[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: BUFR code

On Mon, 10 Jul 2006, Kevin Baggett wrote:

> Robb,
> Thanks for the BUFR code that you provided to me - it looks like it will
> save quite a bit of time.
> I have a couple of questions, tho.
> 1. I do not understand  how Indexing is supposed to work
> Index.java says to look at IndexFormat.txt which I can't find in what
> you gave me.

i haven't created that file yet

> Index.java also says that the index has all the info needed to write a
> netcdf object - that is of interest to me if there is code to do that
> already created.

yes, there is a toolsUI program to create netCDF object. go to:

http://www.unidata.ucar.edu/software/netcdf-java/v2.2/webstart/index.html

lauch netcdfTools with WebStart.  after loading there is a Viewer tab that
you should click on, then use the Folder tab to open a bufr file. there
is Binocular tab to click to give a cdl file. Also if you click on a
parmeter, then right click you can get a netCDF dump of a particular
variable. Also the is a Point tab you should try to open a bufr file.
i haven't coded all the different type of Bufr file, just the Profiler,
Trajectory, and SoundingVerticalSatellite data files.

your will have to download the netcdf-java source at:
http://www.unidata.ucar.edu/software/netcdf-java/

to look at the code.  the bufr code is located at:
<distribution>/src/ucar/nc2/iosp/bufr

> 2. I have used BufrDump.java to output a text file containing BUFR data.
>  The output has size and varCount for each variable, but, for a certain
> number of datasets, will "size/num datasets=varCount" or
> "size/varCount=num datasets" always be a true statement?
"size/varCount=num datasets" should be a true

> I examined some BUFR files (quite painful) and it seems to me that the
> replication factor could be different for each dataset.

that's true especially in the Profiler data sets, that's what makes bufr
hard to handle.

> By itself, for the particular sample file I have, BufrDump outputs,
> say,  all the latitudes grouped together for the number of datasets, but
> there is no way to output
>  a separate grouping of each datasets lat/lon's unless you assume the
> equations above hold.
> So, is it possible for a group of datasets to go, from, say, 100 lat
> observations for dataset 1 to 50 lat observations in dataset 2?
> Not sure if I am totally clear about what I mean here, so I will include
> an example.
> 0-5-2 Latitude size =704 varCount =64
>  -56.13 -55.93 -55.75 -55.56 -55.38 -55.2 -55.02 -54.85 -54.69 -54.52
> -54.36 -54.21 -54.06 -53.91 -53.77 -53.63 -53.49 -53.37 -53.24 -53.12
> -53.0 -52.89 -52.79 -52.68 -52.59 -52.5 -52.41 -52.33 -52.25 -52.18
> -52.11 -52.05 -51.99 -51.94 -51.89 -51.85 -51.82 -51.79 -51.76 -51.74
> -51.72 -51.71 -51.71 -51.71 -51.72 -51.73 -51.74 -51.77 -51.79 -51.82
> -51.86 -51.9 -51.95 -52.0 -52.06 -52.12 -52.19 -52.27 -52.34 -52.43
> -52.51 -52.61 -52.7 -52.81 -55.91 -55.72 -55.53...etc.
>
> ->Latitude starts over here
> So, is it safe to assume, I can go 64 count for the latitude in this
> instance and so the next one will also be 64-count?  Or can the data do
> something funny with its replication factor in the data and go 63
> records one time, and 65 records the next and still add up to 704 in the
> end?

if there are more then one data set in a record, then all the variable
sizes have been the same in all the datasets i have looked at. But bufr is
flexible enough that might not be true in the future. still checking on
it.


> 3. Related to the above, is there a way I can output BufrDump
> differently - each dataset's complete entries, rather than lumping a
> particular variable all together.

sure use ncdump from with in the toolsUI program. Also, you can change the
number of info you want back. ie. 0:25:1 gives records 0-25 with a stride
of 1. so 0:0:1 gives first record back.

Take the time to learn the toolUI program, it will give a good over view
of the bufr data. I'm especially interested in your finding because this
is just beta code.


robb...



>
> Thanks for your help!
> Kevin
>

===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III                      Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================