Incorrect coding of x_start/y_start for space view grid definition about iris-grib

[mo-cbat]

@lbdreyer Right, so I've contacted EUMETSAT and we've deduced that the space view GRIB files that they produce have actually got the wrong values set for (Xo, Yo). Currently they are set to (0, 0), but they should be (1, 1). With the correct values there is no mismatch with the coastlines when plotting using Scitools.

So in conclusion, this pull request is wrong and I'll close it, and Iris is great and works perfectly! Keep up the great work :D

[lbdreyer]

I'm struggling to understand your example so I've tried to draw a diagram to check if this is what you mean:

As I understand there should be two grids: (i) the one that represents the visible full Earth's disk and (ii) the one that represents the sector image. In the above image, the full Earth grid (i) is represented by a 5x5 grid, but the sector image grid (ii) that may be contained in a GRIB file only covers part of the full grid - a smaller 3x3 grid.

The origin of the sector image (Xo,Yo) is at (0,0) (denoted by a yellow circle in the above image.
The Sub satellite location (Xp, Yp) is at (2, 2) (denoted by a yellow cross), which is the centre of the full Earth grid

From, the above diagram, I would have thought a start coordinate of -2 was correct?

[mo-cbat]

Thanks very much for that @lbdreyer! Your example is far better than mine :)

I think the key issue here is that the first cell in grid (i) should be (1, 1), whereas it is (0, 0) in grid (ii). At least that is how I understand the contents of the space view GRIB files that EUMETSAT produce (I'm not aware of any other centres that produce such files with which I could check). So when relating grid (ii) to grid (i) you need to add 1.

I will contact EUMETSAT to see if my understanding is correct and comment here again with my reply.