6

If a coordinate is stored in a geographic coordinate system, it's quite easy to calculate the following things exactly:

  • Distances (geodesic)
  • Point-in-polygon
  • Areas
  • etc.

If you store the coordinates in a projected coordinate system, these things are calculated in a Euclidean way, which means there will be huge distortions if your map covers a larger area of the earth.

So, in my view, storing the coordinates in a geographic coordinate system has the

  • advantage that all calculations can be made exactly, and the
  • disadvantage that they must be projected before they can be rendered in a map.

Since the latter is something that can be done very fast on modern processors, I don't think it would ever be a huge disadvantage.

Are there any other disadvantages? If not, then why does one ever store geographic data in projected systems?

One answer to the second question could be that the projecting of coordinates was not always so cheap. On processors of the 70s, it could be better to store the coordinates in the way they are rendered, to avoid having to project them every time the map is rendered.

  • I agree with you @cheeesus when I was creating a European corporate-wide GIS it made sense for us to store everything in a common geographic reference system in this case (ED50) and depending on the UTM Zone one was working in it was part of our job to supply the governance and assistance to ensure the user knew how to project the data before doing any analysis...this stand ensured that data was much more simple to maintain from a management of CRS perspective. – spk578 Aug 31 '16 at 7:55
  • The most significant disadvantage for me would be users not understanding the difference between geographic vs projected data and a decision being made on the output of a calculation on the incorrect system. However, I do think that is a matter of process and education within an organisation. – spk578 Aug 31 '16 at 7:57
  • Sometimes the source data comes in a projected CRS, or is even measured in one ('official' geodetic measurements, like cadastre, are usually reported in official projected coordinates; satellite images might be provided in UTM, so their footprints will originally be UTM as well). – mkadunc Sep 1 '16 at 1:29
7

When I was involved in creating a European corporate-wide GIS it made sense for us to store geographical data in a common geographic reference system. In this case, the standard was European Datum 1950 (ED50). Depending on the geographic area a user was working in it was the user's responsibility to ensure they knew how to project the data before doing any analysis.

As a GIS Consultant, it was my responsibility to ensure the GIS enabled users to do their work without needing continuous assistance. Therefore we supplied governance (only relevant CRS and their conversions/transformations were made available and specific map documents with those coordinate system settings built in were created for targeted work areas), documentation (how the corporate geodetic system works, who to contact and where to find more information), and training (including third-party courses on CRS) to ensure users understood the problems of not working with coordinate systems correctly for a spatial analysis.

The advantages for me are:

  • Easier data management system
  • Less accidents made with using unprojected or incorrectly projected data
  • More flexible if company decides to incorporate another geographic system

For me, disadvantages are users ignoring the system and running an analysis anyway on unprojected data.

The only reason I think for storing data in a projected coordinate reference system is if the area you are working in is small enough and defined enough to only need the projected CRS. For example, one might use OS National Grid if only working onshore Great Britain.

4

There's a tradeoff; the advantages you mention in precision/accuracy are offset by increased computation. For example, spheroidal calculation of area is much slower than the Euclidean calculation - and if the data are stored in an area-preserving projection, then you will get accurate results more quickly.

When dealing with small areas, the loss of accuracy may well be less important than the computational cost of the results - that's a design decision that depends very much on the application, of course.

It's easy to dismiss coordinate conversion as "easy" or "cheap" - but when you have millions of points to project, it does all add up!

In summary, the best coordinate system to use depends strongly on what processing you're doing. Don't assume that one size fits all!

  • 2
    There's an asterisk that needs to be attached to the assertion that an area-preserving projection yields accurate results for the area. This is only true if the sides of the polygon are well approximated by straight lines in the projection. This may not be true if you're dealing with a polygon with long sides which are geodesics. In this case, it's likely to be faster to calculate the spheroidal area directly. – cffk Aug 31 '16 at 19:15
  • And there's a further asterisk to add to the assertion that people expect long sides to be geodesics: they frequently expect them to be rhumb lines! (See, the northern and southern boundaries of various US states, and US/Canada border.) – Paul Ramsey Sep 12 '16 at 17:47

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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