- As for Geographically Gridded data being reprojected via the EWA algorithm, I think there is utility in apply EWA to general reprojection of projection gridded data.
- In this case, the comparison of source data projection to target data projection results in two ellipses of influence, as was shown for the swath data projection use-case.
- For projection-gridded data, you can imagine the Tissot ellipse for the source data defining an ellipse of influence of the source cell to a geographic region on the earth's surface. And, you can imagine the Tissot ellipse for the target data defining an ellipse of influence for the target cell back to a geographic region - as is the case with swath data.
- Thus there is a strong parallel of applying EWA to projection gridded data, with two elliptical areas of influence, just as it is used for swath data.
- In this case, there might be a larger resulting ellipse of influence, in some areas of the target output grid, than is seen with swath data. This relative difference in application, however, I don't think affects either the validity of the results, or the relative performance of the algorithm. Per source-cell processing might exceed that for swath data, in those regions, but such processing is appropriate and necessary for complete grid coverage. I expect that a similar processing expansion would occur for reverse projection techniques as well, thus EWA remains competitive.
- Re. the ll2rc method, in this case there is the necessity of both the projection of the source grid location to lat/lon values and the projection of geographic location to target grid. Both approaches mentioned for handling Geographically located data are relevant - (A) compute lat/lon data arrays to match the source data grid, and then feed into the existing ll2rc. Or (B) - supplement the source data location projection to the existing ll2rc implementation to create what I would call an xy2rc method.
|