Subscribe to this thread
Home - General / All posts - What is the difference between Lat/Long and Web Mercator
Forest
578 post(s)
#02-May-18 01:35

Both Lat/Long and Web Mercator use WGS84 as a datum and latitude and longitude. The coordinate values are the same for survey pegs in the ground for both projections. This suggests that only the shape of polygons as seen on screen will vary and so will the intrinsic variables for perimeter and area. If you don't care about shape, or the precise values of perimeter or area (e.g. habitat maps), then it really will not matter whether you use web mercator or lat/long. If you look up the definition for web mercator coordinates, they are the same as WGS84 but the coordinate order is switched to x,y.

tjhb

8,273 post(s)
#02-May-18 03:17

There's a lot to be unpacked carefully here, and I am not very good at it.

The first sentence is not quite true. Web Mercator expresses locations in WGS84 coordinates, that is, coordinates over the WGS84 ellipsoid, but then re-interprets them "as if" they lay on a sphere having the same radius as the semi-major axis of the WGS84 datum (so with the same radius at the Equator, but with no flattening).

(Latitude / Longitude measures coordinates in degrees. Web Mercator only pretends to measure in degrees: underlying the pretence it is really in metres, like standard Mercator but less accurate.)

The second and last sentences are true. [Correcting myself. I was wrong before. This is tricky.] Coordinates match, although they will not show in the same place when visualized in the plane (because of the misalignment mentioned above).

The middle parts are true as well. To rephrase, if you don't want to use Web Mercator for (a) planimetric display, (b) measurement or (c) navigation, it is OK. It is fine for storing data, and for visualization on a 3D sphere (as in Google Earth).

Have a look on the EPSG registry for code 3857, especially the stuff off to the side:

Remarks:

Uses spherical development of ellipsoidal coordinates. Relative to WGS 84 / World Mercator (CRS code 3395) errors of 0.7 percent in scale and differences in northing of up to 43km in the map (equivalent to 21km on the ground) may arise.

Scope:

Certain Web mapping and visualisation applications. It is not a recognised geodetic system: for that see ellipsoidal Mercator CRS code 3395 (WGS 84 / World Mercator).

tjhb

8,273 post(s)
#02-May-18 03:46

In my third para,

although they will not show in the same place when visualized in the plane...

the words "in the plane" would be better left out, since the same is also true for visualization on different a sphere or ellipsoid (i.e. other than the sphere used by Web Mercator itself).

Dimitri

5,036 post(s)
#02-May-18 08:38

If you look up the definition for web mercator coordinates, they are the same as WGS84 but the coordinate order is switched to x,y.

Be wary of confusion caused by the recent fashion to use the term "WGS84" to refer not to a datum but to a projection, Latitude / Longitude. Mixing together what used to be precise, separate meanings can cause confusion.

"WGS 84" is the name of a datum, that is, an Earth ellipsoid. "WGS 84" is not a projection, and to use the term as a synonym for "Latitude / Longitude Projection" is a mistake. People do that all the time, usually because they don't know any better, but often because they know perfectly well that WGS 84 is an ellipsoid and not a projection, yet they have succumbed to use of short hand jargon.

As noted in The Earth as an Ellipsoid topic, WGS 84 is just the latest datum in a series of Earth-centered ellipsoids, following WGS 66 and WGS 72.

Before WGS 84 came around, a Latitude / Longitude projection might have used WGS 72 as the base, or WGS 66 as the base, or NAD 27 as the base, or Clarke 66 as the base, or some other ellipsoid. That's part of the message of the Latitude and Longitude are Not Enough essay. Using "WGS 84" as a synonym for "Latitude / Longitude Projection" is proclaiming unawareness that much of the world's Latitude / Longitude data has nothing to do with WGS 84.

Pseudo-Mercator coordinates are not the same as Latitude / Longitude coordinates, even when both projections use the same Earth ellipsoid. Pseudo-Mercator uses meters as units of measure and Latitude / Longitude uses degrees. In addition, the numbers that you get for both for the "same place on Earth" will vary depending on the ellipsoid used, as illustrated in the Latitude and Longitude are Not Enoughessay.

To summarize: the key message I'm trying to convey is to not to use the same phrase to refer to distinctly different concepts, that is, the difference between a projection and an Earth ellipsoid. I realize sometimes it seems hopeless to push back against slacker imprecision in the use of technical language, but everyone can help by putting a finger in the dike against entropy by never, ever using "WGS 84" to mean a projection and not an ellipsoid, that is, a datum or a base.

[To avoid unintended, politically incorrect humor, I use the American/Canadian spelling for an embankment that holds back water, and not the English spelling used in other countries that uses a "y" instead of an "i"....]

Forest
578 post(s)
#03-May-18 03:19

Thanks Dimitri,

I have been through a few datum changes such as AGD66 > AGD84 > GDA94 and am familiar with the concept of datums and projections. The issue is that my data is collected with a GPS in WGS84 in Lat/Long. I store most of my data in external non-spatial databases. I want to be sure that there will not be any issues due to manifold's default use of Web Mercator. It bothers me that the Lat/Long projection in manifold does not specify that it is using a WGS84 datum. WGS84 is defined such that the net sum of continent drifts from continents is averaged. If continental drift is a factor, then a datum that moves with your continent will have to be used. Australia is about to grapple with that as WGS84 positions are now at least 2.5 m out.

I am just thinking that a GPS position shown over a Web Mercator image must also use the Web Mercator projection in order to be visually aligned with the source object. Likewise, if I mark a point on a Web Mercator layer and send it to my GPS, I will be happy if I can navigate to the point in the real world. It does not bother me that if the web mercator position, if overlaid on the surface of the earth would be in the wrong place. We deal with this all the time as vertical datums are often out with respect to sea level or gravitational surfaces of equal potential. All that matters to me that the points picked up in the field will be appear to be in the correct position on a web mercator base map. As long as all data is subject to the same distortions of location, all is fine. I am working at large scales (1:10 000) so curvature of the earth or distortions from projections are not significant. I was just checking that the coordinates for a place on the earths surface measured in Latitude Longitude are the same as the coordinates for the representation of the same place on a Web Mercator map.

tjhb

8,273 post(s)
#03-May-18 05:47

Both of you are still making some significant mistakes BUT I normally would too. The details are really hard to get right, because....

It’s a shitty, shitty projection, almost always to be strictly avoided.

It was a mistake for Manifold 9 to have made it a default (as I half-attempted to say at the time). I still hope that that might be changed. It’s not suitable for serious GIS nor for cartography.

Dimitri

5,036 post(s)
#03-May-18 07:26

It was a mistake for Manifold 9 to have made it a default (as I half-attempted to say at the time). I still hope that that might be changed. It’s not suitable for serious GIS nor for cartography.

If you mean Pseudo-Mercator, the above is a bit like how UNIX people ensured they would become irrelevant to billions of people by criticizing Windows for not being suitable for serious computing. UNIX elitists completely forgot that those unwashed masses of billions of people rather enjoyed the conveniences of Windows, while cursing the wretched inconvenience, as they saw it, of UNIX.

Pseudo-Mercator is the world standard for presentation. Nothing else comes even remotely close to the trillions of views Pseudo-Mercator gets as result of effectively universal use on the web. That's the lay of the land whether anybody likes it or not. I didn't like Pseudo-Mercator at all at first... the tricky "auto" fakery it uses is genuinely annoying from a purist, technical perspective. But I have to admit it has its uses and it sure is convenient.

Pseudo-Mercator *must* be the default option in a situation where those who do not like it can in a nanosecond choose something else, while those billions of people who are looked down about by elitists do not find it so easy to switch from what they regard as non-standard to what they regard as the standard.

In point of fact, Manifold uses not one, but *two* projections that annoy experts: Pseudo-Mercator and Latitude / Longitude. You could make the case that Lat/Lon is the profoundly more inappropriate one. But there again, the reality is that there is so much archival data in Lat/Lon, for all the right reasons, that it too *must* be one of the single-click options.

Which brings me to the last point... if you don't like PM, don't use it. Define your own favorites that you deem suitable for serious GIS and cartography and then you'll have them with one-click ease whenever you want. Takes but a second. :-)

Dimitri

5,036 post(s)
#03-May-18 12:27

Both of you are still making some significant mistakes BUT I normally would too.

I knew I should have spelled it with a "y".... :-)

Apologies, by the way, for ranting on Pseudo-Mercator. I re-read my post and it comes off far more critical than intended. All criticisms of Pseudo-Mercator are absolutely correct. It's still become the world standard.

Can't resist, as the title of this thread reminds me of the perennial variations on the old joke, "what's the difference between a used car salesman and... <target>? "

Question: What is the difference between Lat/Lon and Web Mercator?

Answer: One of them knows when it is lying...

Forest
578 post(s)
#04-May-18 00:42

Lets do an experiment.

I will create a table with a latitude and longitude field and then compose circles using the lat/long data for centres and some distances ranging from one to a few kilometres. The distances were converted to degrees externally to make it possible to compose circles in data that is expressed in degrees.

As Lat/Long with a WGS84 datum and Web Mercator are supposed to use the same Lat/Long coordinates for the same place on the earths surface, I should be able to make two drawings based on the same geom. If I try to do so, I find that creating the first first drawing is fine and works as expected. When I create the second drawing and the change the coordinate system to Web Mercator, I find the circles are in the correct place but are squashed flat. I also find that the coordinate system of the first drawing has also been changed to web mercator. This strongly suggests to me that there is a design issue with Manifold. The coordinate system should be set when a geom field is created and drawing should only be able to inherit this setting. That changing a setting in one component impact other components is a trap for the inexperienced.

If I change the coordinate system back to lat/long. The circles are still squashed flat (blues lines in image). I would have expected two sets of circles that were almost in complete registration as per the famous essay. The

I have attached a M9 map file for anyone who wants to do the experiment.

Attachments:
2018-05-04_Circles.jpg
CoordSysTest.zip

Forest
578 post(s)
#04-May-18 01:30

I think that circles do not survive changes of coordinate system. If I change coordinate system from Lat/Long to UTM or to another geographic coordinate system such as AGD66, the circles still get flattened.

Forest
578 post(s)
#04-May-18 01:57

Boxes survive changes of coordinate system, Circles do not.

TAL6 post(s)
#04-May-18 17:54

Not an expert on this but the circles in your drawing (CoordSysTest.zip) are composed of curved segments, not straight segments. Others can explain this, but here is my explanation:

As I recall circles composed of curved segments (created with the transform - compose circle) do not survive the change of coordinate system. Circles composed of straight segments do survive changes of coordinate system. The curved segment circles can be converted to straight segment circles using transform - buffer.

adamw


8,139 post(s)
#05-May-18 14:26

Exactly.

If you change the coordinate system of a drawing with curves, curves get deleted, the details are lost.

If you first linearize curves using the Linearize transform (you can use the default parameters, I kept tolerance at 0 and increased curve limit to 40), and *then* change the coordinate system, the details are kept (the resulting drawing overlays perfectly with the original lat/lon one and you can only see differences if you zoom enough to see the effects of linearization).

Why do we not linearize curves automatically during reprojection? Perhaps we should, but it's a question of what parameter values to use. We are planning to add some controls for that to the Change Coordinate System dialog.

Forest
578 post(s)
#08-May-18 01:22

Thanks, good info.

I am using circles to indicate uncertainty in locations of observations so as long as something roughly circular survives changes of coordinate system, I am happy. I appreciate the info on geographic circles etc and the complexities of variation in scales but fortunately, I don't have to go there.

Dimitri

5,036 post(s)
#04-May-18 08:16

Boxes survive changes of coordinate system,

Circles drawn on the "surface" of the Earth are indeed circles. Whether they look like circles in a given projection depends on the properties of that projection. In general, unless a projection is specifically tailored to show an accurate circle in the given area of interest it won't show circles on the surface of the Earth as circles on the map or monitor.

If you don't take care to use a projection that indeed displays a circle on the surface of the Earth as a circle on your monitor for the area of interest, then it is highly unlikely that what you draw that looks to be a circle on your monitor will really be a circle on the surface of the Earth.

Example: pop open a drawing in Latitude / Longitude. Freehand, draw what you are convinced is a circular shape covering, say, several European countries. Re-project into Orthographic projection centered on the center of your "circle" .... you'll see it's not a circle at all in real life.

Boxes do not "survive" changes either. Please read the link I recommended to you earlier. As you can see from the illustrations in that topic, what are horizontal and vertical sides to Idaho, forming right angles as a "box" does, in one projection are not horizontal and vertical lines forming right angles in a different projection.

Dimitri

5,036 post(s)
#04-May-18 08:12

The distances were converted to degrees externally to make it possible to compose circles in data that is expressed in degrees.

That is a mistake. Degrees are not the same length all over the earth, and they are not even the same length depending on the direction.

Hire a boat and go to the intersection of the Prime Meridian and the Equator in the ocean off the coast of Africa. Draw a circle that has a radius of one degree and it ends up being about 110 km in radius. It actually looks like a circle, too, if you hover directly above it in your space plane a few hundred km above.

Now, fly to the North Pole in your space plane. Put on your parka and step outside, walking over to stand directly onto the North Pole. Step ten paces towards the South. That would be, of course, any direction from the North Pole since all straight lines leading away from the North pole are heading South.

Suppose one of your "paces" is about 0.6 meters. Ten paces means you've stepped about six meters from the North Pole. A circle with a diameter of 12 meters has a circumference around 37.7 meters. Divide that circle into 360 parts (one part for each degree of longitude) and you get a result of about 10.5 cm for each degree of longitude.

Ten paces from the North pole, one degree of longitude that was 110 km in "length" on the Equator is less than 11 cm in "length".

So now, standing ten paces from the North pole, try drawing a circle that had a radius of 1 degree. What you end up with is a very skinny ellipsoid shape, basically a line at the top and a fatter line at the bottom, which extends 220 km to the south (degrees of latitude are still about 110 km) but which varies from an east west "radius" of 10.5 cm near the top to an east west "radius" of about 1920 cm at the bottom.

That's because the east-west length of a degree varies from about 10.5 cm a few paces from the North Pole to about 1920 cm around 220 km to the south of the North Pole.

The moral of the story: In the Latitude / Longitude projection, using degrees as linear units of measure to specify the location of points each of which is one degree in distance from a central point, does NOT result in a circle. That is because in the Latitude / Longitude projection, the "length" of a degree varies depending on where on Earth that degree is located and also the bearing of that degree. Degrees of latitude are approximately the same, but degrees of longitude are not, so anything except measurements straight up and down, so to speak, will vary.

The shape distortion you report is an accepted and well-known consequence of using different projections. It is one reason why one should choose the right projection for the particular task they have in mind. See the discussion in the About Projections topics.

Dimitri

5,036 post(s)
#03-May-18 07:13

It bothers me that the Lat/Long projection in manifold does not specify that it is using a WGS84 datum.

? Sure it does. See the Example: Re-project a Drawing topic.

http://www.manifold.net/doc/mfd9/images/eg_reproject_a_drawing01_11.png

I want to be sure that there will not be any issues due to manifold's default use of Web Mercator.

Not exactly right. Manifold uses whatever your data says it uses, as about a zillion examples show. See the various introductory examples, plus examples like the above. Import a shapefile in, say, Lat/Lon and that's what the drawing uses.

When you create a new, blank drawing Manifold offers Pseudo-Mercator by default but in all cases you can use whatever you want. A new map automatically takes on whatever projection is used by the first layer added. This is very basic stuff, so if it comes as something new that is a strong indication you should dive into the documentation and read it in order: Getting Started, then Basics and the Examples, starting with the Introductory Examples.

That a Map can show on the fly for display purposes a reprojection of layers in different projections is a convenience. It's not a change in the data.

Manifold User Community Use Agreement Copyright (C) 2007-2017 Manifold Software Limited. All rights reserved.