Tuesday, June 2, 2015

Migration

This is the annual migration pattern of the Rose-Breasted Grosbeak. A pretty fantastic voyage for a little thing -which they will undertake up to twelve times in their lives.  Each sighting (from the Cornell Lab of Ornithology's eBird project) is a little spotlight that reveals the underlying satellite imagery for that month. When seen together, the wonderful emergent properties of a species' migration illuminates a true bird's eye perspective of movement and structure on a continental-scale.

The big seasonal migration of a little bird. Each sighting reveals a spotlight of the underlying satellite image for that month. All together, they illuminate an annual ritual of chasing greenness. Click to open the full-blown version (could take a moment to fully load the animated gif).

Animal migration just blows my mind. The intrinsic commands wired into a bundle of brains and nerves such that they and all their pals are totally in the know and confident of the plan. Except that there is no 'plan.' It also makes me wonder how much of what I do and think in a day is intentional and novel, and how much is me holding a little point of light largely unaware of the larger structure I am a part of.  In many ways I'm not so different from that little bird (though less well-traveled).

Why the Spotlight Style?
When a creature like the Rose-Breasted Grosbeak feels the inner tuggings of the migratory urge, they are still just one element of a larger system. Any one of these birds only get to see what they themselves can see, and while their perspective might perch higher than ours, they have no map and no external device telling them which direction to fly and what the distance.  Their individual view is no bigger than one pinpoint of a spotlight at the continental scale, but they confidently charge ahead. I thought a visual method that revealed visited locations (rather than painting over them) would suit the phenomenon well, hopefully making the viewing a little more personal.

A month by month small multiple of the migration. Now your eyeballs can more efficiently observe the phenomenon rather than be held hostage by the beautiful tyranny of an animated sequence. Click to blow out the full resolution version.

Why?
This was a collaboration with a friend/cohort from Central Michigan University's Geography Department, David Patton. Dave is an actual birder (photographer, guitarist, scientist, gentleman poet), while I am...other things.  Also, special thanks to Daniel Huffman and Joshua Stevens, who both provided terrific cartographic advice along the way.
Dave has done work with the eBird data in the past, with his class, so he had already undertaken the painful process of downloading, parsing, and filtering a multi-gigabyte flat file with five years of observations. He noticed that the Rose-Breasted Grosbeak had a particularly impressive migration pattern.  I wondered if a subtractive, rather than additive, approach might be an interesting way to show this migration. Think scratch-off rather than dots.
[Insider scoop: The seed of this idea came from a concept we used years ago for an unnamed counter-terrorism division of a metropolitan police force.  We used a reverse-heatmap to show, etch-a-sketch style, the buildup of patrol unit movements over time.]

An early test of the scratch-off masking concept. Rather than covering the sighting locations, the sightings became, via masking, the only way to see the locations. You see what the bird sees, and that's it.

How?
First Look
Binning up large amounts of lat/long data (and parsing out dates into months, days, years, etc.) is most quickly and easily done in Excel. Yes, Excel. It is a handy first-pass look at the data, and it takes about two minutes.

An Excel pivot table of sightings by lat/long. You can filter by month, as well, to get a sense of movement over time with just a few clicks.

Here is an initial look at the strength of the migration -also a pivot table in Excel. The Y axis corresponds to latitude (North-Southiness), and the X axis breaks out sightings per month.

Satellite Imagery
The satellite imagery for each month was downloaded from the great NASA Visible Earth resource. They were projected into Lambert Conformal Conic (I tried other more ridiculous perspectives with a stronger horizon to try to echo the feel of looking down from above but good old Lambert was a way better balance of readability and perspective).

Masking
We separated out the sighting points by month in ArcMap, and exported them as PNGs with a transparent background.  Then, in Fireworks, they get a Gaussian glow and are used as a masking layer to reveal the satellite imagery.

Get that month's point cloud and satellite image, apply a Gaussian glow to the points, mask image by points.

Then we calculated the spatial mean of the sightings per month, and masked-out a state lines reference centered over that average location (for each month).

The spatial mean of each month's sightings provide the center of a faint elliptically masked-out state reference.

The combination of these layers, plus a very faint version of the satellite image to provide some context, formed a map for each month of the migration.  They were then stitched together into an animated gif, and small multiple (those things above the fold).

Putting all the layers together.

Charting
While the North/South shift in sightings was pretty obvious in the map animation, itself, I like the behind-the scenes view of proportionality I saw in Excel in the discovery process. So I added it to the map. Why hog all of that chart fun?

The "proportional sightings by latitude" charts were quick bar charts that I copied out of Excel and cleaned up via magic wand waving in Fireworks.

Data
Bird sightings
Dave wrote to the folks at eBird and got a mega-huge file.  Here is their website:
http://ebird.org/content/ebird/

Satellite Imagery
I am a devoted and compulsive user of the cloud-free mosaics from NASA's Visible Earth, made available to a world full of nerds.
http://visibleearth.nasa.gov/view_cat.php?categoryID=1484

State Reference
The US, Canadian, and Mexican state (province) linework came from the precise and generous folks at Natural Earth.
http://www.naturalearthdata.com/downloads/10m-cultural-vectors/

I hope you like it. It was a fun excuse to work with an old friend and a chance to re-visit an old visualization trick with the benefit of perspective; the fact that I am happy with the results is just gravy.


View John Nelson's LinkedIn profile View John Nelson's profile

No comments:

Post a Comment