Yahoo! or iGoogle, or the built-in thumbnail views provided by Google Chrome and Safari 4. Against these competitors, the IA Startpage wins points for the sheer quantity of sites it can display in one screen, and its memorable composition makes muscle-memory point-and-click painless. On the other hand, the IA Startpage is probably not useful on a daily basis for most people, since it is tailored specifically to the needs of its creators.

On the whole we greatly appreciate this artistic portrait of the web and its attempt to bring order to a chaotic landscape. IA provides an interactive version of this map with clickable links, and also has PDF and hi-res JPEG formats available for download, and a poster version of the map is for sale.

There exist a number of subway-map algorithms that have the potential capacity to translate a arbitrary graph structure—such as the web—into a visual map. Maybe the next iterations of this map will include such algorithms, allowing users to generate custom maps of their personal stomping grounds.

note: XKCD has also published a map of the web focused on online communities.

Tokyo subway map

The very elegant Tokyo subway map. Cities are inherently complex, and transit-map publishers face a dizzying set of challenges if they wish to produce a readable but accurate representation of a transit system. The designers of this map deserve particular credit for organizing and clarifying what is necessarily a sprawling collection of subway lines and stations.

At the heart of this map is the Imperial Palace. The inherent structure of the subway system lends itself well to this concentric arrangement, since the tracks themselves circle the palace, but I would also suppose that we could read into this a set of cultural and political assumptions about the importance of this central place and its inhabitants.

This map correlates strongly to how we think of the city. Much research has suggested that human minds think of locations as landmarks, and relationships between locations as simple paths that lack traditional metrics of Euclidean distance or travel time. By translating the magnitude and centrality of the Palace-as-landmark into an appropriate visual representation, we can immediately understand our place on the map as it relates to our understanding of real-world space.

More information, including maps from Osaka and Seoul.