[Met_help] [rt.rap.ucar.edu #41590] History for Wind Speed in Point Stat
RAL HelpDesk {for John Halley Gotway}
met_help at ucar.edu
Tue Oct 19 15:25:11 MDT 2010
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
I want to get stats on wind speed forecasts for several speed thresholds not just for the wind components. I generated the point stat files for various wind thresholds. I want to get contingency table data for these wind speed thresholds, not the components. What should I specify in the stat analysis script to do this?
Thanks
Bob Craig
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
From: John Halley Gotway
Time: Mon Oct 18 14:35:29 2010
Bob,
You are actually able to verify wind speed directly when you run
Point-Stat (and Grid-Stat). For example, in the configuration file,
selecting "fcst_field[] = [ "WIND/Z2" ];" would tell Point-Stat
that you want to verify 2-meter wind speed. Point-Stat will do the
following internally:
(1) Check the input GRIB file for 2-meter wind speed. If it finds
a record, it will use it.
(2) If it doesn't find a record, it finds records for the
corresponding 2-meter U and V components.
(3) If the U and V components are defined relative to the grid,
it'll rotate them to be earth-relative.
(4) Lastly, it'll derive the wind speed field from U and V, and use
it in the verification.
This derivation is done for wind speed, but not for wind direction.
The resulting wind speed field is a scalar field just like many
others. You can specify thresholds of interest and verify in the
same sort of way that you can verify temperature. So Point-Stat can
dump out wind speed matched pairs (MPR line), scalar partial sums and
continuous statistics (SL1L2 and CNT lines), contingency
tables and statistics (CTC and CTS lines), and multi-category
contingency tables (MCTC and MCTS lines) - new in METv3.0. And you
can use STAT-Analysis to aggregate together multiple MPR lines, SL1L2
lines, CTC lines, or MCTC lines and derive the corresponding
statistics from them.
If you're verifying wind speed against point observations (in Point-
Stat), you do need to be sure that there are observations of wind
speed in your NetCDF observation file. When you run PB2NC, you
can request that wind speed be derived by including it in the
"obs_grid_code" setting.
It sounds like right now you have output for the individual components
of wind. There isn't really any way of "combining" them into
statistics using the STAT-Analysis tool. The only thing you can
really do is run STAT-Analysis on the VL1L2 (vector partial sums
lines) to derive errors in the wind direction, not speed. Please take
a look at the MET online tutorial for an example of this type of
job:
http://www.dtcenter.org/met/users/support/online_tutorial/METv2.0/stat_analysis/run.php
(Look at the 3rd job listed on that page)
Mathematically, we could enhance STAT-Analysis to derive errors in the
wind speed in the same sort of way that we do it for the wind
direction. But since you can verify wind speed directly as a
scalar field, no user has requested that that sort of functionality be
added.
Hope that helps clarify.
Thanks,
John
On 10/18/2010 10:03 AM, RAL HelpDesk {for Craig, Robert J Civ USAF
AFWA 16 WS/WXN} wrote:
>
> Mon Oct 18 10:03:31 2010: Request 41590 was acted upon.
> Transaction: Ticket created by craigr at offutt.af.mil
> Queue: met_help
> Subject: Wind Speed in Point Stat
> Owner: Nobody
> Requestors: craigr at offutt.af.mil
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>
>
> I want to get stats on wind speed forecasts for several speed
thresholds not just for the wind components. I generated the point
stat files for various wind thresholds. I want to get contingency
table data for these wind speed thresholds, not the components. What
should I specify in the stat analysis script to do this?
>
> Thanks
>
> Bob Craig
------------------------------------------------
Subject: Wind Speed in Point Stat
From: Craig, Robert J Civ USAF AFWA 16 WS/WXN
Time: Tue Oct 19 14:08:24 2010
Thanks Jon, I tried adding WIND/Z10 then WIND/Z2 to the fcst fields.
I have
threshold values for all three fcst fields. But I am get the error
message
below and no output with this configuration. Again what have I done
wrong?
Forecast File:
/gds/data/global/WXQC/data/wrf/t2/qc_101013_us057g1011t02b060000000
Climatology File: none
Configuration File:
/sphome/qcteam/METv2.0/data/config/point_stat_config_WRF
Observation File: /h/data/global/WXQC/data/obs_prep/2010101306
----------------------------------------
Reading records for WIND/Z2.
For WIND/Z2 found 1 forecast levels and 0 climatology levels.
----------------------------------------
Reading records for UGRD/Z10.
For UGRD/Z10 found 1 forecast levels and 0 climatology levels.
----------------------------------------
Reading records for VGRD/Z10.
For VGRD/Z10 found 1 forecast levels and 0 climatology levels.
----------------------------------------
Searching 228250 observations from 63685 PrepBufr messages.
----------------------------------------
Processing WIND/Z2 versus WIND/Z2, for observation type ANYSFC, over
region
FULL, for interpolation method MAX(4), using 0 pairs.
----------------------------------------
Processing UGRD/Z10 versus UGRD/Z10, for observation type ANYSFC, over
region
FULL, for interpolation method MAX(4), using 2528 pairs.
Computing Categorical Statistics.
ERROR: SingleThresh::check() -> unexpected threshold indicator value
of 0.
The 0 pairs for WIND/Z2 is because the obs files don't have the wind
parameter
yet (I started it yesterday). I would rather verify WIND/Z10 instead
of
WIND/Z2. Are they both available?
Thanks
-----Original Message-----
From: RAL HelpDesk {for John Halley Gotway} [mailto:met_help at ucar.edu]
Sent: Monday, October 18, 2010 3:36 PM
To: Craig, Robert J Civ USAF AFWA 16 WS/WXN
Cc: Pelletier, Kristopher Civ USAF AFWA 16 WS/WXN
Subject: Re: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
Bob,
You are actually able to verify wind speed directly when you run
Point-Stat
(and Grid-Stat). For example, in the configuration file, selecting
"fcst_field[] = [ "WIND/Z2" ];" would tell Point-Stat
that you want to verify 2-meter wind speed. Point-Stat will do the
following
internally:
(1) Check the input GRIB file for 2-meter wind speed. If it finds
a
record, it will use it.
(2) If it doesn't find a record, it finds records for the
corresponding
2-meter U and V components.
(3) If the U and V components are defined relative to the grid,
it'll
rotate them to be earth-relative.
(4) Lastly, it'll derive the wind speed field from U and V, and use
it in
the verification.
This derivation is done for wind speed, but not for wind direction.
The
resulting wind speed field is a scalar field just like many others.
You can
specify thresholds of interest and verify in the
same sort of way that you can verify temperature. So Point-Stat can
dump out
wind speed matched pairs (MPR line), scalar partial sums and
continuous
statistics (SL1L2 and CNT lines), contingency
tables and statistics (CTC and CTS lines), and multi-category
contingency
tables (MCTC and MCTS lines) - new in METv3.0. And you can use STAT-
Analysis
to aggregate together multiple MPR lines, SL1L2
lines, CTC lines, or MCTC lines and derive the corresponding
statistics from
them.
If you're verifying wind speed against point observations (in Point-
Stat), you
do need to be sure that there are observations of wind speed in your
NetCDF
observation file. When you run PB2NC, you
can request that wind speed be derived by including it in the
"obs_grid_code"
setting.
It sounds like right now you have output for the individual components
of
wind. There isn't really any way of "combining" them into statistics
using
the STAT-Analysis tool. The only thing you can
really do is run STAT-Analysis on the VL1L2 (vector partial sums
lines) to
derive errors in the wind direction, not speed. Please take a look at
the MET
online tutorial for an example of this type of
job:
http://www.dtcenter.org/met/users/support/online_tutorial/METv2.0/stat_analysis/run.php
(Look at the 3rd job listed on that page)
Mathematically, we could enhance STAT-Analysis to derive errors in the
wind
speed in the same sort of way that we do it for the wind direction.
But since
you can verify wind speed directly as a
scalar field, no user has requested that that sort of functionality be
added.
Hope that helps clarify.
Thanks,
John
On 10/18/2010 10:03 AM, RAL HelpDesk {for Craig, Robert J Civ USAF
AFWA 16
WS/WXN} wrote:
>
> Mon Oct 18 10:03:31 2010: Request 41590 was acted upon.
> Transaction: Ticket created by craigr at offutt.af.mil
> Queue: met_help
> Subject: Wind Speed in Point Stat
> Owner: Nobody
> Requestors: craigr at offutt.af.mil
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>
>
> I want to get stats on wind speed forecasts for several speed
thresholds not
> just for the wind components. I generated the point stat files for
various
> wind thresholds. I want to get contingency table data for these
wind speed
> thresholds, not the components. What should I specify in the stat
analysis
> script to do this?
>
> Thanks
>
> Bob Craig
------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
From: John Halley Gotway
Time: Tue Oct 19 14:23:36 2010
Bob,
In the config file you sent me, I see the following setting:
fcst_thresh[] = ["NA gt7.7 gt10.3 gt12.9 gt18" , "NA gt7.7 gt10.3
gt12.9 gt18" , "NA gt7.7 gt10.3 gt12.9 gt18"];
Why do you have NA listed as a threshold? I suspect that's what's
causing the problem. Please try remove NA, and let me know if that
does not fix it.
If you're using PREPBUFR files, I suspect that you'll probably ONLY
have 10-meter winds, not 2-meter winds. Just try running PB2NC on a
sample PREPBUFR file and be sure that your configuration file
includes "WIND" in the "obs_grib_code[]" parameter list.
John
On 10/19/2010 02:08 PM, RAL HelpDesk {for Craig, Robert J Civ USAF
AFWA 16 WS/WXN} wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>
> Thanks Jon, I tried adding WIND/Z10 then WIND/Z2 to the fcst fields.
I have
> threshold values for all three fcst fields. But I am get the error
message
> below and no output with this configuration. Again what have I done
wrong?
>
> Forecast File:
> /gds/data/global/WXQC/data/wrf/t2/qc_101013_us057g1011t02b060000000
> Climatology File: none
> Configuration File:
/sphome/qcteam/METv2.0/data/config/point_stat_config_WRF
> Observation File: /h/data/global/WXQC/data/obs_prep/2010101306
>
> ----------------------------------------
>
> Reading records for WIND/Z2.
> For WIND/Z2 found 1 forecast levels and 0 climatology levels.
>
> ----------------------------------------
>
> Reading records for UGRD/Z10.
> For UGRD/Z10 found 1 forecast levels and 0 climatology levels.
>
> ----------------------------------------
>
> Reading records for VGRD/Z10.
> For VGRD/Z10 found 1 forecast levels and 0 climatology levels.
>
> ----------------------------------------
>
> Searching 228250 observations from 63685 PrepBufr messages.
>
> ----------------------------------------
>
> Processing WIND/Z2 versus WIND/Z2, for observation type ANYSFC, over
region
> FULL, for interpolation method MAX(4), using 0 pairs.
>
> ----------------------------------------
>
> Processing UGRD/Z10 versus UGRD/Z10, for observation type ANYSFC,
over region
> FULL, for interpolation method MAX(4), using 2528 pairs.
> Computing Categorical Statistics.
>
>
> ERROR: SingleThresh::check() -> unexpected threshold indicator value
of 0.
>
>
>
> The 0 pairs for WIND/Z2 is because the obs files don't have the wind
parameter
> yet (I started it yesterday). I would rather verify WIND/Z10
instead of
> WIND/Z2. Are they both available?
>
> Thanks
>
> -----Original Message-----
> From: RAL HelpDesk {for John Halley Gotway}
[mailto:met_help at ucar.edu]
> Sent: Monday, October 18, 2010 3:36 PM
> To: Craig, Robert J Civ USAF AFWA 16 WS/WXN
> Cc: Pelletier, Kristopher Civ USAF AFWA 16 WS/WXN
> Subject: Re: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
>
> Bob,
>
> You are actually able to verify wind speed directly when you run
Point-Stat
> (and Grid-Stat). For example, in the configuration file, selecting
> "fcst_field[] = [ "WIND/Z2" ];" would tell Point-Stat
> that you want to verify 2-meter wind speed. Point-Stat will do the
following
> internally:
> (1) Check the input GRIB file for 2-meter wind speed. If it
finds a
> record, it will use it.
> (2) If it doesn't find a record, it finds records for the
corresponding
> 2-meter U and V components.
> (3) If the U and V components are defined relative to the grid,
it'll
> rotate them to be earth-relative.
> (4) Lastly, it'll derive the wind speed field from U and V, and
use it in
> the verification.
>
> This derivation is done for wind speed, but not for wind direction.
The
> resulting wind speed field is a scalar field just like many others.
You can
> specify thresholds of interest and verify in the
> same sort of way that you can verify temperature. So Point-Stat can
dump out
> wind speed matched pairs (MPR line), scalar partial sums and
continuous
> statistics (SL1L2 and CNT lines), contingency
> tables and statistics (CTC and CTS lines), and multi-category
contingency
> tables (MCTC and MCTS lines) - new in METv3.0. And you can use
STAT-Analysis
> to aggregate together multiple MPR lines, SL1L2
> lines, CTC lines, or MCTC lines and derive the corresponding
statistics from
> them.
>
> If you're verifying wind speed against point observations (in Point-
Stat), you
> do need to be sure that there are observations of wind speed in your
NetCDF
> observation file. When you run PB2NC, you
> can request that wind speed be derived by including it in the
"obs_grid_code"
> setting.
>
> It sounds like right now you have output for the individual
components of
> wind. There isn't really any way of "combining" them into
statistics using
> the STAT-Analysis tool. The only thing you can
> really do is run STAT-Analysis on the VL1L2 (vector partial sums
lines) to
> derive errors in the wind direction, not speed. Please take a look
at the MET
> online tutorial for an example of this type of
> job:
>
http://www.dtcenter.org/met/users/support/online_tutorial/METv2.0/stat_analysis/run.php
> (Look at the 3rd job listed on that page)
>
> Mathematically, we could enhance STAT-Analysis to derive errors in
the wind
> speed in the same sort of way that we do it for the wind direction.
But since
> you can verify wind speed directly as a
> scalar field, no user has requested that that sort of functionality
be added.
>
> Hope that helps clarify.
>
> Thanks,
> John
>
>
> On 10/18/2010 10:03 AM, RAL HelpDesk {for Craig, Robert J Civ USAF
AFWA 16
> WS/WXN} wrote:
>>
>> Mon Oct 18 10:03:31 2010: Request 41590 was acted upon.
>> Transaction: Ticket created by craigr at offutt.af.mil
>> Queue: met_help
>> Subject: Wind Speed in Point Stat
>> Owner: Nobody
>> Requestors: craigr at offutt.af.mil
>> Status: new
>> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>>
>>
>> I want to get stats on wind speed forecasts for several speed
thresholds not
>> just for the wind components. I generated the point stat files
for various
>> wind thresholds. I want to get contingency table data for these
wind speed
>> thresholds, not the components. What should I specify in the stat
analysis
>> script to do this?
>>
>> Thanks
>>
>> Bob Craig
>
>
------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
From: Craig, Robert J Civ USAF AFWA 16 WS/WXN
Time: Tue Oct 19 14:25:49 2010
Jon, the NA was for the-non threshold case. What should I use for
this case?
Thanks
-----Original Message-----
From: RAL HelpDesk
{for John Halley Gotway} [mailto:met_help at ucar.edu]
Sent: Tuesday,
October 19, 2010 3:24 PM
To: Craig, Robert J Civ USAF AFWA 16 WS/WXN
Cc: Pelletier, Kristopher Civ USAF AFWA 16 WS/WXN
Subject: Re:
[rt.rap.ucar.edu #41590] Wind Speed in Point Stat
Bob,
In the
config file you sent me, I see the following setting:
fcst_thresh[] = ["NA gt7.7 gt10.3 gt12.9 gt18" , "NA gt7.7 gt10.3
gt12.9 gt18" , "NA gt7.7 gt10.3 gt12.9 gt18"];
Why do you have NA
listed as a threshold? I suspect that's what's causing the problem.
Please try remove NA, and let me know if that does not fix it.
If
you're using PREPBUFR files, I suspect that you'll probably ONLY have
10-meter winds, not 2-meter winds. Just try running PB2NC on a sample
PREPBUFR file and be sure that your configuration file
includes
"WIND" in the "obs_grib_code[]" parameter list.
John
On
10/19/2010 02:08 PM, RAL HelpDesk {for Craig, Robert J Civ USAF AFWA
16 WS/WXN} wrote:
>
> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>
>
Thanks Jon, I tried adding WIND/Z10 then WIND/Z2 to the fcst fields.
I have
> threshold values for all three fcst fields. But I am get
the error message
> below and no output with this configuration.
Again what have I done wrong?
>
> Forecast File:
>
/gds/data/global/WXQC/data/wrf/t2/qc_101013_us057g1011t02b060000000
>
Climatology File: none
> Configuration File:
/sphome/qcteam/METv2.0/data/config/point_stat_config_WRF
>
Observation File: /h/data/global/WXQC/data/obs_prep/2010101306
>
>
----------------------------------------
>
> Reading records for
WIND/Z2.
> For WIND/Z2 found 1 forecast levels and 0 climatology
levels.
>
> ----------------------------------------
>
> Reading
records for UGRD/Z10.
> For UGRD/Z10 found 1 forecast levels and 0
climatology levels.
>
> ----------------------------------------
>
> Reading records for VGRD/Z10.
> For VGRD/Z10 found 1 forecast
levels and 0 climatology levels.
>
>
----------------------------------------
>
> Searching 228250
observations from 63685 PrepBufr messages.
>
>
----------------------------------------
>
> Processing WIND/Z2
versus WIND/Z2, for observation type ANYSFC, over region
> FULL, for
interpolation method MAX(4), using 0 pairs.
>
>
----------------------------------------
>
> Processing UGRD/Z10
versus UGRD/Z10, for observation type ANYSFC, over region
> FULL,
for interpolation method MAX(4), using 2528 pairs.
> Computing
Categorical Statistics.
>
>
> ERROR: SingleThresh::check() ->
unexpected threshold indicator value of 0.
>
>
>
> The 0 pairs
for WIND/Z2 is because the obs files don't have the wind parameter
>
yet (I started it yesterday). I would rather verify WIND/Z10 instead
of
> WIND/Z2. Are they both available?
>
> Thanks
>
>
-----Original Message-----
> From: RAL HelpDesk {for John Halley
Gotway} [mailto:met_help at ucar.edu]
> Sent: Monday, October 18, 2010
3:36 PM
> To: Craig, Robert J Civ USAF AFWA 16 WS/WXN
> Cc:
Pelletier, Kristopher Civ USAF AFWA 16 WS/WXN
> Subject: Re:
[rt.rap.ucar.edu #41590] Wind Speed in Point Stat
>
> Bob,
>
>
You are actually able to verify wind speed directly when you run
Point-Stat
> (and Grid-Stat). For example, in the configuration
file, selecting
> "fcst_field[] = [ "WIND/Z2" ];" would tell Point-
Stat
> that you want to verify 2-meter wind speed. Point-Stat will
do the following
> internally:
> (1) Check the input GRIB file
for 2-meter wind speed. If it finds a
> record, it will use it.
>
(2) If it doesn't find a record, it finds records for the
corresponding
> 2-meter U and V components.
> (3) If the U and V
components are defined relative to the grid, it'll
> rotate them to
be earth-relative.
> (4) Lastly, it'll derive the wind speed field
from U and V, and use it in
> the verification.
>
> This
derivation is done for wind speed, but not for wind direction. The
> resulting wind speed field is a scalar field just like many others.
You can
> specify thresholds of interest and verify in the
> same
sort of way that you can verify temperature. So Point-Stat can dump
out
> wind speed matched pairs (MPR line), scalar partial sums and
continuous
> statistics (SL1L2 and CNT lines), contingency
> tables
and statistics (CTC and CTS lines), and multi-category contingency
>
tables (MCTC and MCTS lines) - new in METv3.0. And you can use STAT-
Analysis
> to aggregate together multiple MPR lines, SL1L2
> lines,
CTC lines, or MCTC lines and derive the corresponding statistics from
> them.
>
> If you're verifying wind speed against point
observations (in Point-Stat), you
> do need to be sure that there
are observations of wind speed in your NetCDF
> observation file.
When you run PB2NC, you
> can request that wind speed be derived by
including it in the "obs_grid_code"
> setting.
>
> It sounds like
right now you have output for the individual components of
> wind.
There isn't really any way of "combining" them into statistics using
> the STAT-Analysis tool. The only thing you can
> really do is run
STAT-Analysis on the VL1L2 (vector partial sums lines) to
> derive
errors in the wind direction, not speed. Please take a look at the
MET
> online tutorial for an example of this type of
> job:
>
http://www.dtcenter.org/met/users/support/online_tutorial/METv2.0/stat_analysis/run.php
> (Look at the 3rd job listed on that page)
>
> Mathematically, we
could enhance STAT-Analysis to derive errors in the wind
> speed in
the same sort of way that we do it for the wind direction. But since
> you can verify wind speed directly as a
> scalar field, no user has
requested that that sort of functionality be added.
>
> Hope that
helps clarify.
>
> Thanks,
> John
>
>
> On 10/18/2010 10:03
AM, RAL HelpDesk {for Craig, Robert J Civ USAF AFWA 16
> WS/WXN}
wrote:
>>
>> Mon Oct 18 10:03:31 2010: Request 41590 was acted upon.
>> Transaction: Ticket created by craigr at offutt.af.mil
>>
Queue: met_help
>> Subject: Wind Speed in Point Stat
>>
Owner: Nobody
>> Requestors: craigr at offutt.af.mil
>> Status:
new
>> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>>
>>
>>
I want to get stats on wind speed forecasts for several speed
thresholds not
>> just for the wind components. I generated the
point stat files for various
>> wind thresholds. I want to get
contingency table data for these wind speed
>> thresholds, not the
components. What should I specify in the stat analysis
>> script to
do this?
>>
>> Thanks
>>
>> Bob Craig
>
>
------------------------------------------------
Subject: Re: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
From: John Halley Gotway
Time: Tue Oct 19 15:14:00 2010
Bob,
You don't need to specify anything in "fcst_thresh" for the non-
thresholded case - i.e. continuous statistics.
Point-Stat uses the following logic:
(1) It checks the output_flag to see if you've requested continuous
statistics. If you have it will compute them (without even looking at
the thresholds).
(2) It checks the output_flag to see if you've requested contingency
table counts and/or statistics. If you have, it will compute them
once for each of the thresholds you're provided in the config file.
Make sense?
Are you perhaps talking about thresholding the wind speeds? If so,
then that is where the "NA" threshold belongs. For example:
fcst_wind_thresh[] = [ "NA", "ge2.0", "ge3.0" ];
obs_wind_thresh[] = [];
Point-Stat only uses these thresholds in computing the vector partial
sums lines. So it checks the output flag to see if you've requested
the VL1L2 partial sums. If you have, it will compute the
VL1L2 data once for each of the "fcst_wind_thresh" thresholds you've
provided.
To summarize, Point-Stat uses...
- The fcst_thresh and obs_thresh settings to define contingency
tables.
- The fcst_wind_thresh and obs_wind_thresh settings to define VL1L2
vector partial sums.
Perhaps what you're really trying to do is a *conditional*
verification of wind speed. Meaning, how do the forecasts a strong
wind speed compare to the forecasts of light wind speed. If that is
the
case, you'd need to decide how to define the conditions. Do you want
to condition on the forecast wind speed value or the observed wind
speed value?
Probably the most flexible way of doing this would be to run Point-
Stat and dump out the raw matched pair values for wind speed (MPR).
Then you can run STAT-Analysis, using the job filtering options,
to define perform all sorts of conditioning. You can use the "-
col_min" and "-col_max" options to condition on the forecast and/or
observation values.
John
On 10/19/2010 02:25 PM, RAL HelpDesk {for Craig, Robert J Civ USAF
AFWA 16 WS/WXN} wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>
> Jon, the NA was for the-non threshold case. What should I use for
this case?
>
> Thanks
>
> -----Original Message-----
> From: RAL HelpDesk {for John Halley Gotway}
[mailto:met_help at ucar.edu]
> Sent: Tuesday, October 19, 2010 3:24 PM
> To: Craig, Robert J Civ USAF AFWA 16 WS/WXN
> Cc: Pelletier, Kristopher Civ USAF AFWA 16 WS/WXN
> Subject: Re: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
>
> Bob,
>
> In the config file you sent me, I see the following setting:
> fcst_thresh[] = ["NA gt7.7 gt10.3 gt12.9 gt18" , "NA gt7.7 gt10.3
gt12.9 gt18" , "NA gt7.7 gt10.3 gt12.9 gt18"];
>
> Why do you have NA listed as a threshold? I suspect that's what's
causing the problem. Please try remove NA, and let me know if that
does not fix it.
>
> If you're using PREPBUFR files, I suspect that you'll probably ONLY
have 10-meter winds, not 2-meter winds. Just try running PB2NC on a
sample PREPBUFR file and be sure that your configuration file
> includes "WIND" in the "obs_grib_code[]" parameter list.
>
> John
>
> On 10/19/2010 02:08 PM, RAL HelpDesk {for Craig, Robert J Civ USAF
AFWA 16 WS/WXN} wrote:
>>
>> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>>
>> Thanks Jon, I tried adding WIND/Z10 then WIND/Z2 to the fcst
fields. I have
>> threshold values for all three fcst fields. But I am get the error
message
>> below and no output with this configuration. Again what have I
done wrong?
>>
>> Forecast File:
>> /gds/data/global/WXQC/data/wrf/t2/qc_101013_us057g1011t02b060000000
>> Climatology File: none
>> Configuration File:
/sphome/qcteam/METv2.0/data/config/point_stat_config_WRF
>> Observation File: /h/data/global/WXQC/data/obs_prep/2010101306
>>
>> ----------------------------------------
>>
>> Reading records for WIND/Z2.
>> For WIND/Z2 found 1 forecast levels and 0 climatology levels.
>>
>> ----------------------------------------
>>
>> Reading records for UGRD/Z10.
>> For UGRD/Z10 found 1 forecast levels and 0 climatology levels.
>>
>> ----------------------------------------
>>
>> Reading records for VGRD/Z10.
>> For VGRD/Z10 found 1 forecast levels and 0 climatology levels.
>>
>> ----------------------------------------
>>
>> Searching 228250 observations from 63685 PrepBufr messages.
>>
>> ----------------------------------------
>>
>> Processing WIND/Z2 versus WIND/Z2, for observation type ANYSFC,
over region
>> FULL, for interpolation method MAX(4), using 0 pairs.
>>
>> ----------------------------------------
>>
>> Processing UGRD/Z10 versus UGRD/Z10, for observation type ANYSFC,
over region
>> FULL, for interpolation method MAX(4), using 2528 pairs.
>> Computing Categorical Statistics.
>>
>>
>> ERROR: SingleThresh::check() -> unexpected threshold indicator
value of 0.
>>
>>
>>
>> The 0 pairs for WIND/Z2 is because the obs files don't have the
wind parameter
>> yet (I started it yesterday). I would rather verify WIND/Z10
instead of
>> WIND/Z2. Are they both available?
>>
>> Thanks
>>
>> -----Original Message-----
>> From: RAL HelpDesk {for John Halley Gotway}
[mailto:met_help at ucar.edu]
>> Sent: Monday, October 18, 2010 3:36 PM
>> To: Craig, Robert J Civ USAF AFWA 16 WS/WXN
>> Cc: Pelletier, Kristopher Civ USAF AFWA 16 WS/WXN
>> Subject: Re: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
>>
>> Bob,
>>
>> You are actually able to verify wind speed directly when you run
Point-Stat
>> (and Grid-Stat). For example, in the configuration file, selecting
>> "fcst_field[] = [ "WIND/Z2" ];" would tell Point-Stat
>> that you want to verify 2-meter wind speed. Point-Stat will do the
following
>> internally:
>> (1) Check the input GRIB file for 2-meter wind speed. If it
finds a
>> record, it will use it.
>> (2) If it doesn't find a record, it finds records for the
corresponding
>> 2-meter U and V components.
>> (3) If the U and V components are defined relative to the grid,
it'll
>> rotate them to be earth-relative.
>> (4) Lastly, it'll derive the wind speed field from U and V, and
use it in
>> the verification.
>>
>> This derivation is done for wind speed, but not for wind direction.
The
>> resulting wind speed field is a scalar field just like many others.
You can
>> specify thresholds of interest and verify in the
>> same sort of way that you can verify temperature. So Point-Stat
can dump out
>> wind speed matched pairs (MPR line), scalar partial sums and
continuous
>> statistics (SL1L2 and CNT lines), contingency
>> tables and statistics (CTC and CTS lines), and multi-category
contingency
>> tables (MCTC and MCTS lines) - new in METv3.0. And you can use
STAT-Analysis
>> to aggregate together multiple MPR lines, SL1L2
>> lines, CTC lines, or MCTC lines and derive the corresponding
statistics from
>> them.
>>
>> If you're verifying wind speed against point observations (in
Point-Stat), you
>> do need to be sure that there are observations of wind speed in
your NetCDF
>> observation file. When you run PB2NC, you
>> can request that wind speed be derived by including it in the
"obs_grid_code"
>> setting.
>>
>> It sounds like right now you have output for the individual
components of
>> wind. There isn't really any way of "combining" them into
statistics using
>> the STAT-Analysis tool. The only thing you can
>> really do is run STAT-Analysis on the VL1L2 (vector partial sums
lines) to
>> derive errors in the wind direction, not speed. Please take a look
at the MET
>> online tutorial for an example of this type of
>> job:
>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv2.0/stat_analysis/run.php
>> (Look at the 3rd job listed on that page)
>>
>> Mathematically, we could enhance STAT-Analysis to derive errors in
the wind
>> speed in the same sort of way that we do it for the wind direction.
But since
>> you can verify wind speed directly as a
>> scalar field, no user has requested that that sort of functionality
be added.
>>
>> Hope that helps clarify.
>>
>> Thanks,
>> John
>>
>>
>> On 10/18/2010 10:03 AM, RAL HelpDesk {for Craig, Robert J Civ USAF
AFWA 16
>> WS/WXN} wrote:
>>>
>>> Mon Oct 18 10:03:31 2010: Request 41590 was acted upon.
>>> Transaction: Ticket created by craigr at offutt.af.mil
>>> Queue: met_help
>>> Subject: Wind Speed in Point Stat
>>> Owner: Nobody
>>> Requestors: craigr at offutt.af.mil
>>> Status: new
>>> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>>>
>>>
>>> I want to get stats on wind speed forecasts for several speed
thresholds not
>>> just for the wind components. I generated the point stat files
for various
>>> wind thresholds. I want to get contingency table data for these
wind speed
>>> thresholds, not the components. What should I specify in the stat
analysis
>>> script to do this?
>>>
>>> Thanks
>>>
>>> Bob Craig
>>
>>
>
------------------------------------------------
Subject: RE: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
From: Craig, Robert J Civ USAF AFWA 16 WS/WXN
Time: Tue Oct 19 15:16:25 2010
Thanks Jon, that answers my questions.
Bob
-----Original
Message-----
From: RAL HelpDesk {for John Halley Gotway}
[mailto:met_help at ucar.edu]
Sent: Tuesday, October 19, 2010 4:14 PM
To: Craig, Robert J Civ USAF AFWA 16 WS/WXN
Cc: Pelletier, Kristopher
Civ USAF AFWA 16 WS/WXN
Subject: Re: [rt.rap.ucar.edu #41590] Wind
Speed in Point Stat
Bob,
You don't need to specify anything in
"fcst_thresh" for the non-thresholded case - i.e. continuous
statistics.
Point-Stat uses the following logic:
(1) It checks the
output_flag to see if you've requested continuous statistics. If you
have it will compute them (without even looking at the thresholds).
(2) It checks the output_flag to see if you've requested contingency
table counts and/or statistics. If you have, it will compute them
once for each of the thresholds you're provided in the config file.
Make sense?
Are you perhaps talking about thresholding the wind
speeds? If so, then that is where the "NA" threshold belongs. For
example:
fcst_wind_thresh[] = [ "NA", "ge2.0", "ge3.0" ];
obs_wind_thresh[] = [];
Point-Stat only uses these thresholds in
computing the vector partial sums lines. So it checks the output flag
to see if you've requested the VL1L2 partial sums. If you have, it
will compute the
VL1L2 data once for each of the "fcst_wind_thresh"
thresholds you've provided.
To summarize, Point-Stat uses...
- The
fcst_thresh and obs_thresh settings to define contingency tables.
-
The fcst_wind_thresh and obs_wind_thresh settings to define VL1L2
vector partial sums.
Perhaps what you're really trying to do is a
*conditional* verification of wind speed. Meaning, how do the
forecasts a strong wind speed compare to the forecasts of light wind
speed. If that is the
case, you'd need to decide how to define the
conditions. Do you want to condition on the forecast wind speed value
or the observed wind speed value?
Probably the most flexible way of
doing this would be to run Point-Stat and dump out the raw matched
pair values for wind speed (MPR). Then you can run STAT-Analysis,
using the job filtering options,
to define perform all sorts of
conditioning. You can use the "-col_min" and "-col_max" options to
condition on the forecast and/or observation values.
John
On
10/19/2010 02:25 PM, RAL HelpDesk {for Craig, Robert J Civ USAF AFWA
16 WS/WXN} wrote:
>
> <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>
> Jon,
the NA was for the-non threshold case. What should I use for this
case?
>
> Thanks
>
> -----Original Message-----
> From: RAL
HelpDesk {for John Halley Gotway} [mailto:met_help at ucar.edu]
> Sent:
Tuesday, October 19, 2010 3:24 PM
> To: Craig, Robert J Civ USAF AFWA
16 WS/WXN
> Cc: Pelletier, Kristopher Civ USAF AFWA 16 WS/WXN
>
Subject: Re: [rt.rap.ucar.edu #41590] Wind Speed in Point Stat
>
>
Bob,
>
> In the config file you sent me, I see the following
setting:
> fcst_thresh[] = ["NA gt7.7 gt10.3 gt12.9 gt18" , "NA
gt7.7 gt10.3 gt12.9 gt18" , "NA gt7.7 gt10.3 gt12.9 gt18"];
>
> Why
do you have NA listed as a threshold? I suspect that's what's causing
the problem. Please try remove NA, and let me know if that does not
fix it.
>
> If you're using PREPBUFR files, I suspect that you'll
probably ONLY have 10-meter winds, not 2-meter winds. Just try
running PB2NC on a sample PREPBUFR file and be sure that your
configuration file
> includes "WIND" in the "obs_grib_code[]"
parameter list.
>
> John
>
> On 10/19/2010 02:08 PM, RAL
HelpDesk {for Craig, Robert J Civ USAF AFWA 16 WS/WXN} wrote:
>>
>>
<URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>>
>> Thanks Jon, I tried adding WIND/Z10 then WIND/Z2 to the fcst
fields. I have
>> threshold values for all three fcst fields. But I
am get the error message
>> below and no output with this
configuration. Again what have I done wrong?
>>
>> Forecast File:
>> /gds/data/global/WXQC/data/wrf/t2/qc_101013_us057g1011t02b060000000
>> Climatology File: none
>> Configuration File:
/sphome/qcteam/METv2.0/data/config/point_stat_config_WRF
>>
Observation File: /h/data/global/WXQC/data/obs_prep/2010101306
>>
>>
----------------------------------------
>>
>> Reading records for
WIND/Z2.
>> For WIND/Z2 found 1 forecast levels and 0 climatology
levels.
>>
>> ----------------------------------------
>>
>>
Reading records for UGRD/Z10.
>> For UGRD/Z10 found 1 forecast levels
and 0 climatology levels.
>>
>>
----------------------------------------
>>
>> Reading records for
VGRD/Z10.
>> For VGRD/Z10 found 1 forecast levels and 0 climatology
levels.
>>
>> ----------------------------------------
>>
>>
Searching 228250 observations from 63685 PrepBufr messages.
>>
>>
----------------------------------------
>>
>> Processing WIND/Z2
versus WIND/Z2, for observation type ANYSFC, over region
>> FULL,
for interpolation method MAX(4), using 0 pairs.
>>
>>
----------------------------------------
>>
>> Processing UGRD/Z10
versus UGRD/Z10, for observation type ANYSFC, over region
>> FULL,
for interpolation method MAX(4), using 2528 pairs.
>> Computing
Categorical Statistics.
>>
>>
>> ERROR: SingleThresh::check() ->
unexpected threshold indicator value of 0.
>>
>>
>>
>> The 0 pairs
for WIND/Z2 is because the obs files don't have the wind parameter
>> yet (I started it yesterday). I would rather verify WIND/Z10
instead of
>> WIND/Z2. Are they both available?
>>
>> Thanks
>>
>> -----Original Message-----
>> From: RAL HelpDesk {for John Halley
Gotway} [mailto:met_help at ucar.edu]
>> Sent: Monday, October 18, 2010
3:36 PM
>> To: Craig, Robert J Civ USAF AFWA 16 WS/WXN
>> Cc:
Pelletier, Kristopher Civ USAF AFWA 16 WS/WXN
>> Subject: Re:
[rt.rap.ucar.edu #41590] Wind Speed in Point Stat
>>
>> Bob,
>>
>>
You are actually able to verify wind speed directly when you run
Point-Stat
>> (and Grid-Stat). For example, in the configuration
file, selecting
>> "fcst_field[] = [ "WIND/Z2" ];" would tell Point-
Stat
>> that you want to verify 2-meter wind speed. Point-Stat will
do the following
>> internally:
>> (1) Check the input GRIB file
for 2-meter wind speed. If it finds a
>> record, it will use it.
>> (2) If it doesn't find a record, it finds records for the
corresponding
>> 2-meter U and V components.
>> (3) If the U and
V components are defined relative to the grid, it'll
>> rotate them
to be earth-relative.
>> (4) Lastly, it'll derive the wind speed
field from U and V, and use it in
>> the verification.
>>
>> This
derivation is done for wind speed, but not for wind direction. The
>> resulting wind speed field is a scalar field just like many others.
You can
>> specify thresholds of interest and verify in the
>> same
sort of way that you can verify temperature. So Point-Stat can dump
out
>> wind speed matched pairs (MPR line), scalar partial sums and
continuous
>> statistics (SL1L2 and CNT lines), contingency
>>
tables and statistics (CTC and CTS lines), and multi-category
contingency
>> tables (MCTC and MCTS lines) - new in METv3.0. And
you can use STAT-Analysis
>> to aggregate together multiple MPR
lines, SL1L2
>> lines, CTC lines, or MCTC lines and derive the
corresponding statistics from
>> them.
>>
>> If you're verifying
wind speed against point observations (in Point-Stat), you
>> do
need to be sure that there are observations of wind speed in your
NetCDF
>> observation file. When you run PB2NC, you
>> can request
that wind speed be derived by including it in the "obs_grid_code"
>>
setting.
>>
>> It sounds like right now you have output for the
individual components of
>> wind. There isn't really any way of
"combining" them into statistics using
>> the STAT-Analysis tool.
The only thing you can
>> really do is run STAT-Analysis on the VL1L2
(vector partial sums lines) to
>> derive errors in the wind
direction, not speed. Please take a look at the MET
>> online
tutorial for an example of this type of
>> job:
>>
http://www.dtcenter.org/met/users/support/online_tutorial/METv2.0/stat_analysis/run.php
>> (Look at the 3rd job listed on that page)
>>
>> Mathematically,
we could enhance STAT-Analysis to derive errors in the wind
>> speed
in the same sort of way that we do it for the wind direction. But
since
>> you can verify wind speed directly as a
>> scalar field,
no user has requested that that sort of functionality be added.
>>
>> Hope that helps clarify.
>>
>> Thanks,
>> John
>>
>>
>> On
10/18/2010 10:03 AM, RAL HelpDesk {for Craig, Robert J Civ USAF AFWA
16
>> WS/WXN} wrote:
>>>
>>> Mon Oct 18 10:03:31 2010: Request
41590 was acted upon.
>>> Transaction: Ticket created by
craigr at offutt.af.mil
>>> Queue: met_help
>>> Subject:
Wind Speed in Point Stat
>>> Owner: Nobody
>>> Requestors:
craigr at offutt.af.mil
>>> Status: new
>>> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=41590 >
>>>
>>>
>>> I want to get stats on wind speed forecasts for several speed
thresholds not
>>> just for the wind components. I generated the
point stat files for various
>>> wind thresholds. I want to get
contingency table data for these wind speed
>>> thresholds, not the
components. What should I specify in the stat analysis
>>> script
to do this?
>>>
>>> Thanks
>>>
>>> Bob Craig
>>
>>
>
------------------------------------------------
More information about the Met_help
mailing list