[Met_help] [rt.rap.ucar.edu #71051] History for Question about ascii2nc and point_stat
Julie Prestopnik via RT
met_help at ucar.edu
Fri Mar 27 09:59:55 MDT 2015
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
Good morning,
I have model forecasts at one-hour snapshots over an 8-hour period, and I
am looking to compare a number of stations at each of those times. I have
read the MET tutorial and the users' guide and I have a couple remaining
questions.
I am using USCRN station data for my point observations, and I can convert
them into a MET-formatted ascii file to use with ascii2nc and then with
point_stat.
1. Can I have more times than desired in my MET-formatted ascii file and
then use the configuration file or command line options for ascii2nc to
select only a certain time range from these observations to include in the
netCDF file, or will it necessarily use all the observations I have in the
ascii file I give to ascii2nc?
2. If I have point observations at multiple times in one ascii or netcdf
file that I give as input to point_stat, can point_stat compare my hourly
forecasts to this hourly data over the whole time period in one call? Or
do I need to have only the observations I want to compare to a given
forecast (a one-time forecast) and just call point_stat for each forecast
time as a separate command, using a separate input file of observations?
Thank you,
Jess
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: Question about ascii2nc and point_stat
From: Julie Prestopnik
Time: Tue Mar 10 12:27:06 2015
Hi Jessica.
You can put observations for as many times as you would like in your
MET-formatted ascii file. ascii2nc will pass them all through the
output
NetCDF file, but they can be filtered out when using Point-Stat.
The Point-Stat configuration file has an obs_window where you can
specify a
begin and end window. The obs_window is defined relative to the
forecast
time. So you could specify (in seconds), for example, one hour plus
or
minus the forecast time. If your forecast hour was noon, that would
tell
it to grab all observations between 11am and 1pm.
Although you could have a ton of data in your ascii2nc output file,
Point-Stat would need to read it in each time, so it could potentially
be
slow.
It wasn't clear if you have 8 forecast files or if your forecast times
are
all in one file. It would be easiest to run Point-Stat if you had 8
separate forecast files. You would call Point-Stat 8 times, once for
each
forecast file you have. If you have multiple forecasts in one file,
please
let us know and we can walk you through running Point-Stat, modifying
the
configuration file, to account for multiple forecasts in one file.
Please let me know if you have any other questions.
Thanks,
Julie
On Tue, Mar 10, 2015 at 9:38 AM, Kunke, Jessica via RT
<met_help at ucar.edu>
wrote:
>
> Tue Mar 10 09:38:01 2015: Request 71051 was acted upon.
> Transaction: Ticket created by jkunke at fas.harvard.edu
> Queue: met_help
> Subject: Question about ascii2nc and point_stat
> Owner: Nobody
> Requestors: jkunke at fas.harvard.edu
> Status: new
> Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051 >
>
>
> Good morning,
>
> I have model forecasts at one-hour snapshots over an 8-hour period,
and I
> am looking to compare a number of stations at each of those times.
I have
> read the MET tutorial and the users' guide and I have a couple
remaining
> questions.
>
> I am using USCRN station data for my point observations, and I can
convert
> them into a MET-formatted ascii file to use with ascii2nc and then
with
> point_stat.
>
> 1. Can I have more times than desired in my MET-formatted ascii file
and
> then use the configuration file or command line options for ascii2nc
to
> select only a certain time range from these observations to include
in the
> netCDF file, or will it necessarily use all the observations I have
in the
> ascii file I give to ascii2nc?
>
> 2. If I have point observations at multiple times in one ascii or
netcdf
> file that I give as input to point_stat, can point_stat compare my
hourly
> forecasts to this hourly data over the whole time period in one
call? Or
> do I need to have only the observations I want to compare to a given
> forecast (a one-time forecast) and just call point_stat for each
forecast
> time as a separate command, using a separate input file of
observations?
>
> Thank you,
> Jess
>
>
--
Julie Prestopnik
National Center for Atmospheric Research
Research Applications Laboratory
Phone: 303.497.8399
Email: jpresto at ucar.edu
------------------------------------------------
Subject: Question about ascii2nc and point_stat
From: Kunke, Jessica
Time: Tue Mar 10 13:22:15 2015
Hi Julie,
Thank you, I think this answers my questions. I wasn't sure whether
the
obs_valid_beg and obs_valid_end command line options for point_stat
would
specify the time range of observations to be applied to a *single
*forecast
time (in which they seem to just directly override the obs_window
setting
in the config file) or whether in a single point_stat call I could use
these command line options to validate multiple forecast times. It
sounds
like the former is true, is that right?
I have multiple forecasts in one file– hourly output from 2013-08-30
06:00:00 UTC to 2013-08-31 00:00:00 UTC, for a total of 18 different
times
(plus the initial output). I may not necessarily want to validate all
the
forecasts, but I would like to validate more than one.
Thank you for your help!
Jess
On Tue, Mar 10, 2015 at 2:27 PM, Julie Prestopnik via RT
<met_help at ucar.edu>
wrote:
> Hi Jessica.
>
> You can put observations for as many times as you would like in your
> MET-formatted ascii file. ascii2nc will pass them all through the
output
> NetCDF file, but they can be filtered out when using Point-Stat.
>
> The Point-Stat configuration file has an obs_window where you can
specify a
> begin and end window. The obs_window is defined relative to the
forecast
> time. So you could specify (in seconds), for example, one hour plus
or
> minus the forecast time. If your forecast hour was noon, that would
tell
> it to grab all observations between 11am and 1pm.
>
> Although you could have a ton of data in your ascii2nc output file,
> Point-Stat would need to read it in each time, so it could
potentially be
> slow.
>
> It wasn't clear if you have 8 forecast files or if your forecast
times are
> all in one file. It would be easiest to run Point-Stat if you had 8
> separate forecast files. You would call Point-Stat 8 times, once
for each
> forecast file you have. If you have multiple forecasts in one file,
please
> let us know and we can walk you through running Point-Stat,
modifying the
> configuration file, to account for multiple forecasts in one file.
>
> Please let me know if you have any other questions.
>
> Thanks,
> Julie
>
> On Tue, Mar 10, 2015 at 9:38 AM, Kunke, Jessica via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > Tue Mar 10 09:38:01 2015: Request 71051 was acted upon.
> > Transaction: Ticket created by jkunke at fas.harvard.edu
> > Queue: met_help
> > Subject: Question about ascii2nc and point_stat
> > Owner: Nobody
> > Requestors: jkunke at fas.harvard.edu
> > Status: new
> > Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051 >
> >
> >
> > Good morning,
> >
> > I have model forecasts at one-hour snapshots over an 8-hour
period, and I
> > am looking to compare a number of stations at each of those times.
I
> have
> > read the MET tutorial and the users' guide and I have a couple
remaining
> > questions.
> >
> > I am using USCRN station data for my point observations, and I can
> convert
> > them into a MET-formatted ascii file to use with ascii2nc and then
with
> > point_stat.
> >
> > 1. Can I have more times than desired in my MET-formatted ascii
file and
> > then use the configuration file or command line options for
ascii2nc to
> > select only a certain time range from these observations to
include in
> the
> > netCDF file, or will it necessarily use all the observations I
have in
> the
> > ascii file I give to ascii2nc?
> >
> > 2. If I have point observations at multiple times in one ascii or
netcdf
> > file that I give as input to point_stat, can point_stat compare my
hourly
> > forecasts to this hourly data over the whole time period in one
call? Or
> > do I need to have only the observations I want to compare to a
given
> > forecast (a one-time forecast) and just call point_stat for each
forecast
> > time as a separate command, using a separate input file of
observations?
> >
> > Thank you,
> > Jess
> >
> >
>
>
> --
> Julie Prestopnik
> National Center for Atmospheric Research
> Research Applications Laboratory
> Phone: 303.497.8399
> Email: jpresto at ucar.edu
>
>
------------------------------------------------
Subject: Question about ascii2nc and point_stat
From: Julie Prestopnik
Time: Tue Mar 10 14:20:20 2015
Hi Jess.
The obs_valid_beg and obs_valid_end command line arguments for
point_stat
are used to explicitly set the matching time window to specific values
and
override the "obs_window" config file setting. You'll want to use
obs_window instead to define the matching time window relative to the
valid
time of the forecast field being evaluated.
Since you have multiple forecasts in one file, you'll want to specify
the
lead_time or the valid_time in the configuration file. So, for
example,
say that you wanted to verify the 2 meter TMP three times (for lead
times
12, 15, and 18), in the field value in the "fcst" field of the
configuration file, you'd want to have:
field = [
{
name = "TMP";
level = [ "Z2" ];
cat_thresh = [ >0.0 ];
lead_time = "12";
},
{
name = "TMP";
level = [ "Z2" ];
cat_thresh = [ >0.0 ];
lead_time = "15";
},
{
name = "TMP";
level = [ "Z2" ];
cat_thresh = [ >0.0 ];
lead_time = "18";
}
];
I hope that helps. Please let us know if you have any other
questions.
Thanks,
Julie
On Tue, Mar 10, 2015 at 1:22 PM, Kunke, Jessica via RT
<met_help at ucar.edu>
wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051 >
>
> Hi Julie,
>
> Thank you, I think this answers my questions. I wasn't sure whether
the
> obs_valid_beg and obs_valid_end command line options for point_stat
would
> specify the time range of observations to be applied to a *single
*forecast
> time (in which they seem to just directly override the obs_window
setting
> in the config file) or whether in a single point_stat call I could
use
> these command line options to validate multiple forecast times. It
sounds
> like the former is true, is that right?
>
> I have multiple forecasts in one file– hourly output from 2013-08-30
> 06:00:00 UTC to 2013-08-31 00:00:00 UTC, for a total of 18 different
times
> (plus the initial output). I may not necessarily want to validate
all the
> forecasts, but I would like to validate more than one.
>
> Thank you for your help!
> Jess
>
> On Tue, Mar 10, 2015 at 2:27 PM, Julie Prestopnik via RT <
> met_help at ucar.edu>
> wrote:
>
> > Hi Jessica.
> >
> > You can put observations for as many times as you would like in
your
> > MET-formatted ascii file. ascii2nc will pass them all through the
output
> > NetCDF file, but they can be filtered out when using Point-Stat.
> >
> > The Point-Stat configuration file has an obs_window where you can
> specify a
> > begin and end window. The obs_window is defined relative to the
forecast
> > time. So you could specify (in seconds), for example, one hour
plus or
> > minus the forecast time. If your forecast hour was noon, that
would tell
> > it to grab all observations between 11am and 1pm.
> >
> > Although you could have a ton of data in your ascii2nc output
file,
> > Point-Stat would need to read it in each time, so it could
potentially be
> > slow.
> >
> > It wasn't clear if you have 8 forecast files or if your forecast
times
> are
> > all in one file. It would be easiest to run Point-Stat if you had
8
> > separate forecast files. You would call Point-Stat 8 times, once
for
> each
> > forecast file you have. If you have multiple forecasts in one
file,
> please
> > let us know and we can walk you through running Point-Stat,
modifying the
> > configuration file, to account for multiple forecasts in one file.
> >
> > Please let me know if you have any other questions.
> >
> > Thanks,
> > Julie
> >
> > On Tue, Mar 10, 2015 at 9:38 AM, Kunke, Jessica via RT <
> met_help at ucar.edu>
> > wrote:
> >
> > >
> > > Tue Mar 10 09:38:01 2015: Request 71051 was acted upon.
> > > Transaction: Ticket created by jkunke at fas.harvard.edu
> > > Queue: met_help
> > > Subject: Question about ascii2nc and point_stat
> > > Owner: Nobody
> > > Requestors: jkunke at fas.harvard.edu
> > > Status: new
> > > Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051
> >
> > >
> > >
> > > Good morning,
> > >
> > > I have model forecasts at one-hour snapshots over an 8-hour
period,
> and I
> > > am looking to compare a number of stations at each of those
times. I
> > have
> > > read the MET tutorial and the users' guide and I have a couple
> remaining
> > > questions.
> > >
> > > I am using USCRN station data for my point observations, and I
can
> > convert
> > > them into a MET-formatted ascii file to use with ascii2nc and
then with
> > > point_stat.
> > >
> > > 1. Can I have more times than desired in my MET-formatted ascii
file
> and
> > > then use the configuration file or command line options for
ascii2nc to
> > > select only a certain time range from these observations to
include in
> > the
> > > netCDF file, or will it necessarily use all the observations I
have in
> > the
> > > ascii file I give to ascii2nc?
> > >
> > > 2. If I have point observations at multiple times in one ascii
or
> netcdf
> > > file that I give as input to point_stat, can point_stat compare
my
> hourly
> > > forecasts to this hourly data over the whole time period in one
call?
> Or
> > > do I need to have only the observations I want to compare to a
given
> > > forecast (a one-time forecast) and just call point_stat for each
> forecast
> > > time as a separate command, using a separate input file of
> observations?
> > >
> > > Thank you,
> > > Jess
> > >
> > >
> >
> >
> > --
> > Julie Prestopnik
> > National Center for Atmospheric Research
> > Research Applications Laboratory
> > Phone: 303.497.8399
> > Email: jpresto at ucar.edu
> >
> >
>
>
--
Julie Prestopnik
National Center for Atmospheric Research
Research Applications Laboratory
Phone: 303.497.8399
Email: jpresto at ucar.edu
------------------------------------------------
Subject: Question about ascii2nc and point_stat
From: Kunke, Jessica
Time: Tue Mar 10 15:43:38 2015
Hi Julie,
Ah, I see. Thank you, that clarifies the obs_window and obs_valid
settings
and also how to handle multiple forecast times. So lead_time is the
number
of hours after the initial time, or the number of forecast snapshots
after
the initial time?
I read the section of the MET users' guide about the point_stat config
file
but I don't fully understand how the config file works. Below is the
example on page 73-74 of the users' guide as an example to consider.
If I
want to use all message types in the ascii input file, for instance,
can I
just omit the message_type line, or do I use empty [] or list them all
within this line? Does the fact that the first five lines within
"fcst"
are outside of "field" mean that they apply to all the variables
listed
within "field"? Does setting a value to [] set it to null or set it
to
whatever values are found within the input ascii file? Is there a
complete
list somewhere of all the fields we can set here like wind_thresh,
message_type, name, level, etc and what they mean? For instance, the
users' guide says wind_thresh filters the winds based on wind speed,
but I
don't understand the syntax of the wind_thresh statement or what
"wind_thresh = [ NA ]" will do. I also don't understand what
thresholds
are in this case or why at least one must be set for each variable
within
"field."
Thank you,
Jess
fcst = {
wind_thresh = [ NA ];
message_type = [ "ADPUPA", "ADPSFC" ];
sid_exc = [];
lead_time = "12";
valid_time = "20070331_12";
field = [
{
name = "TMP";
level = [ "P750-900" ];
cat_thresh = [ <=273, >273 ];
},
{
name = "HGT";
level = [ "P500" ];
cat_thresh = [ >0.0 ];
},
{
name = "UGRD";
level = [ "P500" ];
cat_thresh = [ >5.0 ];
},
{
name = "VGRD";
level = [ "P500" ];
cat_thresh = [ >5.0 ];
}
];
};
obs = fcst;
On Tue, Mar 10, 2015 at 4:20 PM, Julie Prestopnik via RT
<met_help at ucar.edu>
wrote:
> Hi Jess.
>
> The obs_valid_beg and obs_valid_end command line arguments for
point_stat
> are used to explicitly set the matching time window to specific
values and
> override the "obs_window" config file setting. You'll want to use
> obs_window instead to define the matching time window relative to
the valid
> time of the forecast field being evaluated.
>
> Since you have multiple forecasts in one file, you'll want to
specify the
> lead_time or the valid_time in the configuration file. So, for
example,
> say that you wanted to verify the 2 meter TMP three times (for lead
times
> 12, 15, and 18), in the field value in the "fcst" field of the
> configuration file, you'd want to have:
> field = [
> {
> name = "TMP";
> level = [ "Z2" ];
> cat_thresh = [ >0.0 ];
> lead_time = "12";
> },
> {
> name = "TMP";
> level = [ "Z2" ];
> cat_thresh = [ >0.0 ];
> lead_time = "15";
> },
> {
> name = "TMP";
> level = [ "Z2" ];
> cat_thresh = [ >0.0 ];
> lead_time = "18";
> }
> ];
>
> I hope that helps. Please let us know if you have any other
questions.
>
> Thanks,
> Julie
>
>
> On Tue, Mar 10, 2015 at 1:22 PM, Kunke, Jessica via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051 >
> >
> > Hi Julie,
> >
> > Thank you, I think this answers my questions. I wasn't sure
whether the
> > obs_valid_beg and obs_valid_end command line options for
point_stat would
> > specify the time range of observations to be applied to a *single
> *forecast
> > time (in which they seem to just directly override the obs_window
setting
> > in the config file) or whether in a single point_stat call I could
use
> > these command line options to validate multiple forecast times.
It
> sounds
> > like the former is true, is that right?
> >
> > I have multiple forecasts in one file– hourly output from 2013-08-
30
> > 06:00:00 UTC to 2013-08-31 00:00:00 UTC, for a total of 18
different
> times
> > (plus the initial output). I may not necessarily want to validate
all
> the
> > forecasts, but I would like to validate more than one.
> >
> > Thank you for your help!
> > Jess
> >
> > On Tue, Mar 10, 2015 at 2:27 PM, Julie Prestopnik via RT <
> > met_help at ucar.edu>
> > wrote:
> >
> > > Hi Jessica.
> > >
> > > You can put observations for as many times as you would like in
your
> > > MET-formatted ascii file. ascii2nc will pass them all through
the
> output
> > > NetCDF file, but they can be filtered out when using Point-Stat.
> > >
> > > The Point-Stat configuration file has an obs_window where you
can
> > specify a
> > > begin and end window. The obs_window is defined relative to the
> forecast
> > > time. So you could specify (in seconds), for example, one hour
plus or
> > > minus the forecast time. If your forecast hour was noon, that
would
> tell
> > > it to grab all observations between 11am and 1pm.
> > >
> > > Although you could have a ton of data in your ascii2nc output
file,
> > > Point-Stat would need to read it in each time, so it could
potentially
> be
> > > slow.
> > >
> > > It wasn't clear if you have 8 forecast files or if your forecast
times
> > are
> > > all in one file. It would be easiest to run Point-Stat if you
had 8
> > > separate forecast files. You would call Point-Stat 8 times,
once for
> > each
> > > forecast file you have. If you have multiple forecasts in one
file,
> > please
> > > let us know and we can walk you through running Point-Stat,
modifying
> the
> > > configuration file, to account for multiple forecasts in one
file.
> > >
> > > Please let me know if you have any other questions.
> > >
> > > Thanks,
> > > Julie
> > >
> > > On Tue, Mar 10, 2015 at 9:38 AM, Kunke, Jessica via RT <
> > met_help at ucar.edu>
> > > wrote:
> > >
> > > >
> > > > Tue Mar 10 09:38:01 2015: Request 71051 was acted upon.
> > > > Transaction: Ticket created by jkunke at fas.harvard.edu
> > > > Queue: met_help
> > > > Subject: Question about ascii2nc and point_stat
> > > > Owner: Nobody
> > > > Requestors: jkunke at fas.harvard.edu
> > > > Status: new
> > > > Ticket <URL:
> https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051
> > >
> > > >
> > > >
> > > > Good morning,
> > > >
> > > > I have model forecasts at one-hour snapshots over an 8-hour
period,
> > and I
> > > > am looking to compare a number of stations at each of those
times. I
> > > have
> > > > read the MET tutorial and the users' guide and I have a couple
> > remaining
> > > > questions.
> > > >
> > > > I am using USCRN station data for my point observations, and I
can
> > > convert
> > > > them into a MET-formatted ascii file to use with ascii2nc and
then
> with
> > > > point_stat.
> > > >
> > > > 1. Can I have more times than desired in my MET-formatted
ascii file
> > and
> > > > then use the configuration file or command line options for
ascii2nc
> to
> > > > select only a certain time range from these observations to
include
> in
> > > the
> > > > netCDF file, or will it necessarily use all the observations I
have
> in
> > > the
> > > > ascii file I give to ascii2nc?
> > > >
> > > > 2. If I have point observations at multiple times in one ascii
or
> > netcdf
> > > > file that I give as input to point_stat, can point_stat
compare my
> > hourly
> > > > forecasts to this hourly data over the whole time period in
one call?
> > Or
> > > > do I need to have only the observations I want to compare to a
given
> > > > forecast (a one-time forecast) and just call point_stat for
each
> > forecast
> > > > time as a separate command, using a separate input file of
> > observations?
> > > >
> > > > Thank you,
> > > > Jess
> > > >
> > > >
> > >
> > >
> > > --
> > > Julie Prestopnik
> > > National Center for Atmospheric Research
> > > Research Applications Laboratory
> > > Phone: 303.497.8399
> > > Email: jpresto at ucar.edu
> > >
> > >
> >
> >
>
>
> --
> Julie Prestopnik
> National Center for Atmospheric Research
> Research Applications Laboratory
> Phone: 303.497.8399
> Email: jpresto at ucar.edu
>
>
------------------------------------------------
Subject: Question about ascii2nc and point_stat
From: John Halley Gotway
Time: Tue Mar 10 17:10:15 2015
Hello Jess,
This is John Halley Gotway. I work with Julie on met-help support. I
thought I'd jump in here on this one.
I see that you're asking how to set up the configuration file for
point-stat. Please take a look at met-5.0/data/config/README for more
detailed information about the config file settings. Have you been
able to
get point-stat to run and produce meaningful results yet? I suggest
giving
it a shot and then we can help you refine the results.
The message_type setting specifies which observation types should be
used
for each verification task. Setting it once at the top (as shown in
the
example you sent) results in all possible combinations being tried.
Meaning TMP/P750-900 (and all the other fields) are compared to ADPSFC
observations and then also compared to ADPUPA observations. If you
get
matched pairs for both, you'll see two lines in the output - one for
ADPSFC
observations and another for ADPUPA. Each verification task is done
using
exactly one observation type. There is no way to combine them (i.e.
ADPSFC
*and* ADPUPA) directly in point_stat. However, after you've run
point-stat, you could run stat-analysis to aggregate results across
multiple observation types, if that makes sense scientifically.
In the example you sent, message_type is set once at the top and
applied to
all the fields. Alternatively, you could define message_type for each
individual field to be evaluated, as shown below:
fcst = {
wind_thresh = [ NA ];
sid_exc = [];
lead_time = "12";
valid_time = "20070331_12";
field = [
{
name = "TMP";
level = [ "P500" ];
cat_thresh = [ <=273, >273 ];
message_type = [ "ADPUPA" ];
},
{
name = "TMP";
level = [ "Z2" ];
cat_thresh = [ >273 ];
message_type = [ "ADPSFC" ];
}
];
};
obs = fcst;
Typically, you'd compare upper-air forecast (TMP/P500) to upper-air
observation (ADPUPA) and surface forecast (TMP/Z2) to surface
observation
(ADPSFC).
Does that help explain how the message_type setting works?
Here's a selection from met-5.0/data/config/README about
wind_thresh...
// - The "wind_thresh" entry is an array of thresholds used to
filter wind
// speed values when computing VL1L2 vector partial sums. Only
those
U/V
// pairs that meet this wind speed criteria will be included in
the
sums.
// Setting this threshold to NA will result in all U/V pairs being
used.
And here's an example of how you might set it:
wind_thresh = [ >1, >2, >3 ];
Note that this only applies to which U/V pairs are included in the
VL1L2
output line types. Instead of getting 1 VL1L2 output line, you'd get
3...
one with U/V pairs whose wind speed exceed 1m/s, one with wind speed >
2m/s, and one with wind speed > 3m/s. But if you're interested in
wind
direction performance at individual stations, you don't really need to
use
this. The stat-analysis tool in met-5.0 includes some enhancements
that
make this easier. Be sure to dump out the MPR line type for the UGRD
and
VGRD variables and then we can help you to run stat-analysis on them.
Thanks,
John
On Tue, Mar 10, 2015 at 3:43 PM, Kunke, Jessica via RT
<met_help at ucar.edu>
wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051 >
>
> Hi Julie,
>
> Ah, I see. Thank you, that clarifies the obs_window and obs_valid
settings
> and also how to handle multiple forecast times. So lead_time is the
number
> of hours after the initial time, or the number of forecast snapshots
after
> the initial time?
>
> I read the section of the MET users' guide about the point_stat
config file
> but I don't fully understand how the config file works. Below is
the
> example on page 73-74 of the users' guide as an example to consider.
If I
> want to use all message types in the ascii input file, for instance,
can I
> just omit the message_type line, or do I use empty [] or list them
all
> within this line? Does the fact that the first five lines within
"fcst"
> are outside of "field" mean that they apply to all the variables
listed
> within "field"? Does setting a value to [] set it to null or set it
to
> whatever values are found within the input ascii file? Is there a
complete
> list somewhere of all the fields we can set here like wind_thresh,
> message_type, name, level, etc and what they mean? For instance,
the
> users' guide says wind_thresh filters the winds based on wind speed,
but I
> don't understand the syntax of the wind_thresh statement or what
> "wind_thresh = [ NA ]" will do. I also don't understand what
thresholds
> are in this case or why at least one must be set for each variable
within
> "field."
>
> Thank you,
> Jess
>
> fcst = {
> wind_thresh = [ NA ];
> message_type = [ "ADPUPA", "ADPSFC" ];
> sid_exc = [];
> lead_time = "12";
> valid_time = "20070331_12";
>
> field = [
> {
> name = "TMP";
> level = [ "P750-900" ];
> cat_thresh = [ <=273, >273 ];
> },
> {
> name = "HGT";
> level = [ "P500" ];
> cat_thresh = [ >0.0 ];
> },
> {
> name = "UGRD";
> level = [ "P500" ];
> cat_thresh = [ >5.0 ];
> },
> {
> name = "VGRD";
> level = [ "P500" ];
> cat_thresh = [ >5.0 ];
> }
> ];
> };
> obs = fcst;
>
> On Tue, Mar 10, 2015 at 4:20 PM, Julie Prestopnik via RT <
> met_help at ucar.edu>
> wrote:
>
> > Hi Jess.
> >
> > The obs_valid_beg and obs_valid_end command line arguments for
point_stat
> > are used to explicitly set the matching time window to specific
values
> and
> > override the "obs_window" config file setting. You'll want to use
> > obs_window instead to define the matching time window relative to
the
> valid
> > time of the forecast field being evaluated.
> >
> > Since you have multiple forecasts in one file, you'll want to
specify the
> > lead_time or the valid_time in the configuration file. So, for
example,
> > say that you wanted to verify the 2 meter TMP three times (for
lead times
> > 12, 15, and 18), in the field value in the "fcst" field of the
> > configuration file, you'd want to have:
> > field = [
> > {
> > name = "TMP";
> > level = [ "Z2" ];
> > cat_thresh = [ >0.0 ];
> > lead_time = "12";
> > },
> > {
> > name = "TMP";
> > level = [ "Z2" ];
> > cat_thresh = [ >0.0 ];
> > lead_time = "15";
> > },
> > {
> > name = "TMP";
> > level = [ "Z2" ];
> > cat_thresh = [ >0.0 ];
> > lead_time = "18";
> > }
> > ];
> >
> > I hope that helps. Please let us know if you have any other
questions.
> >
> > Thanks,
> > Julie
> >
> >
> > On Tue, Mar 10, 2015 at 1:22 PM, Kunke, Jessica via RT <
> met_help at ucar.edu>
> > wrote:
> >
> > >
> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051 >
> > >
> > > Hi Julie,
> > >
> > > Thank you, I think this answers my questions. I wasn't sure
whether
> the
> > > obs_valid_beg and obs_valid_end command line options for
point_stat
> would
> > > specify the time range of observations to be applied to a
*single
> > *forecast
> > > time (in which they seem to just directly override the
obs_window
> setting
> > > in the config file) or whether in a single point_stat call I
could use
> > > these command line options to validate multiple forecast times.
It
> > sounds
> > > like the former is true, is that right?
> > >
> > > I have multiple forecasts in one file– hourly output from 2013-
08-30
> > > 06:00:00 UTC to 2013-08-31 00:00:00 UTC, for a total of 18
different
> > times
> > > (plus the initial output). I may not necessarily want to
validate all
> > the
> > > forecasts, but I would like to validate more than one.
> > >
> > > Thank you for your help!
> > > Jess
> > >
> > > On Tue, Mar 10, 2015 at 2:27 PM, Julie Prestopnik via RT <
> > > met_help at ucar.edu>
> > > wrote:
> > >
> > > > Hi Jessica.
> > > >
> > > > You can put observations for as many times as you would like
in your
> > > > MET-formatted ascii file. ascii2nc will pass them all through
the
> > output
> > > > NetCDF file, but they can be filtered out when using Point-
Stat.
> > > >
> > > > The Point-Stat configuration file has an obs_window where you
can
> > > specify a
> > > > begin and end window. The obs_window is defined relative to
the
> > forecast
> > > > time. So you could specify (in seconds), for example, one
hour plus
> or
> > > > minus the forecast time. If your forecast hour was noon, that
would
> > tell
> > > > it to grab all observations between 11am and 1pm.
> > > >
> > > > Although you could have a ton of data in your ascii2nc output
file,
> > > > Point-Stat would need to read it in each time, so it could
> potentially
> > be
> > > > slow.
> > > >
> > > > It wasn't clear if you have 8 forecast files or if your
forecast
> times
> > > are
> > > > all in one file. It would be easiest to run Point-Stat if you
had 8
> > > > separate forecast files. You would call Point-Stat 8 times,
once for
> > > each
> > > > forecast file you have. If you have multiple forecasts in one
file,
> > > please
> > > > let us know and we can walk you through running Point-Stat,
modifying
> > the
> > > > configuration file, to account for multiple forecasts in one
file.
> > > >
> > > > Please let me know if you have any other questions.
> > > >
> > > > Thanks,
> > > > Julie
> > > >
> > > > On Tue, Mar 10, 2015 at 9:38 AM, Kunke, Jessica via RT <
> > > met_help at ucar.edu>
> > > > wrote:
> > > >
> > > > >
> > > > > Tue Mar 10 09:38:01 2015: Request 71051 was acted upon.
> > > > > Transaction: Ticket created by jkunke at fas.harvard.edu
> > > > > Queue: met_help
> > > > > Subject: Question about ascii2nc and point_stat
> > > > > Owner: Nobody
> > > > > Requestors: jkunke at fas.harvard.edu
> > > > > Status: new
> > > > > Ticket <URL:
> > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051
> > > >
> > > > >
> > > > >
> > > > > Good morning,
> > > > >
> > > > > I have model forecasts at one-hour snapshots over an 8-hour
period,
> > > and I
> > > > > am looking to compare a number of stations at each of those
> times. I
> > > > have
> > > > > read the MET tutorial and the users' guide and I have a
couple
> > > remaining
> > > > > questions.
> > > > >
> > > > > I am using USCRN station data for my point observations, and
I can
> > > > convert
> > > > > them into a MET-formatted ascii file to use with ascii2nc
and then
> > with
> > > > > point_stat.
> > > > >
> > > > > 1. Can I have more times than desired in my MET-formatted
ascii
> file
> > > and
> > > > > then use the configuration file or command line options for
> ascii2nc
> > to
> > > > > select only a certain time range from these observations to
include
> > in
> > > > the
> > > > > netCDF file, or will it necessarily use all the observations
I have
> > in
> > > > the
> > > > > ascii file I give to ascii2nc?
> > > > >
> > > > > 2. If I have point observations at multiple times in one
ascii or
> > > netcdf
> > > > > file that I give as input to point_stat, can point_stat
compare my
> > > hourly
> > > > > forecasts to this hourly data over the whole time period in
one
> call?
> > > Or
> > > > > do I need to have only the observations I want to compare to
a
> given
> > > > > forecast (a one-time forecast) and just call point_stat for
each
> > > forecast
> > > > > time as a separate command, using a separate input file of
> > > observations?
> > > > >
> > > > > Thank you,
> > > > > Jess
> > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > Julie Prestopnik
> > > > National Center for Atmospheric Research
> > > > Research Applications Laboratory
> > > > Phone: 303.497.8399
> > > > Email: jpresto at ucar.edu
> > > >
> > > >
> > >
> > >
> >
> >
> > --
> > Julie Prestopnik
> > National Center for Atmospheric Research
> > Research Applications Laboratory
> > Phone: 303.497.8399
> > Email: jpresto at ucar.edu
> >
> >
>
>
------------------------------------------------
Subject: Question about ascii2nc and point_stat
From: Kunke, Jessica
Time: Thu Mar 12 10:27:25 2015
Hi John,
Thank you for your thorough response– this helps me a lot. Also thank
you
for pointing me to the README file which I think will be very helpful.
I
will work with this a bit and get back to you with any further
questions.
And thank you, Julie, for all your help too!
Best wishes,
Jess
On Tue, Mar 10, 2015 at 7:10 PM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:
> Hello Jess,
>
> This is John Halley Gotway. I work with Julie on met-help support.
I
> thought I'd jump in here on this one.
>
> I see that you're asking how to set up the configuration file for
> point-stat. Please take a look at met-5.0/data/config/README for
more
> detailed information about the config file settings. Have you been
able to
> get point-stat to run and produce meaningful results yet? I suggest
giving
> it a shot and then we can help you refine the results.
>
> The message_type setting specifies which observation types should be
used
> for each verification task. Setting it once at the top (as shown in
the
> example you sent) results in all possible combinations being tried.
> Meaning TMP/P750-900 (and all the other fields) are compared to
ADPSFC
> observations and then also compared to ADPUPA observations. If you
get
> matched pairs for both, you'll see two lines in the output - one for
ADPSFC
> observations and another for ADPUPA. Each verification task is done
using
> exactly one observation type. There is no way to combine them (i.e.
ADPSFC
> *and* ADPUPA) directly in point_stat. However, after you've run
> point-stat, you could run stat-analysis to aggregate results across
> multiple observation types, if that makes sense scientifically.
>
> In the example you sent, message_type is set once at the top and
applied to
> all the fields. Alternatively, you could define message_type for
each
> individual field to be evaluated, as shown below:
>
> fcst = {
> wind_thresh = [ NA ];
> sid_exc = [];
> lead_time = "12";
> valid_time = "20070331_12";
>
> field = [
> {
> name = "TMP";
> level = [ "P500" ];
> cat_thresh = [ <=273, >273 ];
> message_type = [ "ADPUPA" ];
> },
> {
> name = "TMP";
> level = [ "Z2" ];
> cat_thresh = [ >273 ];
> message_type = [ "ADPSFC" ];
> }
> ];
> };
> obs = fcst;
>
> Typically, you'd compare upper-air forecast (TMP/P500) to upper-air
> observation (ADPUPA) and surface forecast (TMP/Z2) to surface
observation
> (ADPSFC).
>
> Does that help explain how the message_type setting works?
>
> Here's a selection from met-5.0/data/config/README about
wind_thresh...
> // - The "wind_thresh" entry is an array of thresholds used to
filter
> wind
> // speed values when computing VL1L2 vector partial sums. Only
those
> U/V
> // pairs that meet this wind speed criteria will be included in
the
> sums.
> // Setting this threshold to NA will result in all U/V pairs
being
> used.
>
> And here's an example of how you might set it:
> wind_thresh = [ >1, >2, >3 ];
>
> Note that this only applies to which U/V pairs are included in the
VL1L2
> output line types. Instead of getting 1 VL1L2 output line, you'd
get 3...
> one with U/V pairs whose wind speed exceed 1m/s, one with wind speed
>
> 2m/s, and one with wind speed > 3m/s. But if you're interested in
wind
> direction performance at individual stations, you don't really need
to use
> this. The stat-analysis tool in met-5.0 includes some enhancements
that
> make this easier. Be sure to dump out the MPR line type for the
UGRD and
> VGRD variables and then we can help you to run stat-analysis on
them.
>
> Thanks,
> John
>
>
>
> On Tue, Mar 10, 2015 at 3:43 PM, Kunke, Jessica via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051 >
> >
> > Hi Julie,
> >
> > Ah, I see. Thank you, that clarifies the obs_window and obs_valid
> settings
> > and also how to handle multiple forecast times. So lead_time is
the
> number
> > of hours after the initial time, or the number of forecast
snapshots
> after
> > the initial time?
> >
> > I read the section of the MET users' guide about the point_stat
config
> file
> > but I don't fully understand how the config file works. Below is
the
> > example on page 73-74 of the users' guide as an example to
consider. If
> I
> > want to use all message types in the ascii input file, for
instance, can
> I
> > just omit the message_type line, or do I use empty [] or list them
all
> > within this line? Does the fact that the first five lines within
"fcst"
> > are outside of "field" mean that they apply to all the variables
listed
> > within "field"? Does setting a value to [] set it to null or set
it to
> > whatever values are found within the input ascii file? Is there a
> complete
> > list somewhere of all the fields we can set here like wind_thresh,
> > message_type, name, level, etc and what they mean? For instance,
the
> > users' guide says wind_thresh filters the winds based on wind
speed, but
> I
> > don't understand the syntax of the wind_thresh statement or what
> > "wind_thresh = [ NA ]" will do. I also don't understand what
thresholds
> > are in this case or why at least one must be set for each variable
within
> > "field."
> >
> > Thank you,
> > Jess
> >
> > fcst = {
> > wind_thresh = [ NA ];
> > message_type = [ "ADPUPA", "ADPSFC" ];
> > sid_exc = [];
> > lead_time = "12";
> > valid_time = "20070331_12";
> >
> > field = [
> > {
> > name = "TMP";
> > level = [ "P750-900" ];
> > cat_thresh = [ <=273, >273 ];
> > },
> > {
> > name = "HGT";
> > level = [ "P500" ];
> > cat_thresh = [ >0.0 ];
> > },
> > {
> > name = "UGRD";
> > level = [ "P500" ];
> > cat_thresh = [ >5.0 ];
> > },
> > {
> > name = "VGRD";
> > level = [ "P500" ];
> > cat_thresh = [ >5.0 ];
> > }
> > ];
> > };
> > obs = fcst;
> >
> > On Tue, Mar 10, 2015 at 4:20 PM, Julie Prestopnik via RT <
> > met_help at ucar.edu>
> > wrote:
> >
> > > Hi Jess.
> > >
> > > The obs_valid_beg and obs_valid_end command line arguments for
> point_stat
> > > are used to explicitly set the matching time window to specific
values
> > and
> > > override the "obs_window" config file setting. You'll want to
use
> > > obs_window instead to define the matching time window relative
to the
> > valid
> > > time of the forecast field being evaluated.
> > >
> > > Since you have multiple forecasts in one file, you'll want to
specify
> the
> > > lead_time or the valid_time in the configuration file. So, for
> example,
> > > say that you wanted to verify the 2 meter TMP three times (for
lead
> times
> > > 12, 15, and 18), in the field value in the "fcst" field of the
> > > configuration file, you'd want to have:
> > > field = [
> > > {
> > > name = "TMP";
> > > level = [ "Z2" ];
> > > cat_thresh = [ >0.0 ];
> > > lead_time = "12";
> > > },
> > > {
> > > name = "TMP";
> > > level = [ "Z2" ];
> > > cat_thresh = [ >0.0 ];
> > > lead_time = "15";
> > > },
> > > {
> > > name = "TMP";
> > > level = [ "Z2" ];
> > > cat_thresh = [ >0.0 ];
> > > lead_time = "18";
> > > }
> > > ];
> > >
> > > I hope that helps. Please let us know if you have any other
questions.
> > >
> > > Thanks,
> > > Julie
> > >
> > >
> > > On Tue, Mar 10, 2015 at 1:22 PM, Kunke, Jessica via RT <
> > met_help at ucar.edu>
> > > wrote:
> > >
> > > >
> > > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051
>
> > > >
> > > > Hi Julie,
> > > >
> > > > Thank you, I think this answers my questions. I wasn't sure
whether
> > the
> > > > obs_valid_beg and obs_valid_end command line options for
point_stat
> > would
> > > > specify the time range of observations to be applied to a
*single
> > > *forecast
> > > > time (in which they seem to just directly override the
obs_window
> > setting
> > > > in the config file) or whether in a single point_stat call I
could
> use
> > > > these command line options to validate multiple forecast
times. It
> > > sounds
> > > > like the former is true, is that right?
> > > >
> > > > I have multiple forecasts in one file– hourly output from
2013-08-30
> > > > 06:00:00 UTC to 2013-08-31 00:00:00 UTC, for a total of 18
different
> > > times
> > > > (plus the initial output). I may not necessarily want to
validate
> all
> > > the
> > > > forecasts, but I would like to validate more than one.
> > > >
> > > > Thank you for your help!
> > > > Jess
> > > >
> > > > On Tue, Mar 10, 2015 at 2:27 PM, Julie Prestopnik via RT <
> > > > met_help at ucar.edu>
> > > > wrote:
> > > >
> > > > > Hi Jessica.
> > > > >
> > > > > You can put observations for as many times as you would like
in
> your
> > > > > MET-formatted ascii file. ascii2nc will pass them all
through the
> > > output
> > > > > NetCDF file, but they can be filtered out when using Point-
Stat.
> > > > >
> > > > > The Point-Stat configuration file has an obs_window where
you can
> > > > specify a
> > > > > begin and end window. The obs_window is defined relative to
the
> > > forecast
> > > > > time. So you could specify (in seconds), for example, one
hour
> plus
> > or
> > > > > minus the forecast time. If your forecast hour was noon,
that
> would
> > > tell
> > > > > it to grab all observations between 11am and 1pm.
> > > > >
> > > > > Although you could have a ton of data in your ascii2nc
output file,
> > > > > Point-Stat would need to read it in each time, so it could
> > potentially
> > > be
> > > > > slow.
> > > > >
> > > > > It wasn't clear if you have 8 forecast files or if your
forecast
> > times
> > > > are
> > > > > all in one file. It would be easiest to run Point-Stat if
you had 8
> > > > > separate forecast files. You would call Point-Stat 8 times,
once
> for
> > > > each
> > > > > forecast file you have. If you have multiple forecasts in
one
> file,
> > > > please
> > > > > let us know and we can walk you through running Point-Stat,
> modifying
> > > the
> > > > > configuration file, to account for multiple forecasts in one
file.
> > > > >
> > > > > Please let me know if you have any other questions.
> > > > >
> > > > > Thanks,
> > > > > Julie
> > > > >
> > > > > On Tue, Mar 10, 2015 at 9:38 AM, Kunke, Jessica via RT <
> > > > met_help at ucar.edu>
> > > > > wrote:
> > > > >
> > > > > >
> > > > > > Tue Mar 10 09:38:01 2015: Request 71051 was acted upon.
> > > > > > Transaction: Ticket created by jkunke at fas.harvard.edu
> > > > > > Queue: met_help
> > > > > > Subject: Question about ascii2nc and point_stat
> > > > > > Owner: Nobody
> > > > > > Requestors: jkunke at fas.harvard.edu
> > > > > > Status: new
> > > > > > Ticket <URL:
> > > https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=71051
> > > > >
> > > > > >
> > > > > >
> > > > > > Good morning,
> > > > > >
> > > > > > I have model forecasts at one-hour snapshots over an 8-
hour
> period,
> > > > and I
> > > > > > am looking to compare a number of stations at each of
those
> > times. I
> > > > > have
> > > > > > read the MET tutorial and the users' guide and I have a
couple
> > > > remaining
> > > > > > questions.
> > > > > >
> > > > > > I am using USCRN station data for my point observations,
and I
> can
> > > > > convert
> > > > > > them into a MET-formatted ascii file to use with ascii2nc
and
> then
> > > with
> > > > > > point_stat.
> > > > > >
> > > > > > 1. Can I have more times than desired in my MET-formatted
ascii
> > file
> > > > and
> > > > > > then use the configuration file or command line options
for
> > ascii2nc
> > > to
> > > > > > select only a certain time range from these observations
to
> include
> > > in
> > > > > the
> > > > > > netCDF file, or will it necessarily use all the
observations I
> have
> > > in
> > > > > the
> > > > > > ascii file I give to ascii2nc?
> > > > > >
> > > > > > 2. If I have point observations at multiple times in one
ascii or
> > > > netcdf
> > > > > > file that I give as input to point_stat, can point_stat
compare
> my
> > > > hourly
> > > > > > forecasts to this hourly data over the whole time period
in one
> > call?
> > > > Or
> > > > > > do I need to have only the observations I want to compare
to a
> > given
> > > > > > forecast (a one-time forecast) and just call point_stat
for each
> > > > forecast
> > > > > > time as a separate command, using a separate input file of
> > > > observations?
> > > > > >
> > > > > > Thank you,
> > > > > > Jess
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Julie Prestopnik
> > > > > National Center for Atmospheric Research
> > > > > Research Applications Laboratory
> > > > > Phone: 303.497.8399
> > > > > Email: jpresto at ucar.edu
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> > > --
> > > Julie Prestopnik
> > > National Center for Atmospheric Research
> > > Research Applications Laboratory
> > > Phone: 303.497.8399
> > > Email: jpresto at ucar.edu
> > >
> > >
> >
> >
>
>
------------------------------------------------
More information about the Met_help
mailing list