Wednesday, May 7, 2014

The Joys of Georeferencing

Georeferencing can be a lot of fun.  It can also be a pain.  It, like all GIS processes, requires a lot of patience and problem solving.
But first, what is Georeferencing?  Wikipedia defines it as:  to associate something with locations in physical space.
Simply put it is uploading a aerial photo that was taken being digital data was a thing and referencing points that have stayed the same so the old maps can be accessed digitally.  For instance it would be possible to use Google Earth to look at the City of Austin in 1987.

Sounds easy right?   Not Quite.  There are three major problems with it, development, technology, and margin of error.

How much development could have taken place in the last 20 years?  Surely not enough to make it indistinguishable.

OK.  So a few places are no longer recognizable BUT, there are some landmarks that are sill usable (hence being able to find what those houses used to be).  In another 20 years at Austins current rate of growth however it might indeed be problematic).

There are problems with both the 1987 method of taking a picture of the region and with ArcMap's algorithm for calculating what goes where.   

In 1987 the US Government would take photos of cities in an attempt to track their development and in case of a natural disaster.  These images were taken by a plane that would snap a photo with a wide focused lens to capture the entire area.  Then these photos would be matched next to each other and a map would be completed.  

This has problems though as the edges of the pictures are not taken straight while digital photography allows us to do the same thing but every point is taken straight.

Digital Technology is not perfect though; if 3 points are georeferenced in a like the "collinear" aspect of the points throws off the algorithm that ArcMap uses and the picture is skewed.  This means that Georeferencers will have to selective choose what points to reference.  

The final problem with Georeferencing is human error.  While it may seem like an electric pole hasn't moved in 20 years it is very possible it has.  Or the shore line of Town Lake, or how some roads have moved 20 or so feet with new construction.  More Over, sometimes I have to leave a large Margin of Error according to ArcMap even though I could reduce it but have a less accurate map.  Georeferencing is sort of an Art.  The more you do it the better you get at it; even if the computer thinks you're wrong.

In order to have a good map there are a few little tricks that can help
  1. Fit the map to the display once the displays already looks like the map.  The more accurate you can get it the better.  Some of my maps were georeferenced in 4-5 minutes because it fit nicely and hadnt changed too much.  
  2. Select points that are NOT collinear.  The center of the map is the most true so putting a point there is a good start and around the corners.  
  3. Five reference points is really enough.  
    1. Whether judging by ArcMaps calculated degree of error or eyeballing it more than five points really is not needed.  Especially with data like the government supplied photos of Austin.
    2. Be practical.  You can get a degree of error down to 2 feet from 15 feet if you want to and are diligent about referencing.  But for the purpose these maps are being used for there would simply be no reason so.
  4. Georeference chimneys.  Any landmark that doesnt move is good to reference.  Most of my most accurate maps came from referencing chimneys.  

1 comment:

  1. Thats really interesting how the two pictures of the street of houses are similar from a far but very different up close. The way the houses sit in the first picture is very slanted versus the second picture. I wonder if roads have this off set imaging in 1987 because that could cause some land issues such as if I-35 slightly appeared more east which could cause a lot of confusion.