That's the bottom of the sphere, and the second ring should have all surface circles facing the camera but it's not! This is repeated across all the other "rows" on the sphere but the effect only happens at multiples of pi/2 radians so it's not too noticeable but definitely bugs me out.

Some helpful links:

• https://mathworld.wolfram.com/SphericalCoordinates.html
• https://math.stackexchange.com/questions/2314356/finding-the-normal-to-a-sphere-at-any-point-with-spherical-coordinates
• Look at other three.js projects and how they arrange items in a sphere. Surely they deal with wrapping and rotation :) I recall that periodic table arranger having a sphere arrangement that was nifty.

I've decided to use an equirectangular projection when mapping (lat,long) to cartesian coordinates since the mathematics worked out the easiest. This didn't need to be the case, and we could have used any projection equation to do that (ask yourself why).

The code is baked to use a PNG that was created using an equirectangular projection. Find another projection equation (+ PNG) and use that to render the landmasses. Why should we do this? One reason why is that there could be projection calculations that are faster than what we currently have, and we can benchmark the performance of the two.

We don't even have to stop there! We just need some "thing" that can tell us whether there's land at a (lat,long) coordinate. That could be implemented any number of ways. For example, we could send a request to Google Maps with a latitude and longitude, and it can return back whether there's land or water there.