[Met_help] [rt.rap.ucar.edu #94686] History for Follow up to today's MET user meeting: METviewer: using Y dependent (fcst) var to stratify/filter thresholds

Tue Mar 24 12:35:15 MDT 2020

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

Hi Tatiana, thank you for considering the option in the future for using the Y dependent (forecast) variable to stratify/filter all the thresholds in a database so only the threshold values relevant to the requested dependent variable would show up to be selected.  An example is the AWS database mv_ylin_metplus, which contains thresholds for 24h and 3h CTC stats and 24h/6h FSS.  Thresholds for the 24h CTC are 

gt0.254, gt2.54, gt6.35, gt12.7, gt19.05, gt25.4, gt38.1, gt50.8, gt76.2, gt101.6;  (threshold group 1)

For 3h CTC they are
gt0.254,gt0.508,gt1.27,gt2.54,gt6.35,gt12.7,gt19.05,gt25.4; (threshold group 2)

For FSS (6h or 24h) they are 
gt1, gt2, gt5, gt10, gt20, gt50 (threshold group 3)

In the current METviewer, when I choose the threshold as independent variable, all the above values show up in the selection window.  If there’s a way to use the Y dependent variable to sort the thresholds that’d be great, e.g. if the attribute stat selected is NBR_*, then only thresholds in "group 3" above shows up; for non-neighborhood CTC scores, “group 1” for Dependent variable of APCP_24 and “group 2” for APCP_03.  Does that make sense?

Thanks again -

Ying

--
Ying Lin
NCEP/EMC/Verification, Post-processing and Product Generation Branch
Ying.Lin at noaa.gov

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

Subject: Follow up to today's MET user meeting: METviewer: using Y dependent (fcst) var to stratify/filter thresholds
From: Tatiana Burek
Time: Tue Mar 24 08:58:24 2020

Hi Ying

When you select a fixed or independent variable METviewer finds and
displays the values filtered by Dependent (Forecast) Variable. For
example, if you choose APCP_03 - METviewer will display all
fcst_thresh values for this variable ONLY but for all line_types
(statistics).
This is done to keep the response time in the reasonable limits.

As I understand you correctly, you propose to add another level of
filtering to the database query - by the line type (statistic).
This will increase the time to find values and display them in the UI.
It can bring other complications such as:
-user did not select a statistic
-user selected more than one statistic
-user selected a statistic that can be calculated with or without
aggregation (for example, BASER)

I do think filtering by both - variable and line type(statistic) would
improve the user experience and we need to find a way to include it to
METviewer.
I will create a GitHub issue and discuss it with the group.

Tatiana

On Mon Mar 23 13:27:49 2020, ying.lin at noaa.gov wrote:
> Hi Tatiana, thank you for considering the option in the future for
> using the Y dependent (forecast) variable to stratify/filter all the
> thresholds in a database so only the threshold values relevant to
the
> requested dependent variable would show up to be selected.  An
example
> is the AWS database mv_ylin_metplus, which contains thresholds for
24h
> and 3h CTC stats and 24h/6h FSS.  Thresholds for the 24h CTC are
>
> gt0.254, gt2.54, gt6.35, gt12.7, gt19.05, gt25.4, gt38.1, gt50.8,
> gt76.2, gt101.6;  (threshold group 1)
>
> For 3h CTC they are
> gt0.254,gt0.508,gt1.27,gt2.54,gt6.35,gt12.7,gt19.05,gt25.4;
(threshold
> group 2)
>
> For FSS (6h or 24h) they are
> gt1, gt2, gt5, gt10, gt20, gt50 (threshold group 3)
>
> In the current METviewer, when I choose the threshold as independent
> variable, all the above values show up in the selection window.  If
> there’s a way to use the Y dependent variable to sort the thresholds
> that’d be great, e.g. if the attribute stat selected is NBR_*, then
> only thresholds in "group 3" above shows up; for non-neighborhood
CTC
> scores, “group 1” for Dependent variable of APCP_24 and “group 2”
for
> APCP_03.  Does that make sense?
>
> Thanks again -
>
> Ying
>
> --
> Ying Lin
> NCEP/EMC/Verification, Post-processing and Product Generation Branch
> Ying.Lin at noaa.gov

------------------------------------------------
Subject: Follow up to today's MET user meeting: METviewer: using Y dependent (fcst) var to stratify/filter thresholds
From: Ying Lin
Time: Tue Mar 24 12:24:35 2020

Hi Tatiana,

You’re right, I tried again with a ‘fresh’ METviewer, and it does
filter out fcst or obs thresholds not related to the current dependent
(forecast) variable, and as I move between different dependent
variables the available threshold values do get updated.    Earlier on
I was making interactive plots for one dependent variable, then hours
later used the same METviewer window to choose different dependent
variable in the same database, so the “mixed threshold values” I saw
was likely the result of a “stale window/buffer”.  Reloading the
database fixed the problem.

This is great.  Filtering by statistics type in the future might as
you said further improve the user experience, but this is now much
lower priority as far as I’m concerned.  In any case, please feel free
to close this ticket as you see fit.

Thank you for your reply and the explanation -

Ying

--
Ying Lin
NCEP/EMC/Verification, Post-processing and Product Generation Branch
Ying.Lin at noaa.gov

> On Mar 24, 2020, at 10:58 AM, Tatiana Burek via RT
<met_help at ucar.edu> wrote:
>
> Hi Ying
>
> When you select a fixed or independent variable METviewer finds and
displays the values filtered by Dependent (Forecast) Variable. For
example, if you choose APCP_03 - METviewer will display all
fcst_thresh values for this variable ONLY but for all line_types
(statistics).
> This is done to keep the response time in the reasonable limits.
>
> As I understand you correctly, you propose to add another level of
filtering to the database query - by the line type (statistic).
> This will increase the time to find values and display them in the
UI. It can bring other complications such as:
> -user did not select a statistic
> -user selected more than one statistic
> -user selected a statistic that can be calculated with or without
aggregation (for example, BASER)
>
> I do think filtering by both - variable and line type(statistic)
would improve the user experience and we need to find a way to include
it to METviewer.
> I will create a GitHub issue and discuss it with the group.
>
> Tatiana
>
> On Mon Mar 23 13:27:49 2020, ying.lin at noaa.gov wrote:
>> Hi Tatiana, thank you for considering the option in the future for
>> using the Y dependent (forecast) variable to stratify/filter all
the
>> thresholds in a database so only the threshold values relevant to
the
>> requested dependent variable would show up to be selected.  An
example
>> is the AWS database mv_ylin_metplus, which contains thresholds for
24h
>> and 3h CTC stats and 24h/6h FSS.  Thresholds for the 24h CTC are
>>
>> gt0.254, gt2.54, gt6.35, gt12.7, gt19.05, gt25.4, gt38.1, gt50.8,
>> gt76.2, gt101.6;  (threshold group 1)
>>
>> For 3h CTC they are
>> gt0.254,gt0.508,gt1.27,gt2.54,gt6.35,gt12.7,gt19.05,gt25.4;
(threshold
>> group 2)
>>
>> For FSS (6h or 24h) they are
>> gt1, gt2, gt5, gt10, gt20, gt50 (threshold group 3)
>>
>> In the current METviewer, when I choose the threshold as
independent
>> variable, all the above values show up in the selection window.  If
>> there’s a way to use the Y dependent variable to sort the
thresholds
>> that’d be great, e.g. if the attribute stat selected is NBR_*, then
>> only thresholds in "group 3" above shows up; for non-neighborhood
CTC
>> scores, “group 1” for Dependent variable of APCP_24 and “group 2”
for
>> APCP_03.  Does that make sense?
>>
>> Thanks again -
>>
>> Ying
>>
>> --
>> Ying Lin
>> NCEP/EMC/Verification, Post-processing and Product Generation
Branch
>> Ying.Lin at noaa.gov
>
>
>

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