[Met_help] question

Fri Nov 6 12:31:23 MST 2009

Holly,

That's a good question about doing matching in a more sophisticated way.  I'd say that the answer now is no.  We don't currently have that in our future plans.

However, you may be able to perform similar types of stratifications now.  For example, if you have a nice land/sea mask, you could apply it to your data and perform verification separately for points
over land vs points over the sea.  Similarly, if you have a nice elevation field, you could threshold it to define elevation regions and use those to define the masking region.

In fact, WPP has the ability to dump out both a land/sea mask and elevation data on the same grid as your model data.  You could use those fields to define the masking regions.

Look at slides 39 and 40 in the following presentation: http://www.dtcenter.org/met/users/docs/presentations/MET_Tutorial_20090626.pdf
This discusses "Data Masking" where you choose one field and threshold it to define the verification masking region used for another field.  So you might take the elevation, threshold it less than
1000 meters and use that mask to verify some other variable.

Hope that helps.

John

Holly Hassenzahl wrote:
> Thanks for your help, John.  I had some trouble getting NCL to run
> properly.  But we'll figure something out!
> 
> Do you know if there are any future plans for MET to have the ability to
> handle elevation and land vs. water data?  The NDFD uses a "modified"
> nearest neighbor approach when they compute their own verifications.
> Basically, if the nearest neighbor is not within 500ft of the obs site,
> then they try the next nearest neighbor and so on.  Likewise, if the
> nearest neighbor is over water, they do not use it and move on to the
> next.  Someday, we'd like to get as close to that methodology as
> possible when doing our NDFD stats.
> 
> Thanks again,
> Holly
> 
> ___________________________
> Holly C. Hassenzahl
> Meteorologist, Science Analyst
> Weather Central, LLC
> 401 Charmany Drive, Ste 200
> Madison, WI 53719
> (608) 274-5789
> (608) 276-4613 Fax
> www.wxc.com
>  
> 
> -----Original Message-----
> From: John Halley Gotway [mailto:johnhg at ucar.edu] 
> Sent: Thursday, November 05, 2009 9:08 AM
> To: Holly Hassenzahl
> Cc: met_help
> Subject: Re: question
> 
> Holly,
> 
> Right now you are using the Point-Stat tool to compare a gridded
> forecast to point observations.  Point-Stat is the only statistics tool
> in MET that handles point observations.  The MODE tools
> (Grid-Stat and Wavelet-Stat too) can only be run to compare two gridded
> fields - typically a gridded forecast vs gridded observation or analysis
> field.
> 
> MET won't be able to help you generate a plot of the FCST - OBS values.
> If I were in your position, I'd probably do something like the
> following:
> 
> (1) Run my data through Point-Stat and turn on the MPR (matched pair)
> output line in the Point-Stat configuration file.  This dumps out 1 line
> of ASCII output for each FCST - OBS matched pair value
> that MET computed.
> 
> (2) Choose a plotting tool.  I'd probably use NCL (NCAR Command
> Language) or R (www.r-project.org).  But others, like IDL, may work too.
> 
> (3) Use that plotting tool to read the MPR output of Point-Stat.  Each
> MPR line contains lat/lon info.  So I'd process each line, use the
> lat/lon value to place it on a plot, and then plot the
> difference value.
> 
> Hope that helps.
> 
> John
> 
> Holly Hassenzahl wrote:
>> Hi John-
>>
>> Happy Tuesday!  Couple quick questions for you:  We would like to
> create
>> a graphical representation of the difference between the forecast and
>> actual max and min temps.  Is there a way to do this using the MODE
>> tool?  I assume we would have to run both files through PCP Combine
>> first?  And if so, how exactly would we do that since we're dealing
> with
>> temps and not precip?
>>
>> Thanks,
>> Holly
>>   
>>
>> ___________________________
>> Holly C. Hassenzahl
>> Meteorologist, Science Analyst
>> Weather Central, LLC
>> 401 Charmany Drive, Ste 200
>> Madison, WI 53719
>> (608) 274-5789
>> (608) 276-4613 Fax
>> www.wxc.com
>>  
>> -----Original Message-----
>> From: John Halley Gotway [mailto:johnhg at ucar.edu] 
>> Sent: Wednesday, October 28, 2009 2:50 PM
>> To: Holly Hassenzahl
>> Cc: met_help
>> Subject: Re: question
>>
>> Holly,
>>
>> OK, I see the problem.  The projection information is not correct in
> the
>> forecast NetCDF file: fcst18z_06L_TMAX.nc
>>
>>                 :Projection = "Lambert Conformal" ;
>>                 :p1_deg = "39.000000 degrees_north" ;
>>                 :p2_deg = "39.000000 degrees_north" ;
>>                 :p0_deg = "19.753263 degrees_north" ;
>>                 :l0_deg = "168.092865 degrees_east" ;
>>                 :lcen_deg = "-98.000000 degrees_east" ;
>>                 :d_km = "12.000000 km" ;
>>                 :r_km = "6371.200195 km" ;
>>                 :ny = "299 grid_points" ;
>>                 :nx = "447 grid_points" ;
>>
>> I ran this file through the MODE tool, and I just compared it to
> itself:
>> mode fcst18z_06L_TMAX.nc fcst18z_06L_TMAX.nc WrfModeConfig
>>
>> And then I looked at the PostScript file that was generated.  The data
>> sure looks like it's supposed to be over the CONUS, but the MODE
>> PostScript file shows it plotted somewhere else.  That's why
>> you're not getting any matched pairs - Point-Stat thinks that your
>> domain isn't over CONUS.  Take a look at the attached PNG file to see
>> where MODE plots it.
>>
>> I thought maybe that l0_deg = "168.092865 degrees_east" was supposed
> to
>> be "-168.092865 degrees_east" but that didn't do the trick.
>>
>> Do you have some numbers that define what the projection is supposed
> to
>> be?
>>
>> John
>>
>> Holly Hassenzahl wrote:
>>> Hello John-
>>>
>>>  
>>>
>>> I hope this email finds you well!  
>>>
>>>  
>>>
>>> I am having another problem with point stat not finding matched
> pairs.
>>> I have uploaded my files to the ftp site and would love some feed
> back
>>> whenever you get a chance.  The files are:
>>>
>>>  
>>>
>>> PointStatConfig_max12km
>>>
>>> obs_17z.nc
>>>
>>> fcst18z_06L_TMAX.nc
>>>
>>>  
>>>
>>> This was working a couple weeks ago, and is working correctly with a
>>> forecast file from a different model and valid for the same time.  I
>>> viewed both netcdf files and from what I can see, the valid times for
>>> TMAX match up correctly.  If there's something you see, please let me
>>> know.
>>>
>>>  
>>>
>>> Thank you!
>>>
>>> Holly
>>>
>>>  
>>>
>>> ___________________________
>>>
>>> Holly C. Hassenzahl
>>>
>>> Meteorologist, Science Analyst
>>>
>>> Weather Central, LLC
>>>
>>> 401 Charmany Drive, Ste 200
>>>
>>> Madison, WI 53719
>>>
>>> (608) 274-5789
>>>
>>> (608) 276-4613 Fax
>>>
>>> www.wxc.com
>>>
>>>  
>>>
>>>  
>>>
>>>