Re: [thredds] Aggregation of virtual datasets

Hi Rich!

I tried your first suggestion and it was fine to me.

Benno

i know that this data is served aggregated at pcmdi, i just want to make a local server with some data to improve the connection.

thanks folks!



Quoting Rich Signell <rsignell@xxxxxxxx>:

Benno,

On Thu, Dec 30, 2010 at 2:46 PM, Benno Blumenthal
<benno@xxxxxxxxxxxxxxxx> wrote:
FYI - that data is already served aggregated at pcmdi

Where?

one of the mistakes
that one needs to watch for is time coordinates (particularly time bounds)
where the units are different between the three files, e.g. each file starts
at time zero but that value has a different offset in the units.

This case can be handled in NcML by adding timeUnitsChange="true" to
the aggregation specification:

<aggregation dimName="time" type="joinExisting" timeUnitsChange="true">

-Rich


On Dec 28, 2010 9:33 AM, <tnb@xxxxxxxxxxxxxxxx> wrote:

Hi everybode!

I would like to make a local server for some IPCC data and would like to
know if is there any way to aggregate (type "union") virtual data (construct
with aggregate "joinExisting") with netcdf files.

Let me explain with an example:

I have 3 variables, all monthly data, sea surface temperature, salinity and
sea surface height, each one in a diferent directory.

SST are 3 files, sst_1970_1980.nc, sst_1980_1990.nc, sst_1990_2000.nc

salinity are 3 files too, sal_1970_1980.nc, sal_1980_1990.nc,
sal_1990_2000.nc

SSH is 1 file, ssh_1970_2000.nc

Each variable have the same lon, lat axis and in total, all have 30 X 12 =
360 timesteps.

I tried to do a nested aggregation, but it doesn t work...


 <dataset name="IPCC data" ID="ippc_agg" urlPath="ipcc/agg.nc">
   <serviceName>odap</serviceName>
   <netcdf xmlns="http://www.unidata.ucar.edu/namespaces/netcdf/ncml-2.2";>
   <aggregation dimName="time" type="joinExisting">
   <netcdf>
   <aggregation type="union">
     <netcdf location="/data/IPCC/20c3m/ocn/mo/thetao/sst_1970_1980.nc"/>
     <netcdf location="/data/IPCC/20c3m/ocn/mo/so/sal_1970_1980.nc"/>
   </aggregation>
   </netcdf>
   <netcdf>
   <aggregation type="union">
     <netcdf location="/data/IPCC/20c3m/ocn/mo/thetao/sst_1980_1990.nc"/>
     <netcdf location="/data/IPCC/20c3m/ocn/mo/so/sal_1980_1990.nc"/>
   </aggregation>
   </netcdf>
   <netcdf>
   <aggregation type="union">
     <netcdf location="/data/IPCC/20c3m/ocn/mo/thetao/sst_1990_2000.nc"/>
     <netcdf location="/data/IPCC/20c3m/ocn/mo/so/sal_1990_2000.nc"/>
   </aggregation>
   </netcdf>
   <netcdf>
   <aggregation type="union">
     <netcdf location="/data/IPCC/20c3m/ocn/mo/zos/ssh_1970_2000.nc"/>
   </aggregation>
   </netcdf>
   </aggregation>
   </netcdf>
 </dataset>


... I expected that the aggregate data, agg.nc had 3 vars, sst, sal and ssh
and each one with a 360 timesteps axis. But what happens was that agg.nc
have only 2 vars, sst and sal and their time axis have 720 timesteps.

I didn t find any information about how to do this kind of aggregation and
stay wondering how could it be done.

Thanks for attention and happy new year for all!

Tiago Bomventi



_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe,  visit:
http://www.unidata.ucar.edu/mailing_lists/

_______________________________________________
thredds mailing list
thredds@xxxxxxxxxxxxxxxx
For list information or to unsubscribe,  visit:
http://www.unidata.ucar.edu/mailing_lists/




--
Dr. Richard P. Signell   (508) 457-2229
USGS, 384 Woods Hole Rd.
Woods Hole, MA 02543-1598