On 9/7/2012 8:01 AM, Ellyn Montgomery wrote:
for example, i changed to:
S_40(0:105483:1000, 0:0:1, 0:0:1, 0:0:1)
How did you modify S_40 to limit the display? In the above, my only
option is to click NCdump, and everything gets listed in the little
NCdump box.
You directly edit the string "S_40(0:105483:1000, 0:0:1, 0:0:1, 0:0:1)"
in the combo box.
http://geoport.whoi.edu/thredds/dodsC/MBAY_LT/7551mc-a1h.nc is missing
the variable T_20. If you put the ncl file in the "Aggregation" tab,
then hit the "check files" buttom on upper right, you get a report
(see below). there are enough problems to suggest that this collection
of files is not suitable for a "joinExisting" aggregation, which
requires (almost) exact files except or the aggregation coordinate.
I think what you said is that although the aggregation I'm attempting
doesn't crash and burn, it's really failing because the files in the
joinExisting don't have the same content (variables). Best practice
should be to do "check files" first and ensure there aren't a lot of
errors about missing variables.
In theory then, if I removed the files with missing temperature, and
only aggregated on T_20 (leaving out all the other variables that may or
may not be present) it would work?
i think so
So if I want to aggregate temperature from files with variables called
T_20, T_24, T_28 (all with standard_name = sea_water_temperature), I'd
need to rename the variables in ncml, put that new catalog in TDS, then
aggregate from the new spot?
yes
Or is doing a 2-part aggregation a better approach? step 1 might be
aggregating the three names for temperature separately (one for T_20,.
one for T_24, and one for T_28), renaming the output variable to the CF
name each time, then aggregate the 3 files to get one long temperature
record?
i wouldnt do a nested aggregation unless its really needed
also:
1) the extra lat,lon dimension on the variables is actually incorrect.
however it "works" because its equal to 1,1.
We're in the process of re-doing the ncml to represent the data as
discrete samples, and the extra lat and lon dimensions will be removed
there. We kept our data as "grid" representation because in earlier CF
versions I couldn't get PointFeatures to handle the multiple depths in
adcp data correctly.
2) if you are running a TDS server >= 4.2, enabling the cdmremote
service and using it for remote aggregations like this will speed up
response time.
We'll try this.
Thanks again for the help!
you're welcome!
John
