[Met_help] [rt.rap.ucar.edu #85605] History for question about MODE Analysis

John Halley Gotway via RT met_help at ucar.edu
Tue Jun 26 15:50:16 MDT 2018


----------------------------------------------------------------
  Initial Request
----------------------------------------------------------------

Dear MET help desk,

I have a question regarding the MODE Analysis tool. After applying the MODE
tool (for three models) I obtained a lot of statistics for the forecast and
observed objects. Therefore, I used MODE Analysis tool to summarize the
information obtained, like this:

mode_analysis -lookin $DIR -summary -column AREA -column AREA_THRESH
-column LENGTH -column WIDTH -column INTENSITY_90 -fcst -simple -fcst_rad 2
-fcst_thr ">=2.0" -dump_row fcst_model1
(same for the obs)

I'm interested to plot the total object count by forecast lead time for my
models and observation fields aggregated across summer period (respectively
the area objects).

>From MODE I obtained different forecasted and observed number of objects
for all three models. From what I understood the comparisons between
observed and forecast fields may differ for each model, even though I
compare the models with the same observation field. It's correct, right?

Therefore, my question is: how should I compare the models if I have
different number of observed objected?
(For ex: on my graph there are three curves with the models and another
three from the analysis observations ?) Is it correct to average the
observed objects identified by MODE for all three models and plot only the
mean observed identified objects?

Could you please give a hint about this matter?

Thank you,

Mirela


----------------------------------------------------------------
  Complete Ticket History
----------------------------------------------------------------

Subject: question about MODE Analysis
From: John Halley Gotway
Time: Tue Jun 19 15:12:47 2018

Hi Mirela,

I see you have a question about using MODE to compare the output of
three
different models.  If you really are comparing the three models to the
same
set of observations, using the exact same object definition criteria,
then
the number of observed objects should be identical across the three
runs.
By this I mean the number of simple observation objects should be the
same.  Depending on how you've configured the matching/merging logic,
the
number of matched observation objects and resulting cluster objects
may
differ across the 3 runs.

If you are not getting identical simple observation object counts,
please
let me know.  If you have questions, you could send MODE output for a
single case... i.e. 3 PostScript files from MODE, 3 ascii output files
(*_obj.txt) from MODE, and the MODE Config file you used.

If you would like, could configure MODE so that the forecast objects
have
no impact on the matching/merging of the observation objects by
setting:
   match_flag = MERGE_FCST;   // for matching allowing only additional
forecast merging

That prevents the location of the forecast objects from causing
additional
merging in the observation field.  Up to you.

Thanks,
John Halley Gotway


On Tue, Jun 19, 2018 at 2:01 PM Mirela Niculae via RT
<met_help at ucar.edu>
wrote:

>
> Tue Jun 19 14:01:18 2018: Request 85605 was acted upon.
> Transaction: Ticket created by mirela.niculae at gmail.com
>        Queue: met_help
>      Subject: question about MODE Analysis
>        Owner: Nobody
>   Requestors: mirela.niculae at gmail.com
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=85605 >
>
>
> Dear MET help desk,
>
> I have a question regarding the MODE Analysis tool. After applying
the MODE
> tool (for three models) I obtained a lot of statistics for the
forecast and
> observed objects. Therefore, I used MODE Analysis tool to summarize
the
> information obtained, like this:
>
> mode_analysis -lookin $DIR -summary -column AREA -column AREA_THRESH
> -column LENGTH -column WIDTH -column INTENSITY_90 -fcst -simple
-fcst_rad 2
> -fcst_thr ">=2.0" -dump_row fcst_model1
> (same for the obs)
>
> I'm interested to plot the total object count by forecast lead time
for my
> models and observation fields aggregated across summer period
(respectively
> the area objects).
>
> From MODE I obtained different forecasted and observed number of
objects
> for all three models. From what I understood the comparisons between
> observed and forecast fields may differ for each model, even though
I
> compare the models with the same observation field. It's correct,
right?
>
> Therefore, my question is: how should I compare the models if I have
> different number of observed objected?
> (For ex: on my graph there are three curves with the models and
another
> three from the analysis observations ?) Is it correct to average the
> observed objects identified by MODE for all three models and plot
only the
> mean observed identified objects?
>
> Could you please give a hint about this matter?
>
> Thank you,
>
> Mirela
>
>

------------------------------------------------
Subject: question about MODE Analysis
From: Mirela Niculae
Time: Tue Jun 26 15:45:42 2018

Hi John,

Thank you very much for the answer!

I found my mistake and I tested MODE again for the entire period.
After
this step, I applied MODE Analysis to summarize the information and
now,
the number of  simple observed objects is the same for all my models!

Mirela

2018-06-20 0:12 GMT+03:00 John Halley Gotway via RT
<met_help at ucar.edu>:

> Hi Mirela,
>
> I see you have a question about using MODE to compare the output of
three
> different models.  If you really are comparing the three models to
the same
> set of observations, using the exact same object definition
criteria, then
> the number of observed objects should be identical across the three
runs.
> By this I mean the number of simple observation objects should be
the
> same.  Depending on how you've configured the matching/merging
logic, the
> number of matched observation objects and resulting cluster objects
may
> differ across the 3 runs.
>
> If you are not getting identical simple observation object counts,
please
> let me know.  If you have questions, you could send MODE output for
a
> single case... i.e. 3 PostScript files from MODE, 3 ascii output
files
> (*_obj.txt) from MODE, and the MODE Config file you used.
>
> If you would like, could configure MODE so that the forecast objects
have
> no impact on the matching/merging of the observation objects by
setting:
>    match_flag = MERGE_FCST;   // for matching allowing only
additional
> forecast merging
>
> That prevents the location of the forecast objects from causing
additional
> merging in the observation field.  Up to you.
>
> Thanks,
> John Halley Gotway
>
>
> On Tue, Jun 19, 2018 at 2:01 PM Mirela Niculae via RT
<met_help at ucar.edu>
> wrote:
>
> >
> > Tue Jun 19 14:01:18 2018: Request 85605 was acted upon.
> > Transaction: Ticket created by mirela.niculae at gmail.com
> >        Queue: met_help
> >      Subject: question about MODE Analysis
> >        Owner: Nobody
> >   Requestors: mirela.niculae at gmail.com
> >       Status: new
> >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=85605 >
> >
> >
> > Dear MET help desk,
> >
> > I have a question regarding the MODE Analysis tool. After applying
the
> MODE
> > tool (for three models) I obtained a lot of statistics for the
forecast
> and
> > observed objects. Therefore, I used MODE Analysis tool to
summarize the
> > information obtained, like this:
> >
> > mode_analysis -lookin $DIR -summary -column AREA -column
AREA_THRESH
> > -column LENGTH -column WIDTH -column INTENSITY_90 -fcst -simple
> -fcst_rad 2
> > -fcst_thr ">=2.0" -dump_row fcst_model1
> > (same for the obs)
> >
> > I'm interested to plot the total object count by forecast lead
time for
> my
> > models and observation fields aggregated across summer period
> (respectively
> > the area objects).
> >
> > From MODE I obtained different forecasted and observed number of
objects
> > for all three models. From what I understood the comparisons
between
> > observed and forecast fields may differ for each model, even
though I
> > compare the models with the same observation field. It's correct,
right?
> >
> > Therefore, my question is: how should I compare the models if I
have
> > different number of observed objected?
> > (For ex: on my graph there are three curves with the models and
another
> > three from the analysis observations ?) Is it correct to average
the
> > observed objects identified by MODE for all three models and plot
only
> the
> > mean observed identified objects?
> >
> > Could you please give a hint about this matter?
> >
> > Thank you,
> >
> > Mirela
> >
> >
>
>

------------------------------------------------
Subject: question about MODE Analysis
From: John Halley Gotway
Time: Tue Jun 26 15:50:04 2018

Great, thanks for letting me know!

I'll go ahead and resolve this ticket.

John

On Tue, Jun 26, 2018 at 3:46 PM Mirela Niculae via RT
<met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=85605 >
>
> Hi John,
>
> Thank you very much for the answer!
>
> I found my mistake and I tested MODE again for the entire period.
After
> this step, I applied MODE Analysis to summarize the information and
now,
> the number of  simple observed objects is the same for all my
models!
>
> Mirela
>
> 2018-06-20 0:12 GMT+03:00 John Halley Gotway via RT
<met_help at ucar.edu>:
>
> > Hi Mirela,
> >
> > I see you have a question about using MODE to compare the output
of three
> > different models.  If you really are comparing the three models to
the
> same
> > set of observations, using the exact same object definition
criteria,
> then
> > the number of observed objects should be identical across the
three runs.
> > By this I mean the number of simple observation objects should be
the
> > same.  Depending on how you've configured the matching/merging
logic, the
> > number of matched observation objects and resulting cluster
objects may
> > differ across the 3 runs.
> >
> > If you are not getting identical simple observation object counts,
please
> > let me know.  If you have questions, you could send MODE output
for a
> > single case... i.e. 3 PostScript files from MODE, 3 ascii output
files
> > (*_obj.txt) from MODE, and the MODE Config file you used.
> >
> > If you would like, could configure MODE so that the forecast
objects have
> > no impact on the matching/merging of the observation objects by
setting:
> >    match_flag = MERGE_FCST;   // for matching allowing only
additional
> > forecast merging
> >
> > That prevents the location of the forecast objects from causing
> additional
> > merging in the observation field.  Up to you.
> >
> > Thanks,
> > John Halley Gotway
> >
> >
> > On Tue, Jun 19, 2018 at 2:01 PM Mirela Niculae via RT
<met_help at ucar.edu
> >
> > wrote:
> >
> > >
> > > Tue Jun 19 14:01:18 2018: Request 85605 was acted upon.
> > > Transaction: Ticket created by mirela.niculae at gmail.com
> > >        Queue: met_help
> > >      Subject: question about MODE Analysis
> > >        Owner: Nobody
> > >   Requestors: mirela.niculae at gmail.com
> > >       Status: new
> > >  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=85605
> >
> > >
> > >
> > > Dear MET help desk,
> > >
> > > I have a question regarding the MODE Analysis tool. After
applying the
> > MODE
> > > tool (for three models) I obtained a lot of statistics for the
forecast
> > and
> > > observed objects. Therefore, I used MODE Analysis tool to
summarize the
> > > information obtained, like this:
> > >
> > > mode_analysis -lookin $DIR -summary -column AREA -column
AREA_THRESH
> > > -column LENGTH -column WIDTH -column INTENSITY_90 -fcst -simple
> > -fcst_rad 2
> > > -fcst_thr ">=2.0" -dump_row fcst_model1
> > > (same for the obs)
> > >
> > > I'm interested to plot the total object count by forecast lead
time for
> > my
> > > models and observation fields aggregated across summer period
> > (respectively
> > > the area objects).
> > >
> > > From MODE I obtained different forecasted and observed number of
> objects
> > > for all three models. From what I understood the comparisons
between
> > > observed and forecast fields may differ for each model, even
though I
> > > compare the models with the same observation field. It's
correct,
> right?
> > >
> > > Therefore, my question is: how should I compare the models if I
have
> > > different number of observed objected?
> > > (For ex: on my graph there are three curves with the models and
another
> > > three from the analysis observations ?) Is it correct to average
the
> > > observed objects identified by MODE for all three models and
plot only
> > the
> > > mean observed identified objects?
> > >
> > > Could you please give a hint about this matter?
> > >
> > > Thank you,
> > >
> > > Mirela
> > >
> > >
> >
> >
>
>

------------------------------------------------


More information about the Met_help mailing list