Hi all, I read some other topics about data interpolation and I still have a question.
When I request ocean data by submitting a form, I enter locations. I wonder if I can use data from a point that is between the coast and the nearest native ERA5 grid point. Will it be interpolated between the offshore point and an onshore point where there should not be any data on waves ?
Michela Giusti said, on another topic : "The interpolation scheme used by ECMWF for wave data uses a scheme based on weighted distances between the target point and the 4 neighbouring native grid points, *except *when land is present, where only values from sea points are used. However, the scheme does not know at all of the actual land sea mask of the target grid. For this reason, when the data were pre-interpolated on a 0.5x0.5 grid to be stored in the CDS, points along the southern coast of Crete were made to encroach at bit over land. They are to be interpreted as values representative for the closest coastline."
I can re-formulate my question : when I want ocean data (wave state), should I take an extraction point that is further offshore than the first grid point available offshore (to make sure interpolation is done between only offshore points, for sea state) ; or can I request a point closer to the coast, inbetween 4 other grid points where some of these 4 points are inland ?
Thank you
Julien
1 Comment
Jean Bidlot
Hi Julien,
The first thing to realise is that ERA5 is a global dataset. This means that its resolution is still rather coarse. The grid points of the native grid, i.e. the grid that is used to do the actual computation has latitudes spaced every 0.36 degrees, which means that there isn't more details finer than 0.36 degree, or about 40km. There was however an unfortunate decision that was made when the data were prepared and copied to the CDS (from where most users get their data) to actually supply the data on a regular lat-lon grid of 0.5x0.5 degrees. In the middle of the oceans, this would not make much difference, but near coastlines, it further reduced the resolution, leading to apparent degradation in the quality of the product.
If one uses the wave data from the CDS (i.e readily available on disk), I would recommend selecting target points that are on the 0.5x0.5 grid, to avoid further degradation due to another interpolation.
Another possibility, but a lot more time consuming, is to request the data from the native grid. This implies getting the data from tapes rather than from disk. Again, the selection of target points should keep in mind that the native grid has a resolution of 0.36 degrees and the target points should be on that grid. My C3S colleagues should be able to advise on how to request data from tape.