[NelsonMinar]

This map is a great example of how Google Mercator is not always the right projection for a visualization. Antarctica is an important source for meteorite research; black specs are easy to pick off the white ice. But this map projection completely distorts Antarctica so you can't really read that part of the data. Cool map, but sometimes the off-the-shelf map tool isn't the right choice.

[a235]

It's not google. This one is based on http://cartodb.com/, however the argument is still valid

[sp332]

This reply from Google explains why they use Mercator. http://productforums.google.com/d/msg/maps/A2ygEJ5eG-o/KbZr_... Not sure if it's really satisfactory, but at least there's a reason.

[mapmeld]

The projection (how the globe is represented in 2D) is still Google's. Technically it's Web Mercator / EPSG:900913, where 900913 is a rather clever choice by Google.

[gilgoomesh]

> not always

I think "almost never" would be more accurate here.

Mercator is only used because it's the projection you get when you don't apply any projection to longitude/latitude. It's a result of the map creator being lazy.

[ghaff]

It depends on the scale. For a world map, Mercator is, as you suggest, almost never a good choice. For many other situations where it's important to show the correct angular relationship between objects (such as streets), Mercator is the most faithful. And I can understand why there might be issues with using different projections at different scales.

I could equally well say it's lazy to use just use Google Maps or some other convenient mapping service that uses Mercator projections at the world level when those services are not really optimized for displaying at that scale.

[gilgoomesh]

If you had a street that crossed the North Pole you wouldn't want a Mercator Projection. It would look thoroughly deformed.

Many orthogonal, North-aligned projections will look fairly similar same when zoomed into a street at least a couple degrees latitude away from the poles. Ideally, you would always use a 3D projection of that section of the sphere (e.g. Google Earth not Google Maps) or a local two-point equidistant projection since these are always the most accurate. An orthogonal projection only suffices in common usage because there is so little need for maps near the poles.

In any case, I should have qualified my original comment with: "for a whole-earth map or a map showing distributions". Mercator is always a bad projection for a map showing distributions because it seriously misrepresents area, undermining the effort to accurate portray distribution.

[NelsonMinar]

I think you're mistaken; the projection you get when you naively plot lat/lon as x/y is Plate Carrée (aka equirectangular). This visualization is Google Mercator.

You see a lot of Google Mercator because it's the native projection of Google Maps and pretty much all subsequent slippy map libraries and tilesets. It's a great projection when you want something that works easily with map tiles and is zoomable. But it's not a great choice for a whole-earth visualization.