[gradsusr] Open GRIB Files over THREDDS Opendap Server

[Jennifer M Adams]

> On Feb 1, 2018, at 1:48 PM, William Daniel Scheftic <scheftic at email.arizona.edu> wrote:
> 
> The behavior seems to be intermittent.  I tried to open my larger ctl file with most of the variables.  First time it failed. Second time with no changes it opened fine.  I had another connection failure on a data request and then retried it with success.
> 
> I have noticed it's much slower to load the ctl file and plot the variables through this server.  The NCEP GDS server runs very fast for even larger data requests.  Does the THREDDS opendap server do the same work for splicing the data prior to sending it as GDS does?

That is unfortunate behavior! I don’t know how THREDDS servers work. NCEI stopped using GDS operationally because the resources were not available to support updates to GDS so it could meet their needs. It is a bummer for all of us.

I do know that if you look at pre-projected data behind a THREDDS server (say, for the regional model output that is on a lambert grid) the I/O with GrADS will be painfully slow, making that data essentially unusable. That is because they serve the data on the native grid, and Grads must do the PDEF interpolation on the client side, so the I/O is done one grid box at a time, four times per grid point. Fixing that would require some significant changes to the I/O layer in GrADS; it is on my to-do list, but has been back-burnered in favor of other explicitly-funded enhancements. 

> I'll have to check with NCEI because I got an error with an xml dump when trying to plot the same variable in GrADS.
Give them as much information as you can about when/how things went wrong and what tools you were using to access the server. I hope they can help. 

> 
> Also, I'm a little confused about how to extract levels from non-pressure data using the same ctl file.  Can I do this in my ctl file?
> Storm_relative_helicity_height_above_ground_layer_136_Hour_AverageNanalysis=>hlcy0_1000m 0 t,0,y,x ** 0 to 1 km storm rel. helicity
> Storm_relative_helicity_height_above_ground_layer_136_Hour_AverageNanalysis=>hlcy0_3000m 0 t,1,y,x ** 0 to 3 km storm rel. helicity
> The vertical coordinate for this variable is:
>  height_above_ground_layer = 500, 1500 ;
> 
>  height_above_ground_layer_bounds =
>   1000, 0,
>   3000, 0 ;

Yes. I can draw those two grids.
—Jennifer

> 
> 
> 
> > If my sample descriptor works just fine, then the problem is not with GrADS.  Using template strings with URLs is no problem. Just be aware that you’re doing I/O over the internet, and it may be slow. Use define/fwrite/sdfwrite to copy data you’re going to look at more than once to local memory/disk.
> >
> > Since ncdump dies midway through parsing the file, then I would check with tech support at NCEI. The netcdf errror message points to line 473 from vardata.c, which contains the following library call:
> >
> >     NC_CHECK(nc_get_vara(ncid, varid, cor, edg, (void *)valp));
> >
> > Something is funky with that variable; the data owners should be able to track it down.
> > —Jennifer
> 
> _______________________________________________
> gradsusr mailing list
> gradsusr at gradsusr.org
> http://secure-web.cisco.com/1f_X3jXFjQMgQ36l_GyvlW7BpzX582Yk1voOE4_cfHI9NQ-Sut0G2aWZbx4L3vSp39sBiFnZxf1ky23a7-zPqmsAo0M1_OzQYlBTExR6LYQilzi7Z1u2W80j_AnCXIoBHI8pw5EvX9saFyThhVfyNmyGF6RymnRxgU3zk6n9hR8Ac7G7FdEFraNzmX0lnl83q3QcgX4heax5l7GpU7ngxcnnEwCjfvJEbS0wX9CJ-lRjNtIhQg9DNJUTf6ifLngHbXblzaxpx6UuMLkiJtMgcuTYAmYtNvp3IqdkCj7VhpBNxnakBz8OudEc8wUalkGpv5m3Qu4whhjA_XEI9zHwt6tukfDr1mMxgphdZ1C23QsAeQCW-6OyrCc733TIRl8oC/http%3A%2F%2Fgradsusr.org%2Fmailman%2Flistinfo%2Fgradsusr

--
Jennifer Miletta Adams
Center for Ocean-Land-Atmosphere Studies (COLA)
George Mason University