Does anyone use quadratic interpolation methods for geographic coordinates (section 8.3)

[geospatial-jeff]

Question

I'm looking for a NetCDF file, in the wild, that uses quadratic interpolation for geographic coordinates. From here:

A one-dimensional quadratic method for interpolation of the geographic coordinates latitude and longitude, typically used for remote sensing products with geographic coordinates on the reference ellipsoid.

I'm fairly convinced these don't exist. WGS84 is the most common geographic coordinate system, it has its own reference ellipsoid as well. WGS84 represents the world as an oblate-spheroid (an ellipsoid) which is quadratic by definition. But why would a user of CF conventions ever implement this quadratic themselves when they can define WGS84 in the grid_mapping variable? Seems like re-inventing the wheel to me.

I understand the value of section 8.3 is to reduce the number of coordinates stored in the file, but I would guess the complexity of implementing this quadratic yourself far outweighs the reduction in storage costs. Especially when you consider the implementation of "sub areas" (multiple sets of tie points) which gets particularly nasty when you are dealing with a quadratic.

[davidhassell]

Hello,

Whilst datasets encoded with CF's compression by coordinate subsampling may not yet exist, datasets with some different variety of compression by coordinate subsampling do (e.g. these). The CF technique (including the fancier subareas and interpolation parameters features) was developed to standardise what was already being done in the wild. The specifications for operational satellite products are fixed years in advance of data being available, so I would hope and expect that we will start to see its use in the future.

I don't think any wheels are being re-invented (circular or elliptical!). If you already have coordinates defined on the ellipsoid, or in any other coordinate system, compression by coordinate subsampling allows you to encode them in a different way that reduces the size of the dataset on disk, and that's all. For instance, some remote sensing datasets can be reduced in size by a factor of 100, whilst retaining an uncompressed coordinate accuracy to within 0.00002 degrees.

CF doesn't mind what the compression by coordinate subsampling technique is used for, since it doesn't change the logical content of the data.

Cheers,
David

[geospatial-jeff]

Thanks David this is all very helpful! I'll read through that metsat spec!