If you had provided an X and Y grid, then it would be using bilinear interpolation. Since you did not provide a grid, the effect is the same as plain mean() of the four corners.
Once the mean is determined, the value goes through color mapping. If the value is less than the first element of the axes CLim, then it is mapped to the first color in the color map. If it is greater than the last element of the axes CLim then it is mapped to the last element in the colormap. Otherwise it is mapped to (value - CLim(1))/(CLim(end)-CLim(1)) * number of colormap entries -- so proportion of the color range maps to proportion of the color map as an index.
Why that shade of green? Because that is what happens to get looked up in your active colormap. If you were to colormap() a different colormap then it would change.
