2

I draw several points of Lon/Lat(EPSG:4326) on the OpenStreetMap v 4.6.5 (EPSG:3857).

It seems like the distance between the points are closer in the Latitudinal direction then in the Longitudinal direction.

The resolution/distance is approx 5 cm between the horizontal points, and approx 10 cm between the longitudinal points (Measured with the OSM-measure plugin).

distance Near equator this does not seem to be a problem, as the distance between all points seems uniform. I thought plotting lon/lat points on OSM, should be straight forward :)

What is causing this problem and how do i solve this further north of equator?

  • The points are not closer in either direction, it is the longitudes which get closer as you go north, so a line a the equator which crosses 5 degrees would cross 10 degrees at 60 north i.stack.imgur.com/hkOSX.png But both EPSG:4326 and EPSG:3857 would make it look twice as long, while a line which looks the same length is half the length (additionally EPSG:3857 will exaggerate apparent north/south distances). – Mike Aug 22 '19 at 13:21
  • EPSG:4326 is lat/long ~ epsg-registry.org/export.htm?wkt=urn:ogc:def:crs:EPSG::4326 – nmtoken Aug 22 '19 at 19:54
  • @nmtoken OpenLayers defines it as Lon/Lat so in the context of an OpenLayers question that is more understandable – Mike Aug 22 '19 at 21:48
  • OpenLayers isn't in a position to define the axis order, only EPSG can do that. – nmtoken Aug 23 '19 at 6:12
  • Theoretically yes, but using the same definition as the application you are using helps when you want correct output. – Mike Aug 23 '19 at 11:28
7

This is due to the difference between a measurement in degree and a measurement in meters (nothing that you can do about that) and maybe to the projection of the map (all projections induce distortions of the reality, your choice of a projection will depend on what you try to do).

If you assume that the Earth is a sphere, one degree along NS direction will follow a meridian and its size will be 1/360 of a meridian. All meridians have the same length, so you will always have the same distance. However, if you look along WE direction, you are on a parallel, and the size of the parallels decreases when you move towards the poles (size of parallel = size of equator * cosinus(latitude of the parallel). )

So if you have a 1/2 ratio, you are probably somewhere near the + or - 60° of latitude. It is correct that the same angular values of latitude and longitude are not the same euclidian distances on the surface of the Earth if you are not on the equator. I can also guess that your coordinates are truncated at the sixth decimal. This precision is OK for most applications, but if you have a professionnal differential GNSS receiver, then you original precision was degraded by the rounding, and you loose more precision by rounding latitude values than by rounding longitude values.

If you want a square of degrees to look like a square, you can use "Plate carre" projection, but everything else will be distorted. Local projection usually provide the best compromise with minimum distortions of the "reality" (but representing a part of o sphere on a flat screen or piece of paper always create some distortions, just choose those that is the least ennoying to you.)

Quick steps to select a projection:

1) if you work on a "small" area (vs global) use a local coordinate system. For larger regions: cylindrical projections near equator, conical projections at mid-latitude and azimuthal projections around the poles)

2) Select the most important feature for your work, most of the time one of the following A) keep the shape of the object (preserving local angles) => conformal projection B) measuring areas => equal-area projection. It is not possible to combine these true properties on a flat surface.

3) For specific uses, other properties include equidistance (for a given set of lines on the map) or loxodrome representation as straight lines (useful for navigation)

Plate carrée projection is equidistant in the NS direction, but it is neither conformal nor equal-area. Therefore I do not recommend it, except for simplicity.

| improve this answer | |
  • 1
    Image of Plate carrée en.wikipedia.org/wiki/Eq.uirectangular_projection – user30184 Jul 1 '19 at 10:17
  • Thanks for a very good answer. To me the "true representation" of the figure/pattern of the points is what is important. I don't need it to be a map either (could be a Cartesian coordinate system). What are my best options for visualizing the square as you say, any suggestions? Reading this: gis.stackexchange.com/questions/25802/… makes me confused. Does REAL "Plate Carre" (not pseudo) has its own EPSG code? – otk Jul 1 '19 at 10:34
  • @otk - it sounds like you need a local projection for the area where your points are – Ian Turton Jul 1 '19 at 10:50
  • Any good suggestions on what tools exists for local projection? Openlayers? Excel? Others?`PowerBI? – otk Jul 1 '19 at 11:46
  • use a GIS software (e.g. QGIS). But you can also define a local projection in openlayer – radouxju Jul 1 '19 at 11:48

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.