http://local.wasp.uwa.edu.au/~pbourke/geometry/sphere/

thank you, but that doesn't address the question.

If you have to have the number of points relative to the changing circumference of each latitude of the sphere, you are going to need more than a 2-dimensional array.

However, if you can have the same number of points at each latitude, but at an increasing distance from each other, it is possible to have an equation that let's you determine the (x,y,z) of each point and store them in a 2-dimensional array. eg. nootropic's link above.

___________________
-Richard David Wong
http://www.rdwong.net

Then 2d array was a guess(hope).  I just don't know how to approach the problem.
If you go here: http://mrob.com/pub/comp/xmorphia/ you'll see the type of equation I'm dealing with.  I can grow this equation on a surface by using simple u,v,w; however, as the surface becomes more curved, the pattern becomes more distorted...pinched.  My thought is to use less pixels as the curve increases, the problem is that each pixel must be able to query its neighbors for each iteration. I'm unsure how to do that.  I've considered making a point class with pointers to it's neighbors, that still requires finding the neighbors.  I'm hoping the brighter, more experienced than me, members of this forum can offer some pointers.

Squares onto a sphere do not map well. It inevitably leads to discontinuities, overlaps or significant distortion of shape. Can you consider triangles instead? There are plenty of triangular tessellations of spheres available and since each triangle will have three edge sharing neighbours, diffusion models should be possible.