All 5 books, Edward Tufte paperback $180
All 5 clothbound books, autographed by ET $280
Visual Display of Quantitative Information
Seeing With Fresh Eyescatalog + shopping cart
Edward Tufte e-books
Immediate download to any computer:
Visual and Statistical Thinking $5
The Cognitive Style of Powerpoint $5
Seeing Around + Feynman Diagrams $5
Data Analysis for Politics and Policy $9catalog + shopping cart
New ET Book
Seeing with Fresh Eyes:catalog + shopping cart
Meaning, Space, Data, Truth
All 5 books + 4-hour ET online video course, keyed to the 5 books.
It will take a little while to download even with a fast connection. It is amazing.
[Link updated February 2005]
-- Edward Tufte
It is amazing. However, by making the planes many, many, times larger than their scale size, it gives one the impression that the sky is full of planes, when that is clearly not the case.
-- Mark Kohler (email)
As with the space-debris drawing (see this thread for additional discussion), the planes, although not to scale, are required to locate the simulated plane on the map. A mark which was to scale would likely be too small to display on a computer screen.
-- Scott Zetlan (email)
The animation is fascinating. It would be interesting to compare it to a day when a storm steals a common flight path or closes a hub airport for a few hours.
Rather than the relatively small aircraft, it is the safety buffers around aircraft that fill the skies. The required horizontal separation between aircraft can be between 5 and 30 nautical miles. The vertical separation is a few thousand feet. That brings the animated flights much closer to scale.
This page from the FAA on "Longitudinal Separation" describes the separation rules based on aircraft speed, same/converging/crossing courses, ascent/descent, and navigation equipment. The text is a bit dull, but the graphs make the rules very clear.
[Link updated February 2005]
-- David A. Nash (email)
I couldn't see this particular movie, but I have seen FACET movies before. Nice way to follow the sun. You don't want to be a small plane following closely behind a 747. The wake vortices from the wingtips are potent for many miles.
-- Rick Kwan (email)
This was fascinating indeed. Besides reminding me of the ants that presently have invaded my kitchen it struck me the scale issue mentioned earlier and the uneven motion and sudden apperance/disapperance of the jet icons was interfering with some of the more subtle information that this kind of animation could reveal.
Other than the correlation of time of day (represented only by a digital counter) with quantity of flights and the relative density of air travel in the east as compared to the west there was little more that this animation could offer. I realize that this represents a huge amount of information being crunched and is probably a landmark in representing this kind of data but I can only compare it to the work Benjamin Fry is doing with words (http://acg.media.mit.edu/people/fry/valence/index.html) and I can imagine this sort of metaphor for air travel appearing presently.
This animation is a first step in that direction.
[Link updated February 2005]
-- David Bishop (email)
I saw similar animations at the Hacker's conference last year. We saw both a normal day, like the above link and an animation of the skies clearing directly after the Sept. 11 hijackings. It was stunning. Every single plane disappeared from the screen.
-- David A. Smith (email)
Impressive. One thing I found confusing though, was the display of UTC Time. Although flights worldwide sync to UTC it would have been helpful to be able to select a local time zone instead of having to do a mental conversion to figure out when the busy/quiet times are. Maybe this is selectable from the menu, but we cannot tell from this animation.
-- Craig Pickering (email)
The local times for all the displayed time zones could be effectively indicated by showing the passing daylight and darkness the way a geochron does.
-- David A. Nash (email)
Planes coming up the eastern seaboard appear to have an odd crab about them. Most of the planes in the graphic seem to be flying with their nose (longitudinal axis) oriented along their track, but the planes coming up the eastern seaboard are look to be crabbed maybe 20 degrees to the right of their track. If there was a jet stream that day, one might expect them to be crabbed 20 degrees in the opposite direction. And none of the other aircraft appear to be crabbed at all.
-- Jerry Asher (email)
As one of our kindly contributors mentioned, there is a similar animation for the grounding of all planes on 9/11. Can someone please provide the URL?
-- Edward Tufte
NASA has an "Air Traffic Shutdown Study" page of links to contrail studies that
includes GIF animations of hourly air traffic density for a normal day and for September 11, 2001.
The animations are similar to this graph:
Average daily frequncy of commercial airplanes over Continental USA
NASA's page of links: http://www-pm.larc.nasa.gov/sass/airtraffic_shutdown.html
normal day animation: http://www-pm.larc.nasa.gov/sass/airtraffic_shutdown_data/010903_anim.gif
9/11/01 animation: http://www-pm.larc.nasa.gov/sass/airtraffic_shutdown_data/010911_anim.gif
-- David A. Nash (email)
There are similar animations on ADP's web site (the company managing Paris airports), it requires Flash and the interface is in French: http://query.adp.fr/adpv3/fr/OPA/home.htm
-- Alex Drahon (email)
Another interesting air traffic animation can be found at http://www.aaronkoblin.com/work/faa/. There's a good deal of art here, but it's data driven and quite interesting. Estes
At first glance, I assumed the plane icons were flying over a map of the U.S., but of course that's not quite the case. It's not until things get really dense, around UTC 22:00, that you can see the distinct outlines of the east coast, Florida, Texas, California. Anybody know what the boundaries are in that underlying map? They seem to extend pretty far offshore.
Interesting animation of FedEX planes avoiding thunderstorm (via jorn at robotwisdom). Accompanied by some ridiculous music, which should be turned down before viewing.
-- Edward Tufte
I'm curious, is there a name for the type of visualization shown in David Nash's post? I've always naively referred to it as a "heat map", but I'm sure that's not the correct term. Would anyone know what the correct term is? Further, would anyone be able to recommend software that aids in creating this type of visualization?
MATLAB can produce these and refers to them as a "filled contour" plot (CONTOURF function), many other packages must also be able to do this. It is the same technique as a topographic map so I suppose one could call it that, too....
-- Dan Goldner (email)
Another excellent example: the Tazman Curve
Three variables are plotted: time, space (latitude), and atmospheric carbon dioxide concentration. Variability at different observing stations, cyclic yearly variation, and the longer-term increase are all made visible in this quicktime movie of a pair of 2-d plots. It is the most elegant, transparent presentation of this data set that I have seen.
-- Alexey Merz (email)
The CO2 movie is interesting but I found the inset and the animated pie chart distracting, especially because the inset x-axis (time) is different to the main axis (latitude). I would suggest that after each year's data is animated an annual average line be added to the animation to emphasize the trend with respect to the rest of the animation. A crude illustration is shown below - the data is screen shots grabben approximately every two years. The default playback in my viewer was rather fast though I know I can slow or pause the animation. Alternatively, a series of small multiples could be used with error bars to show spread at each station.
-- Simon Shutter (email)
Belated answer to Jesse -
The mapped area is indeed the continental US (CONUS in the lingo). The pink boundaries outline the areas of control of Air Route Traffic Control Centers (ARTCCs), which control aircraft enroute. The red boundaries outline sectors (and here, I believe, some merged sectors) within the ARTCCs, which define the scope of control of a team of (usually 2) controllers.
One place you can find the identities of the Centers is
The codes are shown on the map, and the cities in which they are located are in the pulldown list under "Centers". There are a few oddities - Chicago Center is ZAU because the actual site is Aurora IL, Atlanta Center is ZTL, Salt Lake City is ZLC - but otherwise they match up fairly obviously.
I hope this answers your question.
I'm sure the readers would find much to comment on on the above site; there are many options for information to be overlaid on the map. Not much explanation, though. Very quickly, Special Use Airspace is mostly that used by the military (MOA = Military Operations Area, MTR = Miltiary Training Route). Some is shared with civil air traffic, some is open to civil traffic when not actively in use by the military ("hot"). SUA is defined in three spatial dimensions and sometimes in time as well, in http://www.faa.gov/airports_airtraffic/air_traffic/publications/at_orders/media/SUA.pdf
-- Ken Schulz (email)
A belated reply to Simon and Alexey:
Watching the movie and viewing the screenshot summary should provoke questions for any viewer of those data. The obvious conclusion is that as time goes on, CO2 is going up. But is there more than this?
The first thing to note is the support set of the data: almost all the points are in the Northern Hemisphere; the latitudes with an "S" are remarkably thin when it comes to data concentration.
The second obvious thing is the variation: it is wild and obviously cyclic in viewing the Northern Hemisphere; the lack of seasonality in the Southern Hemisphere data, eventhough there are few of them, is striking, perhaps unnerving, unsettling.
So, of course, there are more stations in the Northern, hence more data support. So, let's look at the data. Also, most of the south is covered by water; can that make a difference?
I went to the site listed on Simon's screen shot summary: it was forbidden for me, and also changed its name. Moving up the directory structure, I was able to get to http://www.esrl.noaa.gov/gmd/ccgg/globalview and found a color version of the movie that Alexey presented.
The descriptor on the page says: "GLOBALVIEW data products are designed to enhance the spatial and temporal distribution of atmospheric observations of CO2, CH4 and other related atmospheric measurements. GLOBALVIEW products are specifically intended as tools for use in carbon cycle modeling studies. A GLOBALVIEW data product is derived from measurements but contains no actual data. [sic]" Furthermore, a footnote in the movie says "Extended Records". I wonder what that means?
So, these are a "data product" "derived from measurements" but containing "no actual data"?!? This is starting to sound like "processed cheese food".
The website references a paper by Masarie and Tans ("Extension and integration of atmospheric carbon dioxide data into a globally consistent measurement record", J Geophysical Research, 1995). The paper describes how data can be wrangled to produce a "global atmospheric CO2 database that should aid modelers in the pursuit of a better understanding of the global carbon budget."
The data, therefore, are not the data at all; they are summaries, fit to assumptions. The Benchmark Trend Method "extends CO2 site records using information from site climatologies and a benchmark trend derived from a single network site that is representative of the global trend in atmospheric CO2". Wow.
There are more movies than just the one for CO2. There are also movies for CH4 (that's methane, right?), CO (carbon monoxide, famed in driver's ed as "odorless, tasteless, colorless"), and CO2C13 (I have no idea what that is). The movies are all similar: little support in the South, seasonality, general rise over time.
The CH4 one is interesting: it is S-shaped, with the North substantially higher than the South. CO does this too, but the CH4 one is really dramatic. So is it just that the North produces more methane? Or is there more going on here? What does this mean? What information does it contain about the data product production process itself?
In 2006, I took part in an American Statistical Association Data Expo which looked at atmospheric data from Jan 1995 to Dec 2000. We had monthly values for temp and ozone and clouds and such over a 24x24 grid of points that ran from approximately 36N to 21S and was centered over the Americas. After all was said and done, there was one consistent finding across all the entries in the Expo: something funky happened to the data between May and June 1998.
Fast-forward to today when Rafe is looking at the CH4 data movie. Starting in 1994, the Southern hemisphere trend line bounces up and down annually from about 1680 to about 1730 prior to about 1998. From 1998 to the end of the data stream, the line bounces from about 1710 to about 1750. The change is not smooth and gradual but takes place at the same time as the change seen in the Data Expo data.
Something about the model changed in 1998. What are the consequences for that?
I'm not a climatologist; I'm a statistician. I'm not going to say that the Masarie and Tans paper is wrong or even misleading. It is what it is: a method to extrapolate data to create a data product. It appears quite slick actually, with fancy smoothing and some wonderful math.
But the product is not the data. The product is the data run through a filter of assumptions. We need to be careful about the consequences of assumptions, certainly when they are applied to such a heated topic (pun intended).
Ok, that's that.
-- rafe donahue (email)