Jun 27, 2007
Eddie Elliott's Cab Spots
I've been talking a bit with Eddie Elliott, local designer/technologist and all-around raconteur whose beautiful digitial work predates the web, about Cabspotting lately. We (Stamen) keep meaning to get back to the project and do some new investigation, but something else (i.e. paid work that we like to do) keeps getting in the way. Eddie's stepped into this gap, and been using the Cabspotting API to produce some really stunning work. You can read more about it at a page he made for it at http://cabs.lightmoves.net/—in particular I like the calculations of the center of gravity over time, and he's pointed out an error in how we're plotting latitude vs. longitude (ouch)—but it's the high-res long-term point maps of San Francisco that make me the happiest, and provide some interesting new ways to look at the city through the data it throws off.
So first of all Eddie's using dots instead of lines, which makes sense over longer periods of time. This image shows 5,744,623 cab-spots (what Eddie calls individual cab GPS locations), recorded over the course of 31 days (44,640 minutes) from March 21st to April 21st, 2007, and it's just lovely:
I like this well enough to think that we might want to consider using this method for generating the background tiles on Cabspotting. The lines are cool, and demonstrate flow really well - but they're a bit messy. We'd have to figure out how long to show dots for; I don't think four hours (what we use now) would be enough to show the city, but it'd definitely show the changes in use patterns over time.
But oh! The detail in this thing! The freeways, deliciously outlined, medians completely clear as day, wispy circles of freeway off-ramps... and the fuzzy bits at the corners, the places where the seams are turn out, again, to draw my eyes in, and prompt me to ask more questions.
Seams, again
For example, in the image above, the spots generally tend to stay within the boundaries of the streets, and are crisp and detailed. The image below is focused on downtown, though, and is wierdly fuzzy, with points all over the place. Obviously the cabs aren't driving through buildings willy-nilly. Eddie thinks (I do too) that this is because downtown buildings are so high and close together that GPS signals can't make it down to the ground with very much accuracy, bounce around off the glass and steel, and give "bad" results. Fair enough; downtown's not so accurate. But what it means in terms of urban area chartings, where cabs tend to stay in very narrow street slots, is that you can use a visualization like this to tell immediately where the high buildings are by the degree of fuzziness in the map, and if you mapped the height of the buildings over this image, they'd probably overlap prety much one-to-one. Which is nice, right? And wierdly, this kind of fuzziness looks almost exactly like an out-of-focus photo, so here's another real-world to data-visualization metaphor to go along with hearts beating and ice melting.
Even better, there's a bit of fuzziness in the upper-left hand corner of the image that seems slightly out of place - it's not quite within the downtown area, and the buildings over there are tall, but not super tall. So why the fuzz? A quick overlay of a google map (you see where this is going) shows that this is the Fairmont Hotel, at the top of California Street:
One click on google's street view later, and we get a view of the Fairmont's extensive portico area, where (of course) lots of people get into and pick up taxis:
So: you and I live in a world where normal people can look at complex data visualizations of urban environments, notice anomalies in the display, go to the web to find information about where that place is, and then make pretty good guesses as to why the data is showing up the way it is. It needs smart people with some non-trivial technical knowhow to make these particular views on it possible, sure. But once that's done, there's a very quick path available to free information that can be used to reinforce, disprove, or generally poke at the way that the world is, and why it is that way, and it's fluid and easy and you can start asking real questions very quickly.
I think this is a new thing.
At the very least I found myself immediately asking why the display looked the way it did, looking to google maps to see if I was right, and then returning to the maps that Eddie made to see if I was right. For example, Eddie mentions that "It’s not clear what is going on with the slight swelling at street intersections in lower-rise areas of town," and this had me puzzled for a while too, until Tom suggested that this is probably directly related to the speed that the cabs are going: there's still a little bit of error in all the GPS signals, and if a cab stays at an intersection for a while, those little errors eventually add up and form a swelling at those intersections. On average, the ones where taxis are moving will have less of this kind of swelling, so the streets and freeways generally look clean and crisp.
Start with the visualization, figure out what it tells you about the world, ask new questions, go back to the visualization. Repeat.
But wait, there's more
A similar situation, right by our studio: two lighter blobs, right at the corners of a block (in the middle of the picture below):
Google maps overlay; they're at the corners of South Van Ness and 16th and 17th Streets, respectively (the green arrow is our studio):
Where there happens to be a combination gas station/car wash on one corner:
Gas Station: South Van Ness Ave & 16th St
And another gas station on the other:
Gas Station: South Van Ness & 17th Street
There's lots more of this kind of enjoyment to be had in the high res image, the rest of the project at http://cabs.lightmoves.net/, and on Eddie's site. Enjoy.
roadvis