[Met_help] [rt.rap.ucar.edu #41159] History for WRF + MET
RAL HelpDesk {for Paul Oldenburg}
met_help at ucar.edu
Mon Oct 11 15:29:56 MDT 2010
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
Dear Sirs,
I'm writing my Bachelor's Thesis about the computation of the confidence
intervals at the Technical University of Vienna, Austria (www.*tuwien*.ac.at).
I'm quite newby with MET but I've some experience with WRF-ARW. Maybe my
question is quite dumm but I'd appreciate your help to clarify myself, in
order to get a more precise picture.
I'm trying to compare the output of WRF-ARW of some region of central Europe
with some measured data from several weatherstations of the same region. My
first task is the transformation of the values of the weathter stations
(with ASCII2NC) into NetCDF format (the values are provided in ASCII
format). Because of my measured data are not gridded, I suppose that the
best (or unique) possibility is the use of 'Point Stat' tool (am I right?).
The question is: how should I use the output of WRF-ARW? I mean: from the
description of the 'point_stat' tool (
http://www.dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_810/point_stat/index.php)
the forecasted data are from PCP-Combine, but I don't understand why must
one use PCP-Combine. Is it not possible to use directly the output from
WRF-ARW? Maybe my reasoning is wrong somewhere. Please, could you help me to
solve this question.
Thank you very much in advance and best regards from Vienna,
Ferdinand
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #41159] WRF + MET
From: Paul Oldenburg
Time: Tue Sep 28 10:54:25 2010
Ferdinand,
We assure you that your questions are not dumb, and we are happy to
help you. You are correct that the MET utility
point_stat is the most appropriate tool for verifying WRF-ARW output
using weather station data. To use the output of
WRF-ARW in MET, you must first use the WRF post-processor (WPP) to
generate GRIB model data. There are instructions for
how to do this in chapter 7 of the WRF-NMM User's Guide:
http://www.dtcenter.org/wrf-
nmm/users/docs/user_guide/V3/users_guide_nmm_chap1-7.pdf
Please note that the information presented in this guide will also
work for the ARW model. The pcp_combine tool is only
used if you need to change the precipitation accumulation intervals.
This step should be performed, if necessary, after
you run WPP. If you have any other questions, please let us know.
Thanks,
Paul
RAL HelpDesk {for Ferdinand Rennen} wrote:
> Tue Sep 28 10:20:54 2010: Request 41159 was acted upon.
> Transaction: Ticket created by ferdinand.rennen at gmail.com
> Queue: met_help
> Subject: WRF + MET
> Owner: Nobody
> Requestors: ferdinand.rennen at gmail.com
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41159 >
>
>
> Dear Sirs,
>
> I'm writing my Bachelor's Thesis about the computation of the
confidence
> intervals at the Technical University of Vienna, Austria
(www.*tuwien*.ac.at).
> I'm quite newby with MET but I've some experience with WRF-ARW.
Maybe my
> question is quite dumm but I'd appreciate your help to clarify
myself, in
> order to get a more precise picture.
>
> I'm trying to compare the output of WRF-ARW of some region of
central Europe
> with some measured data from several weatherstations of the same
region. My
> first task is the transformation of the values of the weathter
stations
> (with ASCII2NC) into NetCDF format (the values are provided in ASCII
> format). Because of my measured data are not gridded, I suppose that
the
> best (or unique) possibility is the use of 'Point Stat' tool (am I
right?).
>
> The question is: how should I use the output of WRF-ARW? I mean:
from the
> description of the 'point_stat' tool (
>
http://www.dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_810/point_stat/index.php)
> the forecasted data are from PCP-Combine, but I don't understand why
must
> one use PCP-Combine. Is it not possible to use directly the output
from
> WRF-ARW? Maybe my reasoning is wrong somewhere. Please, could you
help me to
> solve this question.
>
> Thank you very much in advance and best regards from Vienna,
>
> Ferdinand
------------------------------------------------
Subject: WRF + MET
From: Ferdinand Rennen
Time: Wed Sep 29 04:05:30 2010
Dear Paul,
thank you very much for your fast answer. It helped a lot. I've
installed
WPP. Now I can tranfsform the output of WRF-ARW into GRIB format and
compare
it with the measured data.
I have now a new problem (please, let me know is you consider that I'm
abusing of your help). My objetive is to estimate the confidence
intervals
of usual forecasted parameters (temperature, wind speed, etc.) at the
measured places, depending on the lapsed time of the forecast (let's
say 3,
6, 12, 24, 48 and 72 hours). I have access to the WRF-ARW prognoses
and the
measured data (hourly) of all the year 2009. The questions are:
1. In order to get the C.I. of each (3, 6, 12, 24, 48, and 72 hours)
forecasted parameter, should I compare the measurements with each one
of the
sampled times? If I'm right, should I compare all the year 6 times,
one for
each forecast periode, and I'd like to ask you, if you know a faster
method.
2. Taking advantage of the big amount of data, I think it could be
interesting to try to find some correlations between the seasonality
or the
difficulty of the forecast and the quality of the forecast. For sure
I'm not
the first in thinking about it. I'd appreciate your advice about my
trying.
Perhaps your have a tip or some references to reads.
Thank you very much and best regards,
Ferdinand
2010/9/28 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
> Ferdinand,
>
> We assure you that your questions are not dumb, and we are happy to
help
> you. You are correct that the MET utility
> point_stat is the most appropriate tool for verifying WRF-ARW output
using
> weather station data. To use the output of
> WRF-ARW in MET, you must first use the WRF post-processor (WPP) to
generate
> GRIB model data. There are instructions for
> how to do this in chapter 7 of the WRF-NMM User's Guide:
>
>
> http://www.dtcenter.org/wrf-
nmm/users/docs/user_guide/V3/users_guide_nmm_chap1-7.pdf
>
> Please note that the information presented in this guide will also
work for
> the ARW model. The pcp_combine tool is only
> used if you need to change the precipitation accumulation intervals.
This
> step should be performed, if necessary, after
> you run WPP. If you have any other questions, please let us know.
>
> Thanks,
>
> Paul
>
>
> RAL HelpDesk {for Ferdinand Rennen} wrote:
> > Tue Sep 28 10:20:54 2010: Request 41159 was acted upon.
> > Transaction: Ticket created by ferdinand.rennen at gmail.com
> > Queue: met_help
> > Subject: WRF + MET
> > Owner: Nobody
> > Requestors: ferdinand.rennen at gmail.com
> > Status: new
> > Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41159 >
> >
> >
> > Dear Sirs,
> >
> > I'm writing my Bachelor's Thesis about the computation of the
confidence
> > intervals at the Technical University of Vienna, Austria
(www.*tuwien*.
> ac.at).
> > I'm quite newby with MET but I've some experience with WRF-ARW.
Maybe my
> > question is quite dumm but I'd appreciate your help to clarify
myself, in
> > order to get a more precise picture.
> >
> > I'm trying to compare the output of WRF-ARW of some region of
central
> Europe
> > with some measured data from several weatherstations of the same
region.
> My
> > first task is the transformation of the values of the weathter
stations
> > (with ASCII2NC) into NetCDF format (the values are provided in
ASCII
> > format). Because of my measured data are not gridded, I suppose
that the
> > best (or unique) possibility is the use of 'Point Stat' tool (am I
> right?).
> >
> > The question is: how should I use the output of WRF-ARW? I mean:
from the
> > description of the 'point_stat' tool (
> >
>
http://www.dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_810/point_stat/index.php
> )
> > the forecasted data are from PCP-Combine, but I don't understand
why must
> > one use PCP-Combine. Is it not possible to use directly the output
from
> > WRF-ARW? Maybe my reasoning is wrong somewhere. Please, could you
help me
> to
> > solve this question.
> >
> > Thank you very much in advance and best regards from Vienna,
> >
> > Ferdinand
>
>
>
------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #41159] WRF + MET
From: Paul Oldenburg
Time: Wed Sep 29 11:06:14 2010
Ferdinand,
We assume that you have point observations that you want to compare to
model data using MET, since you refer to
"measured places". Is that correct? If so, it sounds like you should
run the MET tool point_stat for each valid time
to generate verification statistics. If you are interested in
handling the matched pair data (model & obs), you must
turn on the MPR line type in the point_stat configuration. Be advised
that doing so may result in a large amount of
output data, depending on your situation.
Once you complete the step of running point_stat and generating the
verification statistics and matched pair data, there
are several different ways to proceed with analyzing this information
using the stat_analysis tool:
1. Aggregate the verification statistics:
stat_analysis -job aggregate_stat -line_type SL1L2 -out_line_type CNT
-fcst_var TMP -fcst_lead 03 ...
stat_analysis -job aggregate_stat -line_type MPR -out_line_type CNT
-fcst_var TMP -fcst_lead 03 ...
2. Print out matched pair data to perform your own calculations:
stat_analysis -job filter -dump_row out.stat -fcst_lead 03 -fcst_var
TMP ... time ranges
MET calculates normal and bootstrap confidence intervals for
verification statistics, whenever possible. This
information is also in the MET output files.
Please let us know if you have any further questions.
Thanks,
Paul
RAL HelpDesk {for Ferdinand Rennen} wrote:
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41159 >
>
> Dear Paul,
>
> thank you very much for your fast answer. It helped a lot. I've
installed
> WPP. Now I can tranfsform the output of WRF-ARW into GRIB format and
compare
> it with the measured data.
>
> I have now a new problem (please, let me know is you consider that
I'm
> abusing of your help). My objetive is to estimate the confidence
intervals
> of usual forecasted parameters (temperature, wind speed, etc.) at
the
> measured places, depending on the lapsed time of the forecast (let's
say 3,
> 6, 12, 24, 48 and 72 hours). I have access to the WRF-ARW prognoses
and the
> measured data (hourly) of all the year 2009. The questions are:
>
> 1. In order to get the C.I. of each (3, 6, 12, 24, 48, and 72 hours)
> forecasted parameter, should I compare the measurements with each
one of the
> sampled times? If I'm right, should I compare all the year 6 times,
one for
> each forecast periode, and I'd like to ask you, if you know a faster
method.
>
> 2. Taking advantage of the big amount of data, I think it could be
> interesting to try to find some correlations between the seasonality
or the
> difficulty of the forecast and the quality of the forecast. For sure
I'm not
> the first in thinking about it. I'd appreciate your advice about my
trying.
> Perhaps your have a tip or some references to reads.
>
> Thank you very much and best regards,
>
> Ferdinand
>
>
> 2010/9/28 RAL HelpDesk {for Paul Oldenburg} <met_help at ucar.edu>
>
>> Ferdinand,
>>
>> We assure you that your questions are not dumb, and we are happy to
help
>> you. You are correct that the MET utility
>> point_stat is the most appropriate tool for verifying WRF-ARW
output using
>> weather station data. To use the output of
>> WRF-ARW in MET, you must first use the WRF post-processor (WPP) to
generate
>> GRIB model data. There are instructions for
>> how to do this in chapter 7 of the WRF-NMM User's Guide:
>>
>>
>> http://www.dtcenter.org/wrf-
nmm/users/docs/user_guide/V3/users_guide_nmm_chap1-7.pdf
>>
>> Please note that the information presented in this guide will also
work for
>> the ARW model. The pcp_combine tool is only
>> used if you need to change the precipitation accumulation
intervals. This
>> step should be performed, if necessary, after
>> you run WPP. If you have any other questions, please let us know.
>>
>> Thanks,
>>
>> Paul
>>
>>
>> RAL HelpDesk {for Ferdinand Rennen} wrote:
>>> Tue Sep 28 10:20:54 2010: Request 41159 was acted upon.
>>> Transaction: Ticket created by ferdinand.rennen at gmail.com
>>> Queue: met_help
>>> Subject: WRF + MET
>>> Owner: Nobody
>>> Requestors: ferdinand.rennen at gmail.com
>>> Status: new
>>> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41159 >
>>>
>>>
>>> Dear Sirs,
>>>
>>> I'm writing my Bachelor's Thesis about the computation of the
confidence
>>> intervals at the Technical University of Vienna, Austria
(www.*tuwien*.
>> ac.at).
>>> I'm quite newby with MET but I've some experience with WRF-ARW.
Maybe my
>>> question is quite dumm but I'd appreciate your help to clarify
myself, in
>>> order to get a more precise picture.
>>>
>>> I'm trying to compare the output of WRF-ARW of some region of
central
>> Europe
>>> with some measured data from several weatherstations of the same
region.
>> My
>>> first task is the transformation of the values of the weathter
stations
>>> (with ASCII2NC) into NetCDF format (the values are provided in
ASCII
>>> format). Because of my measured data are not gridded, I suppose
that the
>>> best (or unique) possibility is the use of 'Point Stat' tool (am I
>> right?).
>>> The question is: how should I use the output of WRF-ARW? I mean:
from the
>>> description of the 'point_stat' tool (
>>>
>>
http://www.dtcenter.org/met/users/support/OnLinePractical/OnLinePractical_810/point_stat/index.php
>> )
>>> the forecasted data are from PCP-Combine, but I don't understand
why must
>>> one use PCP-Combine. Is it not possible to use directly the output
from
>>> WRF-ARW? Maybe my reasoning is wrong somewhere. Please, could you
help me
>> to
>>> solve this question.
>>>
>>> Thank you very much in advance and best regards from Vienna,
>>>
>>> Ferdinand
>>
>>
------------------------------------------------
More information about the Met_help
mailing list