Global Mapper Pro

splitting continents results in 5-10% reduction in measured area?

JSL Global Mapper UserTrusted User
edited November 2014 in Vector Data
I'm trying to get polygons of just the five major continental landmasses, and Natural Earth is the closest I've found (1:10m Physical Vectors | Natural Earth). However, their land shapefile comes with North America joined to South America, and Eurasia joined to Africa. The problem is that no matter what technique I use, splitting the continents results in a 5-10% decrease in the area of the split continents obtained through the "Update measure attributes" command. I've tried creating a line from the unsplit continent, splitting the outline with another line, and then creating new areas from the splits, but even though the split continents look exactly the same as the unsplit continents, with only a line separating them, there's a massive difference. When I split Africa-Eurasia into two separate polygons, I got a difference of 10495226 sq km between the unsplit landmass and the two splitoffs, and when I split North America-South America, I got a difference of 7973777 sq km.

Attached is a workspace with the original Natural Earth shapefile and the split polygons.

Any help would be appreciated--I'm trying to find percentages of areas compared to the individual continents, and it's difficult to know which measurements I should trust.


  • tjhb
    tjhb Global Mapper User Trusted User
    edited November 2014
    I'm trying to find percentages of areas compared to the individual continents

    Trying to find percentages of areas of what compared to the individual continents? Do you mean for each continent, its percentage of their total area?

    In any case, you should change the projection for your workspace to one that is equal area and applicable to the whole world. For example, Mollweide or Eckert IV. (There aren't very many other choices.) Then measure areas again.

    In these projections you should find that the split and combined areas agree very closely. (I can't do a quick check of that since the unsplit versions don't appear to have been embedded in the .gmw file, only the split versions.)

    The results for a given area should also agree fairly closely between Mollweide and Eckert IV. How closely depends on whether you use a spherical or ellipsoidal datum. Neither projection has an ellipsoidal form (to my knowledge). Using datum Sphere (radius 6378137m) the areas for Africa are respectively 29268631 km² and 29268633 km², while using WGS84 the results are 29341981 and 29440609.

    (I'm not sure why the results are so much wider apart for WGS84; and clearly the formulas in GM don't simply ignore flattening, at least not by using a sphere with radius equal to the WGS84 semi-major axis of 6378137m, or otherwise we would get the same results as when using Sphere (radius 6378137m) explictly.)

    There really isn't a single "correct" answer to areas at this scale, because of the way their measurement depends on projection and datum. But you would expect that the sum of two areas, measured separately, would be equal to their combined area, measured together (as you say).
  • tjhb
    tjhb Global Mapper User Trusted User
    edited November 2014
    P.s. it would be interesting to know from one of the developers wat method GM uses for calculating measure attributes of unprojected objects (in unprojected space). Does it, for example, reproject each object to a local Lambert Azimuthal Equal Area projection (centred at its centroid) to measure area, and use Vincenty's alogorithm for lengths?

    (And any comment about what happens when measurements in Mollweide and Eckert IV are made "using" (or not) an ellipsoidal datum would also be a bonus.)
  • tjhb
    tjhb Global Mapper User Trusted User
    edited November 2014
    P.p.s. I think two better choices for the equal area projection to be used for global area measurements are Lambert Azimuthal Equal Area and Lambert Cylindrical Equal Area. Both have the advantage that they have ellipsoidal forms.

    Some quick cross-checking suggests that LCEA, with an ellipsoidal datum, may be the best choice for accurate area measurements at global scale.
  • JSL
    JSL Global Mapper User Trusted User
    edited November 2014
    Thank you very much for that information! I have a number of large drainage basins in different continents, and I’m trying to find out what percentage their area is of the continent’s surface.
    I tested out the South America-North America polygon using the projections you suggested, comparing the split and unsplit polygons, and got the results below:

    Attachment not found.

    So it seems like either of the three you mentioned work equally well, but now I’m wondering which projection I should use, since they all give different measurements for the same polygon. What projection in GM should I use for getting a true measurement of the earth’s surface?
Sign In or Register to comment.