There are a number of sites that describe the way these coordinate systems work in theory, but this single Powerpoint presentation combines the theory with the practical side better than any other single resource I've found on the topic:

Navigating the Disaster

There's a sustained conversation about using grid systems vs. decimal degrees (lat,lon points) for dealing with on-the-ground emergencies (grids apparently seriously reduced human error when finding points), and a thorough test case for grids during the Katrina disaster response. The worst part of the test? That the 9th Ward in New Orleans was, of course, cut in two by two UTM zones.

Here are pictures from the ppt:


UTM is apparently lots of little mini-projections so that we have equal area (?) blocks to slap onto a map that seem to keep their distance at scale. This causes some real issues where zones overlap, however, which receive more illustrative attention within the Powerpoint linked to, above.

Here, apparently the pictured overlap was too much for the FEMA GIS Coordinator to handle, and they went from the Military Grid Reference System (MGRS) to decimal degrees instead.

What should have been done with the overlap is below. Note the dark line cutting through what's ultimately hexagonal transitional tiles. In FEMA's defense, you admittedly might still have to occasionally figure out where 15 is in 16 and vice versa if someone's GPS reported they were still in one zone after stepping past that bold line. That's a task that, unfortunately, would take especially competent responders (FEMA office and everyone in the field).



So New Orleans' 9th Ward was literally a worst-case scenario for the MGRS. Why, then, was using lat,lon potentially a very bad idea?

The presentation here argues that the potential benefits for grid use (at least where there isn't overlap, or where overlap can be dealt with properly) appear immense. I'm going to include two more pictures, one showing grids vs. lat,lon "blunders" (red is over 150 meters, green under 15) and another picturing the difference.





Those last two are taken from a publication from the Powerpoint's author:
Terry, 1994, Photogrammetric Engineering & Remote Sensing, Vol 63, No. 4, Apr 1997, p.381 - 383.

Now, again, the Powerpoint author doesn't relate those findings to using the grids in multizone areas like the 9th Ward of New Orleans. I imagine those numbers would change quite a bit. At least lat, lon is unambiguous -- there's no mental calculation when traversing arbitrarily placed bold lines (arbitrary in that the MGRS could have started a mile father east and spared at least New Orleans the trouble).

(Aside: If Sweden can have its own UTM zone exception, it might be worth putting one in for New Orleans as well. But even so, it's hard to think that a little preventative education wouldn't've allowed grids to work well -- likely significantly better than lat,lon -- even in this worst case.)

What's the connection to programming? I inherited code with a horrible, inaccurate, buggy UTM to MGRS translator, and I'm trying to make sure I'm educated enough to test the at least excellent looking new library our team's found to make sure it's a solid improvement. That link also has excellent descriptions of what's going on with MGRS and a walkthrough of his code.

Labels: ,