You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 21 Next »

netCDF/OPeNDAP are the perfect file format/web service for gridded data. Gridded data exist in various flavours of complexity.

Grid topology

  • Orthogonal grid
    • An orthogonal lat-lon grid: climate data products often appear in this topology. There is a separate lat vector (e.g. -180:1:180 deg) and a separate lon vector (e.g. -90:1:90 deg) that span a part of the globe. Support for these grids is always available. Example:
      http://opendap.deltares.nl/thredds/catalog/opendap/deltares/FEWS-IPCC/catalog.html

      The minimal netCDF dump of this grid type looks lile 
      dimensions:
	lon = 3 ;
	lat = 5 ;
variables:
	single VARIABLE(lon,lat), shape = [3 5]
		VARIABLE:grid_mapping = "wgs84" 
\\ lat-lon vectors with mapping info
	single lon(lon), shape = [3]
	single lat(lat), shape = [5]
	int32 wgs84([]), shape = [1]
    • An orthogonal x-y grid: national data products often appear in this topology. There is a separate x vector (e.g. 0:100:280000 m)
      and a separate y vector (e.g. 624800:-100:300000) that span a part in a local projection. Support to plot these grids in lat-lon space is less often available, as it assumes that the client can deal with the two methods to prescribe lat-lon coordinates. The first, implicit option is assuming that the client can do the conversion from the local projection to the a WGS84 lat-lon projection. The CF convention describes how to add projection meta-data to the netCDF file (grid_mapping attribute). In addition, the data producer can transform the x and y vectors to complete lat and lon grids, prescribing the WGS84 position of each grid cell separately. The CF convention describes how to attach these global coordinates to the dataset (coordinates attribute). OpenEarth recommend to supply both options in netCDF files. This option is in fact the same as considering the orthogonal x-y grid as a curvi-linear lat-lon grid, as discussed next. Example:
      http://opendap.deltares.nl/thredds/catalog/opendap/tno/ahn100m/catalog.html with clearly defined local projection (Dutch RD) and with full lat-lon matrices included.
      http://opendap.deltares.nl/thredds/catalog/opendap/knmi/NOAA/mom/1990_mom/5/catalog.html without clearly defined local projection (a polar stereographic one) but with full lat-lon matrices included.
  • Curvi-linear grid: plotting support for curvi-linear grids depends on the client. Curvi-linear grids often exist, key examples are: (info) individual satellite images that are orthogonal in pixel space, but fully warped in lat-lon space and (ii) curvi-linear coastal circulation models that are orthogonal in grid index space, but that have been warped to fit local coast lines.
  • Unstructured grid: Support for unstructured grids has not been standardized. A proposals can be found at the Deltares netCDF wiki

Single grid vs. tiled grids

For large areas with high resolutions often single netCDF files become too big. Also when there are large parts with _Fillvalue, netCDF3 files can become too large. netCDF4 of course solves this issue by allowing for internal zipping of variables, but not all applications can handle netCDF4 yet. So tiling of the full coverage into sub netCDF files is a solution.

>> url = 'http://opendap.deltares.nl/thredds/catalog/opendap/rijkswaterstaat/vaklodingen/catalog.xml'
>> L = opendap_catalog(url)
>> disp(L{1})
ans = http://opendap.deltares.nl/thredds/dodsC/opendap/rijkswaterstaat/vaklodingen/vaklodingenKB138_0706.nc
>> nc_dump(L{1})

Time dependency

Most 'GIS'-minded people produce grids without a time dimension, but simply with some kind of time annotation. This does not easily allow to automatic processing of temporal data. Instead, model output, spatially binned remote sensing products and repeated depth sounding products do produce datasets with a clear and explicit time dimension.

Password restrictions

  • No labels

Disclaimer