Map Projections, A Necessary Evil

Lately, I’ve been spending a lot of time working with map projections. This isn’t a favorite pastime, but a necessary evil. Map projections don’t produce discoveries, improve security, or really do anything in and of themselves. They’re tools, intermediate steps in the process of the real work. They’re a lot like file formats. What I want is the data, but I have to deal with the formats to get to the content. And all of that is fine. There isn’t a perfect map projection or file format any more than there is a perfect language. There are good reasons for them all.

But most of that is cold comfort after hours of trying to figure out how the data are arranged relative to “standard” map projections and data sets. Alaska is still drifting hundreds of miles off course and Madagascar still looks more like a rounding error. If it were easy, everyone would be doing it, but I would prefer the data behave and let me work on the more difficult, and more important, scientific pieces. I have finally forced things to line up at this point, but I thought it might be useful to list the important terminology of geospatial data location as it has occurred to me through the process.

1. Geolocated – This is more of a marketing term than a scientific one. There is no real meaning or certification behind it; you get nothing. It’s like saying a hot dog is 100% beef. So are ribeye steaks, but that doesn’t make them in any way comparable. It is used primarily for describing the view of a non-quantitative piece of information or imagery. “Picture of Africa”. Okay, fine. From an airplane? From the ISS? A subset of a meteorological satellite view? A ground-based photo? Maybe it even has coordinates, like “40 N, 105 W”. I still don’t know the scope or nature of the date. Geolocation is a start, but raises more questions than it answers.
2. Georeferenced or Projected – This is usually much more useful. We know location of the data, with specs for the coordinate system (lat/lon, or E/N UTM, etc.), units (degrees, feet, meters, etc.) and what map projection was used and with what parameters such as datums, ellipsoids, latitude of true scale, and more. We might not like how the data are arranged, but at least we know what we have.
3. Orthorectified or RPC – These two are not equal but often get tangled. If the data account for the exact geometry of the surface and how it has been traced from there to the sensor, through the processing, and on to the screen in a standardized map projection, we call it orthorectified. If the data use a mathematical approximation to do this instead of a geometric model, the process is called “RPC” or “Rational Polynomial Coefficient”. RPCs are usually faster and the techniques for producing and using them have been improved so much that accuracy differences are inconsequential for all but the most rugged terrain (think the Himalaya or Hindu Kush) or most unusual map projections. Unfortunately this has lead to many people using the terms “ortho” and “RPC” interchangeably, which is misleading at best.
4. IGM or GLT – Input Geometry Models and Geometric (Geographic) Look up Tables skip the whole projection problem and substitute a data storage problem instead. Rather than have the data arranged in a well-describe map projection, the data have actual map coordinates stored for each location and users can use and project the data however they like. This introduces storage problems if x and y coordinates must be stored for each pixel in a 5,000 x 5,000 pixel image, opens up the user to projection and resampling issues, and can greatly increase processing time. However, it also eliminates a lot of early data processing and let’s users choose the final state of the data.

None of these standards are “the best” for storing or distributing data. “The Best”, as is so often the case,has more to do with the end goal and who’s involved than anything else. The important part is to be sure what you’re starting with and choose the final state based on what you’re trying to accomplish. Monitoring polar sea ice changes? Go for Equal Area Polar Stereographic. Looking to match historic survey data? Maybe State Plane is the way to go. I can’t think of a Goode reason to use Interrupted, but I suppose it’s an option, too. Me? I’m going to see if I can’t make it a little easier for my colleague to get their data. Besides, neither Alaska nor Madagascar should ever look like that. It’s unseemly.