9

I obtained a netCDF raster file, but I couldn't get any metadata to get the name of the coordinate system the raster has been built on. The raster itself doesn't have any coordinate system embedded. I thought it would just be a WGS84, and it looked like that at first glimpse, but with further investigation in ArcMap, I saw that it is a rather not common system. Here is how it displays: enter image description here

The orange raster is a normal raster in WGS84 which I have inserted here for comparison purposes. The purple one is the raster with the unknown coordinate system. Do you have any clue what this might be?

Some updates: Here is the netCDF raster: https://www.dropbox.com/s/nottbl9yt6dwss6/sic_average_nclimate.nc?dl=0 I was also able to get some metadata from the image provider:

netcdf sic_average_nclimate {
dimensions:
    nlon = 361 ;
    nlat = 90 ;
    nseas = 4 ;
variables:
    float SIC_Change(nlat, nlon) ;
        SIC_Change:Title = "Gridded Multi-Model Ensemble Mean Annual Mean Change in Ice Concentration 21C-20C" ;
    float SIC_Season_Change(nseas, nlat, nlon) ;
        SIC_Season_Change:Title = "Gridded Multi-Model Ensemble Mean Seasonal Mean Change in Ice Concentration 21C-20C" ;
    float SIC_Change_STD(nlat, nlon) ;
        SIC_Change_STD:Title = "Gridded Multi-Model Standard Deviation of the Annual Mean Change in Ice Concentration 21C-20C" ;
    float SIC_Season_Change_STD(nseas, nlat, nlon) ;
        SIC_Season_Change_STD:Title = "Gridded Multi-Model Standard Deviation of the Seasonal Mean Change in Ice Concentration 21C-20C" ;
    float LAT(nlat) ;
        LAT:Title = "Latitude" ;
    float LON(nlon) ;
        LON:Title = "Longitude" ;

// global attributes:
        :Title = "Ice Concentration metrics for Model subset as in Figure 1 of NCLIMATE paper" ;

They show the boundary lat-long, but apparently not any information regarding the coordinate system.

  • 1
    Antarctica is usually in Polar Stereographic Projection so everything is going north from south pole. nsidc.org/data/polar_stereo/ps_grids.html – Mapperz Dec 16 '14 at 15:06
  • 1
    Weird, looks like 0-360, but why is it plotting in the northern hemisphere? – mkennedy Dec 16 '14 at 17:12
  • 1
    If the reported extents in the NetCDF header look like lat/lon, but are lat: 0 to ?, lon: 0 to 360. Can you hack them to lat: -90 to (? - 90) and lon: -180 to +180? – mkennedy Dec 16 '14 at 18:33
  • @Mapperz & mkennedy, I tried to assign it a Polar Stereographic Projection, but it didn't work. I can probably mirror it vertically downwards, but it's right side half will be still beyond 180E longitudes. I opened the netCDF with notepad, but could not find any information related to lat,long. I am attaching the raster in the question if you're curious to open it on your side. – multigoodverse Dec 17 '14 at 8:34
  • That link doesn't work - dropbox says "the owner hasn't authorised access". – BradHards Dec 19 '14 at 9:43
5
+25

You can shift the data into the correct position using these GDAL commands:

gdal_translate -a_srs EPSG:4326 -a_ullr 0 0 360 -90 NETCDF:"sic_average_nclimate.nc":SIC_Change change360.tif
gdalwarp -t_srs WGS84 change360.tif change180.tif  -wo SOURCE_EXTRA=1000 --config CENTER_LONG 0

(with a little help by Frank Warmerdam: How to reproject raster from 0 360 to -180 180 with cutting 180 meridian)

and the result looks like this in a stereographic projection:

enter image description here

1

(This isn't an answer, but rather a process suggestion that is too long to be a comment)

You could geo-reference it in QGIS using the GeoReferencer tool as you appear to be able to identify common points on both rasters. This would generate a GDAL script with a set of GCP points, and transformation parameters. You could add this script with a sample of these points to your question, as it might explain what the difference between the two projections is.

Worst case scenario, you would have a rough estimate of the values based on the repositioned raster which might help you with your goal.

  • + 1 and yes, georeferencing might be the last bullet, but the process would yield some errors because it's difficult to match control points. Additionally, the netCDF raster would have to be split because it does not have the same prime meridian. For example, the tail of Antarctica is on the right, while the orange raster shows it on the left. – multigoodverse Dec 19 '14 at 11:59
0

The lower left corner is at (0,0) coordinates, but for me (0,0) is the coordinate of the upper right corner in your case and I don't know any geographic coordinate system with its origin at -90.

As you work in ArcGIS, you could use the georeferencing toolbar to move your data at the right location. Move the upper right to (0,0) and the lower left to (-360, -90) and this should yield an exact transform but you would still have a problem with the dateline.

enter image description here

You can thus try this link which should help you to create a custom geographic coordinate system. To sum it up, in your reference system, create a custom WGS 84 with the prime meridian at -180, then a custom transform using longitude rotation.

enter image description here

Georeferencing in this new system would consist in setting the origin at (-180, -90). One control point is enough because you just need a translation. Carefully select the bottom left pixel, right click and add the exact XY values). Then it should look like this

enter image description here

As a final remark, there is no certainty that the coordinate system is based on WGS 84. But it is for sure a geographic coordinate system in degrees and the resolution of your raster is probably coarser than the potential error due to a wrong datum selection.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.