Description
The Global Change Observation Mission - Climate (GCOM-C) data have content like below (simplified):
group: Geometry_data {
variables:
float Latitude(161, 126)
string Unit = "degree"
string Dim0 = "Line grids"
string Dim1 = "Pixel grids"
int Resampling_interval = 10
float Longitude(161, 126)
string Unit = "degree"
string Dim0 = "Line grids"
string Dim1 = "Pixel grids"
int Resampling_interval = 10
}
group: Image_data {
variables:
ushort SST(1599, 1250) // Note: different size than (latitude, longitude) variables.
string dim0 = "Line grids"
string dim1 = "Pixel grids"
string Unit = "degree"
}
The size of latitude and longitude variables is not the same than the size of image data. In this case, even if reader correctly identifies Latitude and Longitude as the variables to use for building a localization grid, we are still unable to associate the SST variable to those axes because they have no dimension in common. However if we interpret dim0 and dim1 attributes as "Name of dimension 0" and "Name of dimension 1" respectively, then we can associate the same dimension names to all those variables: namely "Line grids" and {"Pixel grids"}} in above example. Using those names, we deduce that the (data_y, data_x) dimensions in the SST variable are mapped to the (grid_y, grid_x) dimensions in the localization grid.
The feature is an extension to CF-conventions, as I'm not aware of equivalent mechanism in CF at this time. The coordinates attributes is not exactly the same since it tells us which variable to use, but not which dimensions (an ambiguity still exists if the variable has 2 or more dimensions).
Attachments
Issue Links
- is depended upon by
-
SIS-443 Give access to netCDF raster data as GridCoverage
-
- Closed
-