Area-weight in spatial correlation?
Dear all,
I have been using spatial correlation with the s2dv tool (function ACC.R) as well as with other languages. I wonder now whether we should consider an area-weighting for the grid cells in the computation of this correlation.
In one sense, it seems correct to check the model skill equally for each grid point where the 1-D physical computations of the model are applied. On the other hand, it could be justified to check the model skill per unit of surface and in this case, we should consider to give more weight to the large grid cells.
Anybody would have a point of view about that?
Best regards
Martin