[Met_help] [rt.rap.ucar.edu #80525] History for series-analysis
John Halley Gotway via RT
met_help at ucar.edu
Mon May 22 09:31:43 MDT 2017
----------------------------------------------------------------
Initial Request
----------------------------------------------------------------
I am having some problem with "series analysis" and not sure what is causing it. I am trying to verify model NGAC against MODIS satellite for month of January. I averaged NGAC forecast data for each day (using CDO) and using it against daily satellite gridded files. When I am trying to run series analysis, it shows the following error :
WARNING: MetGrib2DataFile::data_plane() - No matching record found for 'AOTK/A0'
WARNING:
WARNING:
WARNING: get_series_entry() -> No match found for AOTK/A0 in file: /scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/MET6/Regrid_NGACv2_00h18_Modf/20170128/ngac.t00z_a2df20170128.grib2
But when I individually look at those grib2 files, using "plot_data_plane" it reads them correctly and generates figure. My files are in Theia at :
/scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/MET6
and was using "series_stat_Jan17_MODIS_2.sh" to run series analysis. Any help on this is appreciated.
Thanks.
----------------------------------------------------------------
Complete Ticket History
----------------------------------------------------------------
Subject: series-analysis
From: John Halley Gotway
Time: Thu May 18 14:01:02 2017
Partha,
I see that you've having some difficulty running the Series-Analysis
tool to compare NGAC forecasts to MODIS.
I took a look at your config and think I found the problem. Please
try removing this setting:
GRIB2_process=62000;
You were telling Series-Analysis to only use GRIB2 records where the
process value is "62000". Using wgrib2, I can see the process is set
to a value of 2 in this data:
wgrib2
/scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/MET6/Regrid_NGACv2_00h18_Modf/20170101/ngac.t00z_a2df20170101.grib2
-process
1:0:code table 4.3=2 fcst
2:129803:code table 4.3=2 fcst
3:259606:code table 4.3=2 fcst
I'm guessing that setting was just a holdover from a previous run.
Hope that does the trick.
Thanks,
John
------------------------------------------------
Subject: series-analysis
From: Partha Bhattacharjee - NOAA Affiliate
Time: Thu May 18 20:09:50 2017
Thanks John. It did run without showing previous error. But I am
getting
nearly empty fields when I looked into Bias, RMSE etc fields. Attached
is a
figure of RMSE generated by this run indicate no matching pairs
between
model and satellite (it has missing values between swaths).
Thanks.
Partha
On Thu, May 18, 2017 at 4:01 PM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:
> Partha,
>
> I see that you've having some difficulty running the Series-Analysis
tool
> to compare NGAC forecasts to MODIS.
>
> I took a look at your config and think I found the problem. Please
try
> removing this setting:
>
> GRIB2_process=62000;
>
> You were telling Series-Analysis to only use GRIB2 records where the
> process value is "62000". Using wgrib2, I can see the process is
set to a
> value of 2 in this data:
>
> wgrib2 /scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/
> MET6/Regrid_NGACv2_00h18_Modf/20170101/ngac.t00z_a2df20170101.grib2
> -process
> 1:0:code table 4.3=2 fcst
> 2:129803:code table 4.3=2 fcst
> 3:259606:code table 4.3=2 fcst
>
> I'm guessing that setting was just a holdover from a previous run.
Hope
> that does the trick.
>
> Thanks,
> John
>
--
Partha S. Bhattacharjee, Ph.D.
I.M. Systems Group
Environmental Modeling Centre
NCWCP
5830 University Research Court.
College Park, MD 20740
------------------------------------------------
Subject: series-analysis
From: John Halley Gotway
Time: Fri May 19 10:08:49 2017
Partha,
I logged on to theia and looked at your scripts. I wanted to get a
better
understanding of what the MODIS data looks like, so I ran the
following
commands to generate an animated gif that I've attached to this
mesage:
module use /scratch4/BMC/dtc/MET/modulefiles
module load met/6.0
rm -rf plot; mkdir plot
foreach nc ( `ls -1
/scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/MET6/MODIS_daily/g4.subsetted.MYD08_D3_6_AOD_550_Dark_Target_Deep_Blue_Combined_Mean.*`
)
set ps = `basename $nc | sed 's/.nc/.ps/g'`
plot_data_plane $nc plot/$ps
'name="MYD08_D3_6_AOD_550_Dark_Target_Deep_Blue_Combined_Mean";
level="(0,*,*)";' -plot_range 0 3
end
convert -rotate 90 -background white -dispose Background -delay 100
plot/*.ps plot/modis.gif
I see the strips of missing data between the swaths that you've
mentioned.
So we just need to give some more thought to how we want Series-
Analysis to
handle missing data. This is handled by the config file setting named
"vld_thresh". You set that to a value between 0 and 1 to indicate the
ratio of pairs that must contain valid data for it to be used. You
currently have it set to 1.0:
vld_thresh = 1.0;
Since your series is length 31, you're telling Series-Analysis to only
compute statistics at grid points where all 31 pairs contain valid
data.
With all that missing data between the swaths, that seldom happens.
And
that's why you're getting so few points in the output. It's up to you
to
decide how you want to set that ratio, but you might try 0.5. Or even
just
set it to 0.01 so that you get statistics even if there's only 1
matched
pair.
I reran with vld_thresh = 0.01 and attached a plot showing the number
of
matched pairs at each grid point:
plot_data_plane output_series_201701_Modis/series_fh018_201701_3.nc
total.ps
'name="series_cnt_TOTAL"; level="(*,*)";' -plot_range 0 31
You can see you now have a lot more data points!
I also noticed in your config file that you're requesting categorical
stats:
fho = [ "F_RATE", "O_RATE", "TOTAL" ];
cts = [ "CSI", "GSS", "FAR" ];
However, you haven't defined any categorical thresholds to be applied
when
computing them:
cat_thresh = [ NA ];
Computing categorical stats is fine, just be sure to pick one or more
meaningful thresholds.
Thanks,
John
On Thu, May 18, 2017 at 8:09 PM, Partha Bhattacharjee - NOAA Affiliate
via
RT <met_help at ucar.edu> wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=80525 >
>
> Thanks John. It did run without showing previous error. But I am
getting
> nearly empty fields when I looked into Bias, RMSE etc fields.
Attached is a
> figure of RMSE generated by this run indicate no matching pairs
between
> model and satellite (it has missing values between swaths).
>
> Thanks.
> Partha
>
> On Thu, May 18, 2017 at 4:01 PM, John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
> > Partha,
> >
> > I see that you've having some difficulty running the Series-
Analysis tool
> > to compare NGAC forecasts to MODIS.
> >
> > I took a look at your config and think I found the problem.
Please try
> > removing this setting:
> >
> > GRIB2_process=62000;
> >
> > You were telling Series-Analysis to only use GRIB2 records where
the
> > process value is "62000". Using wgrib2, I can see the process is
set to
> a
> > value of 2 in this data:
> >
> > wgrib2 /scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/
> >
MET6/Regrid_NGACv2_00h18_Modf/20170101/ngac.t00z_a2df20170101.grib2
> > -process
> > 1:0:code table 4.3=2 fcst
> > 2:129803:code table 4.3=2 fcst
> > 3:259606:code table 4.3=2 fcst
> >
> > I'm guessing that setting was just a holdover from a previous run.
Hope
> > that does the trick.
> >
> > Thanks,
> > John
> >
>
>
>
> --
> Partha S. Bhattacharjee, Ph.D.
> I.M. Systems Group
> Environmental Modeling Centre
> NCWCP
> 5830 University Research Court.
> College Park, MD 20740
>
>
------------------------------------------------
Subject: series-analysis
From: Partha Bhattacharjee - NOAA Affiliate
Time: Fri May 19 16:40:57 2017
Many thanks John for your suggestions. I tried different valid
thresholds
and will use that when compare with satellite data that have data
gaps.
Partha
On Fri, May 19, 2017 at 12:08 PM, John Halley Gotway via RT <
met_help at ucar.edu> wrote:
> Partha,
>
> I logged on to theia and looked at your scripts. I wanted to get a
better
> understanding of what the MODIS data looks like, so I ran the
following
> commands to generate an animated gif that I've attached to this
mesage:
>
> module use /scratch4/BMC/dtc/MET/modulefiles
> module load met/6.0
>
> rm -rf plot; mkdir plot
> foreach nc ( `ls -1
> /scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/
> MET6/MODIS_daily/g4.subsetted.MYD08_D3_6_AOD_550_Dark_
> Target_Deep_Blue_Combined_Mean.*`
> )
> set ps = `basename $nc | sed 's/.nc/.ps/g'`
> plot_data_plane $nc plot/$ps
> 'name="MYD08_D3_6_AOD_550_Dark_Target_Deep_Blue_Combined_Mean";
> level="(0,*,*)";' -plot_range 0 3
> end
> convert -rotate 90 -background white -dispose Background -delay 100
> plot/*.ps plot/modis.gif
>
> I see the strips of missing data between the swaths that you've
mentioned.
> So we just need to give some more thought to how we want Series-
Analysis to
> handle missing data. This is handled by the config file setting
named
> "vld_thresh". You set that to a value between 0 and 1 to indicate
the
> ratio of pairs that must contain valid data for it to be used. You
> currently have it set to 1.0:
> vld_thresh = 1.0;
>
> Since your series is length 31, you're telling Series-Analysis to
only
> compute statistics at grid points where all 31 pairs contain valid
data.
> With all that missing data between the swaths, that seldom happens.
And
> that's why you're getting so few points in the output. It's up to
you to
> decide how you want to set that ratio, but you might try 0.5. Or
even just
> set it to 0.01 so that you get statistics even if there's only 1
matched
> pair.
>
> I reran with vld_thresh = 0.01 and attached a plot showing the
number of
> matched pairs at each grid point:
>
> plot_data_plane output_series_201701_Modis/series_fh018_201701_3.nc
> total.ps
> 'name="series_cnt_TOTAL"; level="(*,*)";' -plot_range 0 31
>
> You can see you now have a lot more data points!
>
> I also noticed in your config file that you're requesting
categorical
> stats:
> fho = [ "F_RATE", "O_RATE", "TOTAL" ];
> cts = [ "CSI", "GSS", "FAR" ];
>
> However, you haven't defined any categorical thresholds to be
applied when
> computing them:
> cat_thresh = [ NA ];
>
> Computing categorical stats is fine, just be sure to pick one or
more
> meaningful thresholds.
>
> Thanks,
> John
>
> On Thu, May 18, 2017 at 8:09 PM, Partha Bhattacharjee - NOAA
Affiliate via
> RT <met_help at ucar.edu> wrote:
>
> >
> > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=80525 >
> >
> > Thanks John. It did run without showing previous error. But I am
getting
> > nearly empty fields when I looked into Bias, RMSE etc fields.
Attached
> is a
> > figure of RMSE generated by this run indicate no matching pairs
between
> > model and satellite (it has missing values between swaths).
> >
> > Thanks.
> > Partha
> >
> > On Thu, May 18, 2017 at 4:01 PM, John Halley Gotway via RT <
> > met_help at ucar.edu> wrote:
> >
> > > Partha,
> > >
> > > I see that you've having some difficulty running the Series-
Analysis
> tool
> > > to compare NGAC forecasts to MODIS.
> > >
> > > I took a look at your config and think I found the problem.
Please try
> > > removing this setting:
> > >
> > > GRIB2_process=62000;
> > >
> > > You were telling Series-Analysis to only use GRIB2 records where
the
> > > process value is "62000". Using wgrib2, I can see the process
is set
> to
> > a
> > > value of 2 in this data:
> > >
> > > wgrib2 /scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/
> > >
MET6/Regrid_NGACv2_00h18_Modf/20170101/ngac.t00z_a2df20170101.grib2
> > > -process
> > > 1:0:code table 4.3=2 fcst
> > > 2:129803:code table 4.3=2 fcst
> > > 3:259606:code table 4.3=2 fcst
> > >
> > > I'm guessing that setting was just a holdover from a previous
run.
> Hope
> > > that does the trick.
> > >
> > > Thanks,
> > > John
> > >
> >
> >
> >
> > --
> > Partha S. Bhattacharjee, Ph.D.
> > I.M. Systems Group
> > Environmental Modeling Centre
> > NCWCP
> > 5830 University Research Court.
> > College Park, MD 20740
> >
> >
>
>
--
Partha S. Bhattacharjee, Ph.D.
I.M. Systems Group
Environmental Modeling Centre
NCWCP
5830 University Research Court.
College Park, MD 20740
------------------------------------------------
Subject: series-analysis
From: John Halley Gotway
Time: Mon May 22 09:31:05 2017
Great to hear. Glad that did the trick.
I'll go ahead and resolve this ticket.
Thanks,
John
On Fri, May 19, 2017 at 4:40 PM, Partha Bhattacharjee - NOAA Affiliate
via
RT <met_help at ucar.edu> wrote:
>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=80525 >
>
> Many thanks John for your suggestions. I tried different valid
thresholds
> and will use that when compare with satellite data that have data
gaps.
>
>
> Partha
>
> On Fri, May 19, 2017 at 12:08 PM, John Halley Gotway via RT <
> met_help at ucar.edu> wrote:
>
> > Partha,
> >
> > I logged on to theia and looked at your scripts. I wanted to get
a
> better
> > understanding of what the MODIS data looks like, so I ran the
following
> > commands to generate an animated gif that I've attached to this
mesage:
> >
> > module use /scratch4/BMC/dtc/MET/modulefiles
> > module load met/6.0
> >
> > rm -rf plot; mkdir plot
> > foreach nc ( `ls -1
> > /scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/
> > MET6/MODIS_daily/g4.subsetted.MYD08_D3_6_AOD_550_Dark_
> > Target_Deep_Blue_Combined_Mean.*`
> > )
> > set ps = `basename $nc | sed 's/.nc/.ps/g'`
> > plot_data_plane $nc plot/$ps
> > 'name="MYD08_D3_6_AOD_550_Dark_Target_Deep_Blue_Combined_Mean";
> > level="(0,*,*)";' -plot_range 0 3
> > end
> > convert -rotate 90 -background white -dispose Background -delay
100
> > plot/*.ps plot/modis.gif
> >
> > I see the strips of missing data between the swaths that you've
> mentioned.
> > So we just need to give some more thought to how we want Series-
Analysis
> to
> > handle missing data. This is handled by the config file setting
named
> > "vld_thresh". You set that to a value between 0 and 1 to indicate
the
> > ratio of pairs that must contain valid data for it to be used.
You
> > currently have it set to 1.0:
> > vld_thresh = 1.0;
> >
> > Since your series is length 31, you're telling Series-Analysis to
only
> > compute statistics at grid points where all 31 pairs contain valid
data.
> > With all that missing data between the swaths, that seldom
happens. And
> > that's why you're getting so few points in the output. It's up to
you to
> > decide how you want to set that ratio, but you might try 0.5. Or
even
> just
> > set it to 0.01 so that you get statistics even if there's only 1
matched
> > pair.
> >
> > I reran with vld_thresh = 0.01 and attached a plot showing the
number of
> > matched pairs at each grid point:
> >
> > plot_data_plane
output_series_201701_Modis/series_fh018_201701_3.nc
> > total.ps
> > 'name="series_cnt_TOTAL"; level="(*,*)";' -plot_range 0 31
> >
> > You can see you now have a lot more data points!
> >
> > I also noticed in your config file that you're requesting
categorical
> > stats:
> > fho = [ "F_RATE", "O_RATE", "TOTAL" ];
> > cts = [ "CSI", "GSS", "FAR" ];
> >
> > However, you haven't defined any categorical thresholds to be
applied
> when
> > computing them:
> > cat_thresh = [ NA ];
> >
> > Computing categorical stats is fine, just be sure to pick one or
more
> > meaningful thresholds.
> >
> > Thanks,
> > John
> >
> > On Thu, May 18, 2017 at 8:09 PM, Partha Bhattacharjee - NOAA
Affiliate
> via
> > RT <met_help at ucar.edu> wrote:
> >
> > >
> > > <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=80525 >
> > >
> > > Thanks John. It did run without showing previous error. But I am
> getting
> > > nearly empty fields when I looked into Bias, RMSE etc fields.
Attached
> > is a
> > > figure of RMSE generated by this run indicate no matching pairs
between
> > > model and satellite (it has missing values between swaths).
> > >
> > > Thanks.
> > > Partha
> > >
> > > On Thu, May 18, 2017 at 4:01 PM, John Halley Gotway via RT <
> > > met_help at ucar.edu> wrote:
> > >
> > > > Partha,
> > > >
> > > > I see that you've having some difficulty running the Series-
Analysis
> > tool
> > > > to compare NGAC forecasts to MODIS.
> > > >
> > > > I took a look at your config and think I found the problem.
Please
> try
> > > > removing this setting:
> > > >
> > > > GRIB2_process=62000;
> > > >
> > > > You were telling Series-Analysis to only use GRIB2 records
where the
> > > > process value is "62000". Using wgrib2, I can see the process
is set
> > to
> > > a
> > > > value of 2 in this data:
> > > >
> > > > wgrib2 /scratch4/NCEPDEV/global/noscrub/Partha.Bhattacharjee/
> > > >
MET6/Regrid_NGACv2_00h18_Modf/20170101/ngac.t00z_a2df20170101.grib2
> > > > -process
> > > > 1:0:code table 4.3=2 fcst
> > > > 2:129803:code table 4.3=2 fcst
> > > > 3:259606:code table 4.3=2 fcst
> > > >
> > > > I'm guessing that setting was just a holdover from a previous
run.
> > Hope
> > > > that does the trick.
> > > >
> > > > Thanks,
> > > > John
> > > >
> > >
> > >
> > >
> > > --
> > > Partha S. Bhattacharjee, Ph.D.
> > > I.M. Systems Group
> > > Environmental Modeling Centre
> > > NCWCP
> > > 5830 University Research Court.
> > > College Park, MD 20740
> > >
> > >
> >
> >
>
>
> --
> Partha S. Bhattacharjee, Ph.D.
> I.M. Systems Group
> Environmental Modeling Centre
> NCWCP
> 5830 University Research Court.
> College Park, MD 20740
>
>
------------------------------------------------
More information about the Met_help
mailing list