I have a problem which is sort of similar to the one here:

Dealing with data that spreads across UTM Zones?

Essentially I am looking for a projection for this region:


The projection should have as little distortion as possible so I can use geospatial data structures. I tried using UTM, but the rectangle spreads across the zones 32,V and 33,V. Could you recommend some other projection? More generally, do you know how to get a good projection for an arbitrary bounding rectangle?

Is there a projection which is already implemented in jscience or a similar library?

  • What datum is your data in?
    – Mintx
    Apr 11, 2016 at 13:54
  • WGS84. Sorry, I did not think of that :)
    – hfhc2
    Apr 11, 2016 at 13:57
  • 2
    "How to get a good projection for an arbitrary" region is both too vague and too broad to be answerable. It is vague in that it does not specify what "good" means and it is broad in that the answer depends on the size, location, and orientation of the rectangle and (to some extent) on the shapes and orientations of the features it contains. The first part of this question would benefit from an explanation of what kind of distortion should be minimized: areas, angles, bearings, shapes, a graticule, or something else?
    – whuber
    Apr 11, 2016 at 20:03
  • Well, I want to perform map matching. I need to find closest points so distances should not be distorted. If in addition I take angles into account they should not be distorted too much either..
    – hfhc2
    Apr 12, 2016 at 11:14

2 Answers 2


If you're looking for an established projection and you don't mind accuracy to around 2 meters, I would go with EPSG::3006

This assumes your data is in the SWEREF99 datum, but the difference between that datum and WGS84 is very slight.


A very simple method is to put up a custom transverse mercator projection centered on the center of your study area:

+proj=tmerc +lat_0=51.4 +lon_0=7 +k=1 +x_0=0 +y_0=0 +ellps=WGS84 +towgs84=0,0,0,0,0,0,0 +units=m +no_defs

This works for about +/- 15° of longitude. On longitudinal larger areas, a Lambert projection with one or two standard parallels would be better.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.