Appropriate Projections for Cartograms
Posted 05 October 2007 - 03:45 PM
I am creating a succession of 12 world cartograms (1993-2005) that distort area based on amount of crude oil (barrels) exported to the United States. I am going to use Eric Wolf's cartogram creator script for ARCmap. What is an appropriate projection for my world shapefile to be in before I create the cartogram? I imagine that the ideal projection would be one which is equal area and preserves shape well, but I was wondering if there is a accepted standard projection for world cartograms.
Posted 06 October 2007 - 03:28 PM
Posted 08 October 2007 - 01:58 PM
I don't know about this algorithm, but equal-area and preserved shape don't go hand in hand. I wouldn't worry about equal-area, just ignore it. I would start with miller cylindrical or gall, which both portray shape reasonably well. You will need to experiment, but I would consider looking into removing joint boundaries/attachments between countries at some places - that creates too much distortion. Russia is often problematic for cartograms, as well as some of the bridges between continents (central america, middle east).
An interesting question; I would think you should first choose which projected coordinate system (PCS) you want to use, produce the cartogram using that PCS, and then be stuck with it's flaws. I don't think preserving shape is an issue since the cartogram is going deform the shapes anyway; though I can see that reprojecting the cartogram data into some other PCS could ruin the results.
I would have assumed equal area projections would be good for cartograms--I even remember learning that, since the point of doing a cartogram in the first place is to substitute geographic area with another variable; projections that do not represent area equally will visually skew the representation of the represented variable.
I guess the smart thing to do is pick a PCS that makes sense to you and you can test it by doing an ordered correlation between the cartogram geometry's area for each feature and the value of the original variable for each feature. If the correlation is high, then you're okay. I would suspect Cylindrical, Conic, and Polyconic PCSs would be very poor for world cartograms, while conformal and equal area might work decently. I think it would be facinating to create a table of results for a few standard projections just to prove which PCSs would work best.
Software Products Department
ESRI, Redlands, California
Posted 09 October 2007 - 04:00 AM
Maintaining the shape is a priority when you prepare the cartogram, it might be distorted, but you want to preserve as much as possible of the original shape as possible. Take New Zeeland, or Australia as an example, not having any boundary with any other country - you would want that to look quite undistorted, just sized to the variable. That's why I wouldn't suggest using Winkel Tripel, Robinson etc.
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users