Friday, July 27, 2012

Tweet Dissection for Nerds

Twitter, in addition to being a guilty time-suck as I stare into my phone at the quips of strangers while by beloved children pull on my arm for attention including eye contact, happens to be a great confluence of data dimensions for viz nerds.

Like what?  Well, there is a geographic component (often), a time component, the textual content of the message, and the ontological nature of retweets.  Gold!  While each of these are valuable dimensions for illustrating data, the combination is so much sweeter than the sum of their parts.

Here's a recent more example using tweets containing the word, "torch."  I've zoomed in on London to find a pretty well defined cluster of folks noting the passing of the Olympic torch on its way to the opening ceremony.  The intersection of space (lat long) with time (tweet timestamp) shows a clear trending of a phenomenon (keyword) with more insight than those three can independently (check out this other, probably tighter, example).  Even just a basic handling of those dimensions, like in the following snapshots...

Thumbs scramble to tweet descriptions of the "torch" passing through the streets of suburban London.  A distinct geographic pattern appears.  Using relative age to color the tweets reveals a vector from west to east and a relative frequency through the timeline (you may notice my local EST offset as I observe from across the pond here in Lansing, Michigan).

 Later on, folks describe the sight of the torch chugging down the Thames.  Another benefit of the concurrent view of the same data across multiple visual dimensions in their linkage through interaction.  As I mouse around in one, the counterpart pings in the other panels providing a thread across views.

Here's an illustration of how interaction with a data element causes instances in other visual dimensions to pulse, providing a link between the three.

Thursday, July 26, 2012

Interactive Tornado Tracks Map is Live

or jump right in, here:

So lots of questions surrounding the original tornado tracks map sent me back to the data a few times for some answers.  Turns out, the very best way to interrogate the tornado tracks data set was to drop it into our viz software, where I could slice and dice it on the fly (rather than via a chain of queries in a GIS or as a universe of pivot tables in Excel) and see the results right then and there.

We plopped the source data (which is freely available at and goes from 1950 through 2011) into SQL and put lots of criteria for filtering right in the web page so we/you could visualize whatever combination sounds interesting/useful.

Anyways, have at it...but be advised that it is a pretty big dataset so when you ask for every tornado in the past 61 years, it could take a moment.  But it will be worth it!

Here are some things I was curious about, in no particular order.  If you find something interesting, by all means share it and include the #tornadotracks tag in Twitter so we can see it too!

Every recorded tornado since 1950.

The terrible 2011 season.

A close-up of Joplin, showing how frequently tornadoes have passed through over the past 61 years.  The track of 2011 is highlighted at right.

1974 was another particularly bad year for tornadoes.  The rash moving north through the Midwest actually occurred over one day...

The night of April 3, 1974.

Tornadoes in New York?  Not Manhattan, it turns out.

The most expensive (in terms of property damage) tornadoes since 1950.  Here is every tornado that caused more than $50 million in property damage.

Every known F5 (or EF5) since 1950.

Recorded Hawaiian tornadoes since 1950.

You may have seen this before, but it is an example of creating a set of related images in the website and stitching them together into a more narrative infographic.  There images were made using this website.  We did something similar but for time slices and dropped the images into a couple YouTube movies, you can find here.

Thursday, July 19, 2012

Major Fires Flip-Book Style

So map data is great and time data is great.  And each of those can be interesting on their own.  But when you show geographic dispersion through time (or time dispersion within geographic subsets), that movement can go a long way toward facilitating understanding (or at least dimensionally richer insight).
Anyways, here's the decade or so of fire data from the recent Major Fires Since 2001 map, veneered off into discrete years and clunkily flipped through in an animated GIF.
(Who ever thought the animated GIFs would make a comeback?  The same people who decided plugins didn't belong in mobile, I suppose.  Oh Flash, how I miss thee...)
I should also mention that the reason fires are showing up beyond the US borders is because the data source provides instances within an overfetch buffer.

Major fires since 2001.  It is pretty big and may take a moment to fully load.

Or, if you rather, here they are via a not-so-small small multiple.  Click for the unreasonably tall version...

Friday, July 13, 2012

Major Fires Since 2001

There is a sweet sweet machine orbiting the earth soaking up it's emitted photons at various wavelengths.  One of the benefits of this is it's ability to, with reasonable confidence, pinpoint the location and intensity of a "thermal anomaly," or fire. So, armed with over a decade of these events, you can make a map of literal hot spots.
As it happens, the same satellite that scans for fires also provided the basemap image.  I heart NASA.

Anyways, each dot represents a moment of pretty extreme heat, down to the one square kilometer level (I only retained fires greater than 100KW MW and of those only fires that the system was more than 50% confident of).  They've been colored and scaled by "units" of the typical American nuclear power plant's summertime capacity to provide some sort of baseline of the fires' magnitude.
There are a couple temporal charts in there, too.  The seasonal curve I would expect, but the overall upwards trend was interesting (and 2012 is only half through).  Is it related to a lag-offset El Niño or La Niña effect?
It's almost like the universe balancing itself out, considering the general density of historical tornadoes.  Except poor Florida gets a heaping pile of both.  Can there be a fire tornado?  If so, look out, Tampa.  Also, what's cooking in the south Mississippi basin?  And check out the firewall of the Cascades -You shall not pass!

Here's an animated version of this map, isolating each year's fires and highlighting their aggregated intensities on the chart.


Other Sort-Of Related Maps
If this map is interesting to you, you might be amenable to plunging down the rabbit hole of related visualizations...