[Met_help] Request for Assistance-UPDATE (UNCLASSIFIED)

John Halley Gotway johnhg at ucar.edu
Wed Apr 28 10:54:44 MDT 2010


John,

Glad that did the trick.  Just to let you know, I did include that change in the development version of MET.  So Point-Stat should behave the way you're expecting in the next release.

John

Raby, John (Civ, ARL/CISD) wrote:
> Classification: UNCLASSIFIED
> Caveats: NONE
> 
> John and Bob - 
> 
> We just tested the recompiled version and it is working fine! We receive
> warnings vice errors and the results are being output as expected. Thanks
> for your help.
> 
> R/
> John and Bob
> 
> 
> -----Original Message-----
> From: John Halley Gotway [mailto:johnhg at ucar.edu] 
> Sent: Wednesday, April 28, 2010 9:41 AM
> To: Raby, John (Civ, ARL/CISD)
> Cc: Flanigan, Robert (Civ, ARL/CISD); met_help; Brown, Bob (Civ, ARL/CISD)
> Subject: Re: [Met_help] Request for Assistance-UPDATE (UNCLASSIFIED)
> 
> John,
> 
> So it does sound like the patch I sent for some reason hasn't been applied
> to the binary you're running.  I've re-attached that patched version of
> point_stat.cc.  Please do the following:
> - Use the attached version of point_stat.cc to overwrite the file
> METv2.0/src/point_stat/point_stat.cc.
> - In the top-level METv2.0 directory, do a "make clean" to remove the
> existing executables in METv2.0/bin
> - In the top-level METv2.0 directory, do a "make" to rebuild those
> executables.
> - Try running Point-Stat on the command line by pointing to a real forecast
> file and bogus observation files.  Something like:
>    METv2.0/bin/point_stat real_fcst.grb bogus_obs1.nc PointStatConfig
> -ncfile bogus_obs2.nc
> 
> You should see 2 WARNING messages like this:
> WARNING: process_obs_file() -> can't open observation netCDF file:
> bogus_obs1.nc
> WARNING: process_obs_file() -> can't open observation netCDF file:
> bogus_obs2.nc
> 
> But if you continue to see ERROR messages instead of WARNING messages, then
> something funny is going on.
> 
> Thanks,
> John
> 
> Raby, John (Civ, ARL/CISD) wrote:
>> Classification: UNCLASSIFIED
>> Caveats: NONE
>>
>> John -
>>
>> We recompiled MET with the fix yesterday and then tested it. We still 
>> are getting the error rather than the warning. Please call us at (575)
> 678-2004.
>> Thanks.
>> John
>>  
>>
>> -----Original Message-----
>> From: John Halley Gotway [mailto:johnhg at ucar.edu]
>> Sent: Tuesday, April 27, 2010 3:43 PM
>> To: Raby, John (Civ, ARL/CISD)
>> Cc: Brown, Bob (Civ, ARL/CISD); met_help at ucar.edu
>> Subject: Re: [Met_help] Request for Assistance-UPDATE (UNCLASSIFIED)
>>
>> John,
>>
>> I believe the problem you're seeing should also be solved by the fix I
> sent.
>> Basically, you were calling Point-Stat by passing it two observation 
>> files, let's call them file1.nc and file2.nc.  The fact that one was 
>> in the "mandatory" position and one was in the "optional" position 
>> (using -ncfile
>> argument) should be irrelevant.  In the code, they're just put into a 
>> list of files.  When processing each of those observation files in the 
>> list, originally Point-Stat would error out if one of the files didn't 
>> exist with a message like:
>>    ERROR: process_obs_file() -> can't open observation netCDF file:
>> ../out/ascii2nc/sample_ascii2.nc
>>
>> However, the change I sent you replaces that error with a simple 
>> warning message.  Now, rather than erroring out because of a 
>> non-existent file, it'll print a warning and continue processing the 
>> other files.  The warning looks something like this:
>>
>>    WARNING: process_obs_file() -> can't open observation netCDF file:
>> ../out/ascii2nc/sample_ascii2.nc
>>
>> Please try running with that fix and see if it addresses this issue.
>>
>> Hope that helps.
>>
>> John
>>
>> Raby, John (Civ, ARL/CISD) wrote:
>>> Classification: UNCLASSIFIED
>>> Caveats: NONE
>>>
>>> John -
>>>
>>> We are confused as our sysadmin, Bob Flanigan, was in the process of 
>>> compiling and found that it appears that the patch was already 
>>> compiled on
>>> 21 APR 2009. He is going ahead with the recompile to make sure it's 
>>> implemented.
>>>
>>> Yesterday, we found a mirror image of the same problem. This problem 
>>> was a situation where the required data was fully populated for every 
>>> hour, but the optional data was mostly empty (1 out of 25 hourly 
>>> files).  Point-Stat only produced results for the 1 hour where there 
>>> was optional and required data.
>>>
>>> It appears that Point-Stat will only produce results where both 
>>> datasets have data for a given hour.
>>>
>>> Please call us at (575) 678-2004.
>>>
>>> Thanks.
>>>
>>> R/
>>> John
>>>
>>>
>>> -----Original Message-----
>>> From: John Halley Gotway [mailto:johnhg at ucar.edu]
>>> Sent: Monday, April 26, 2010 10:40 AM
>>> To: Raby, John (Civ, ARL/CISD)
>>> Cc: Brown, Bob (Civ, ARL/CISD); met_help at ucar.edu
>>> Subject: Re: [Met_help] Request for Assistance-UPDATE (UNCLASSIFIED)
>>>
>>> John,
>>>
>>> Yes, it should print a warning message to the screen when an 
>>> observation file is missing, and then process the observations in any 
>>> additional observation files.  I definitely think logging the output 
>>> of the MET tools is a good idea.  Please let me know if you 
>>> experience any more issues or need more clarification.
>>>
>>> Thanks,
>>> John
>>>
>>> Raby, John (Civ, ARL/CISD) wrote:
>>>> Classification: UNCLASSIFIED
>>>> Caveats: NONE
>>>>
>>>> John -
>>>>
>>>> Thanks for the call last Friday and for your quick analysis and
> solution.
>>>> With the fix, you say it should print a warning then just continue 
>>>> processing instead of exiting. Does "continue processing" mean it 
>>>> will go ahead and process any additional observation files for that 
>>>> same hour passed via the optional argument?
>>>>
>>>> Since we were calling Point-Stat 25 times in succession, we were 
>>>> unable to note the writing of errors in the info which was scrolling 
>>>> on the screen. We have re-written our script to log the output and 
>>>> error statements so we can go back and read it.
>>>>
>>>> We will let you know about the results after recompiling with your fix.
>>>>
>>>> R/
>>>> John
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: John Halley Gotway [mailto:johnhg at ucar.edu]
>>>> Sent: Friday, April 23, 2010 3:22 PM
>>>> To: Raby, John (Civ, ARL/CISD)
>>>> Cc: Brown, Bob (Civ, ARL/CISD); met_help at ucar.edu
>>>> Subject: Re: [Met_help] Request for Assistance-UPDATE (UNCLASSIFIED)
>>>>
>>>> John,
>>>>
>>>> OK, I've made a minor change in the Point-Stat source code.  The 
>>>> updated version is attached.
>>>>
>>>> Here's the original piece of code:
>>>>       cerr << "\n\nERROR: process_obs_file() -> "
>>>>            << "can't open observation netCDF file: "
>>>>            << obs_file[i_nc] << "\n\n" << flush;
>>>>       exit(1);
>>>>
>>>> And here's the modified piece of code:
>>>>       cout << "\n\nWARNING: process_obs_file() -> "
>>>>            << "can't open observation netCDF file: "
>>>>            << obs_file[i_nc] << "\n\n" << flush;
>>>>       return;
>>>>
>>>> When one of the observation files passed to it doesn't exist, 
>>>> instead of exiting with an error message, it'll just print a warning 
>>>> message and process the remaining observation files.  I suspect that 
>>>> if you look closely at the output for those hours where the PREPBUFR 
>>>> file does not exist, you'll find an error message alerting you to that
> fact.
>>>> Please try using this version of the code, recompile MET, and rerun.  
>>>> And please let me know if this solves your problem.
>>>>
>>>> Thanks,
>>>> John
>>>>
>>>> Raby, John (Civ, ARL/CISD) wrote:
>>>>> Classification: UNCLASSIFIED
>>>>> Caveats: NONE
>>>>>
>>>>> John -
>>>>>
>>>>> We are unable to put together the data you requested now. 
>>>>>
>>>>> Ignoring the source of the observational (PrepBUFR or MADIS) the 
>>>>> main issue is that if the required data argument contains no data 
>>>>> for a given hour the software ignored the optional data for that 
>>>>> same
>> hour.
>>>>> Conversely, if there was required data for a given hour, then the 
>>>>> optional data for that hour was ingested and used by Point-Stat. In 
>>>>> short, the software refuses to add optional data to missing 
>>>>> required data
>>>> field.
>>>>> Why should the lack of data in the required data for a given hour 
>>>>> preclude the ingest of the optional data for that same hour?
>>>>>
>>>>> Due to the complexity of all of this, it would probably help to 
>>>>> discuss it verbally. Please call me at this number (575) 678-2004.
>>>>>
>>>>> Thanks.
>>>>>
>>>>> John
>>>>>
>>>>>
>>>>> -----Original Message-----
>>>>> From: John Halley Gotway [mailto:johnhg at ucar.edu]
>>>>> Sent: Friday, April 23, 2010 10:35 AM
>>>>> To: Raby, John (Civ, ARL/CISD)
>>>>> Cc: Brown, Bob (Civ, ARL/CISD); met_help at ucar.edu
>>>>> Subject: Re: [Met_help] Request for Assistance-UPDATE 
>>>>> (UNCLASSIFIED)
>>>>>
>>>>> John,
>>>>>
>>>>> The order in which you pass point observation files to Point-Stat 
>>>>> should have no impact on how the matched pairs are accumulated.  
>>>>> For each observation file passed to Point-Stat (the mandatory one 
>>>>> or additional optional ones), we just call the "process_obs_file"
>>>>> routine once.  If the behavior of Point-Stat changes depending on 
>>>>> the order of the point observation files, there may be a bug.
>>>>>
>>>>> I'm not entirely clear on the exact details of your situation.  
>>>>> Would you be able to provide me with an example of running 
>>>>> Point-Stat two different ways, where changing the order of the 
>>>>> point observation files has an impact on the results?  The simpler 
>>>>> you're able to make the example, the easier it'll be to debug the
> issue.
>>>>> For this, I'd need:
>>>>> - 1 forecast file
>>>>> - 2 point observation files
>>>>> - 1 Point-Stat configuration file
>>>>> - 1 example of how you call Point-Stat on the command line
>>>>>
>>>>> And you could post them to our ftp following the directions I sent 
>>>>> previously.
>>>>>
>>>>> Thanks,
>>>>> John
>>>>>
>>>>> Raby, John (Civ, ARL/CISD) wrote:
>>>>>> Classification: UNCLASSIFIED
>>>>>> Caveats: NONE
>>>>>>
>>>>>> John -
>>>>>>
>>>>>> UPDATE: 
>>>>>>
>>>>>> We solved the problem by processing the PrepBUFR data to include 
>>>>>> the metar data (ADPSFC) in addition to the ADPUPA" and "ANYAIR"
>>>>>> data mentioned in the email below. This ensured that there were 
>>>>>> all
>>>>>> 25 files of data present data provided in the required argument.
>>>>>>
>>>>>> We changed the MADIS data collection to remove the metar data and 
>>>>>> only receive the mesonet data and added this using the optional 
>>>>>> data argument (-ncfile).
>>>>>>
>>>>>> Now the Point-Stat results are produced for all 25 hours (06Z-06Z).
>>>>>>
>>>>>> In summary, if the required PrepBUFR data argument contains no 
>>>>>> data for a given hour the software ignored the optional MADIS data.
>>>>>>
>>>>>> Is this supposed to happen? Why should the lack of data in the 
>>>>>> PrepBUFR data (required argument) preclude the ingest all 25 hours 
>>>>>> of the
>>>>> optional data?
>>>>>> Please call me if we can clarify more for you or let me know if 
>>>>>> you still want the data ftp'd to you.
>>>>>>
>>>>>> Thanks.
>>>>>>
>>>>>> R/
>>>>>> John
>>>>>>
>>>>>> Mr John W. Raby, Meteorologist
>>>>>> U.S. Army Research Laboratory
>>>>>> White Sands Missile Range, NM 88011
>>>>>> (575) 678-2004 DSN 258-2004
>>>>>> FAX (575) 678-1230 DSN 258-1230
>>>>>> Email: john.raby at us.army.mil
>>>>>>
>>>>>> Please begin using the email address within this signature block 
>>>>>> for all future correspondence.  The current @arl.army.mil address 
>>>>>> will be decommissioned in upcoming months.
>>>>>>
>>>>>> "When you can measure what you are speaking about and express it 
>>>>>> in numbers, you know something about it, but when you cannot 
>>>>>> measure it, when you cannot express it in number, your knowledge 
>>>>>> is of a meagre and unsatisfactory kind". - Lord Kelvin
>>>>>>
>>>>>>  
>>>>>>
>>>>>> -----Original Message-----
>>>>>> From: John Halley Gotway [mailto:johnhg at rap.ucar.edu]
>>>>>> Sent: Thursday, April 22, 2010 8:58 AM
>>>>>> To: Raby, John (Civ, ARL/CISD)
>>>>>> Cc: met_help at ucar.edu
>>>>>> Subject: Re: Request for Assistance
>>>>>>
>>>>>> John,
>>>>>>
>>>>>> Sure, we'd be happy to help you figure this out.  Would you be 
>>>>>> able to send me some sample ASCII MADIS data, a sample forecast 
>>>>>> file, and the Point-Stat configuration file you're using?
>>>>>>
>>>>>> I'll try running ASCII2NC and Point-Stat here and figure out 
>>>>>> what's going on.
>>>>>>
>>>>>> You could post this sample data to our ftp site as follows:
>>>>>> ftp ftp.rap.ucar.edu
>>>>>> username=anonymous
>>>>>> password="your email address"
>>>>>> cd incoming/irap/met_help/raby_data put "your data files"
>>>>>> bye
>>>>>>
>>>>>> Please let me know when the data is available, and I'll go take a
> look.
>>>>>> Thanks,
>>>>>> John
>>>>>>
>>>>>>
>>>>>> We are attempting to use both PrepBUFR and MADIS observational 
>>>>>> data in MET Point_Stat to evaluate WRF performance over Utah for 2 
>>>>>> domains, one nested inside the other.
>>>>>>
>>>>>> We process the PrepBUFR data for a 24 hour period (06Z to 06Z) We 
>>>>>> explicitly retain only the message strings "ADPUPA" and "ANYAIR"
>>>>>> using the PB2NC configuration file settings in order to collect 
>>>>>> upper air data (raobs, aircraft, acars, etc). After running PB2NC, 
>>>>>> we typically find only 7 hourly files (12Z, 15Z, 17Z, 19Z, 20Z, 
>>>>>> 00Z and
>>>>>> 01Z) out of
>>>>>> 25 possible which contain data (presumably because the upper air 
>>>>>> data is not present for every hour).  These 7 include the 00Z and 
>>>>>> 12Z raob data for 2 upper air stations inside the larger of our 
>>>>>> domains (KSLC and
>>>>> KLKN).
>>>>>> We process the MADIS data to retain only surface mesonet and metar 
>>>>>> data in
>>>>>> 25 hourly text files (Point observation format), then run ASCII2NC 
>>>>>> to convert the files to netcdf.
>>>>>>
>>>>>> We ran Point-Stat with the PrepBUFR upper air data files specified 
>>>>>> in the required argument for the "obs_file" with the MADIS 
>>>>>> (surface
>>>>>> data) specified as additional NETCDF point observation files using 
>>>>>> the
>>>> "-ncfile"
>>>>> argument.
>>>>>> The Point-Stat output contains results for only the hours 12Z, 
>>>>>> 15Z, 17Z, 19Z, 20Z, 00Z and 01Z.
>>>>>>
>>>>>> We then tried running Point-Stat with the MADIS data specified in 
>>>>>> the required argument and the PrepBUFR data specified as the 
>>>>>> optional (additional observation file) argument, but this gave 
>>>>>> identical
>>> results.
>>>>>> We expect results for all 25 hourly periods since we know that 
>>>>>> there are MADIS surface observations for every hour.
>>>>>>
>>>>>> Can you assist us in resolving this problem?
>>>>>>
>>>>>> Thanks.
>>>>>>
>>>>>> R/
>>>>>> John
>>>>>>
>>>>>> Mr John W. Raby, Meteorologist
>>>>>> U.S. Army Research Laboratory
>>>>>> White Sands Missile Range, NM 88011
>>>>>> (575) 678-2004 DSN 258-2004
>>>>>> FAX (575) 678-1230 DSN 258-1230
>>>>>> Email: john.raby at us.army.mil
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> Classification: UNCLASSIFIED
>>>>>> Caveats: NONE
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> ------------------------------------------------------------------
>>>>>> -
>>>>>> -
>>>>>> -
>>>>>> -
>>>>>> --
>>>>>>
>>>>>> _______________________________________________
>>>>>> Met_help mailing list
>>>>>> Met_help at mailman.ucar.edu
>>>>>> http://mailman.ucar.edu/mailman/listinfo/met_help
>>>>> Classification: UNCLASSIFIED
>>>>> Caveats: NONE
>>>>>
>>>>>
>>>> Classification: UNCLASSIFIED
>>>> Caveats: NONE
>>>>
>>>>
>>> Classification: UNCLASSIFIED
>>> Caveats: NONE
>>>
>>>
>> Classification: UNCLASSIFIED
>> Caveats: NONE
>>
>>
> Classification: UNCLASSIFIED
> Caveats: NONE
> 
> 


More information about the Met_help mailing list