Forums

Back

RE: GEFS Raw vs BC

JS
Jack Settelmaier, modified 1 Year ago.

GEFS Raw vs BC

Youngling Posts: 67 Join Date: 11/2/12 Recent Posts
Happened to notice this afternoon that the plot line (dark blue) for the BC GEFS Average looks to be basically a max of the members plotted here, whereas the Raw GEFS looks more correctly like the Average of the plotted members.  

Is that due to a time-lagged approach to bias-correcting, at that far out forecast time step or something?  Also, with the GEFS membership now up to ~32, is it by choice just 20 are plotted here, or for readability?




--
Jack Settelmaier
Digital Techniques Meteorologist
NOAA/NWS, Southern Region HQ
Fort Worth, TX
DM
David T Miller, modified 1 Year ago.

RE: GEFS Raw vs BC

Youngling Posts: 10 Join Date: 11/7/13 Recent Posts
Jack,

Hey thanks very much for pointing this out to us!  It's good to know folks are looking at our displays and keeping us on our toes when there seem to be issues ;)

In investigating your questions, it turns out that the two issues are related.  Getting to your second question first, we discovered you're correct in that there are now 30 BC QPF06 perturbation members along with two controls, the average, and the standard deviation in the GRIB files.  The number of GEFS raw and BC members were increased when NAEFS v7 became operational on Dec 5.  As the average and member values are in different files for the raw GEFS QPF GRIB values, our data processing scripts adjusted for the increased number of members in those files.  However, the GEFS BC QPF GRIB files contain the  member and average values in one file, which is different from the raw as they're in separate files.  So the scripts that process the GEFS BC QPF GRIB files didn't adjust to the increased number of members as 20 were still expected.  Since the processing scripts for the raw GEFS member values adjusted with the increase in members without issue, I think we were just unaware that the processing scripts for the BC GEFS QPF data had not adjusted.  Note that, and somewhat in our defense, NCEP NCO online documentation still lists these GEFS BC QPF GRIB files as only having 20 members along with the controls, average/mean, and standard deviation.

So, to answer your first question, our data processing script hasn't been processing the correct GRIB message in the GEFS BC QPF GRIB v7 files due to the increased number of members in the files as the average/mean value is now in a different GRIB message number.

These issues should be fixed with today's GEFS 12Z run, which should start processing around 18-19Z.  

Again, thanks for letting us know!

Dave M

On Mon, Apr 22, 2024 at 5:57 PM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Happened to notice this afternoon that the plot line (dark blue) for the BC GEFS Average looks to be basically a max of the members plotted here, whereas the Raw GEFS looks more correctly like the Average of the plotted members.  

Is that due to a time-lagged approach to bias-correcting, at that far out forecast time step or something?  Also, with the GEFS membership now up to ~32, is it by choice just 20 are plotted here, or for readability?




--
Jack Settelmaier
Digital Techniques Meteorologist
NOAA/NWS, Southern Region HQ
Fort Worth, TX

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36670061VLab.Notifications@noaa.gov


--

David T. Miller

Research Associate, Programmer/Analyst

National Weather Service, Meteorological Development Laboratory

Cooperative Institute for Research in the Atmosphere/Colorado State University

1325 East West Highway |  Silver Spring, MD 20910  |  Phone (GV): (202) 936-5659  |  Fax: 301-713-9316

JS
Jack Settelmaier, modified 1 Year ago.

RE: GEFS Raw vs BC

Youngling Posts: 67 Join Date: 11/2/12 Recent Posts
Thanks Dave, for the detailed response.    I look forward to seeing the changes, as I peruse the site for the next rainfall event.

Hey, while I have your ear, do you do your processing of these GEFS files on AWS?  There was a question that came in via NODD circles (see text below), asking about the ability to subset S3 data stores (by variable, and by bounded area) so that one needn't download TBs of data across an archive just to get one variable in one small area.  You have any sample code that may get a user closer to that?

"I am encountering a problem where I need five months of just one "b" variable (SST) in a small bounded area from GEFS v11.0 archived data. I cannot find any way to do this without downloading an estimated 18 TB of data and then processing it which seems very wasteful of computational resources given the final subset size. Is there any way to directly download a subset of the "b" variable files?"

On Tue, Apr 23, 2024 at 10:03 AM David T Miller <VLab.Notifications@noaa.gov> wrote:
Jack,

Hey thanks very much for pointing this out to us!  It's good to know folks are looking at our displays and keeping us on our toes when there seem to be issues ;)

In investigating your questions, it turns out that the two issues are related.  Getting to your second question first, we discovered you're correct in that there are now 30 BC QPF06 perturbation members along with two controls, the average, and the standard deviation in the GRIB files.  The number of GEFS raw and BC members were increased when NAEFS v7 became operational on Dec 5.  As the average and member values are in different files for the raw GEFS QPF GRIB values, our data processing scripts adjusted for the increased number of members in those files.  However, the GEFS BC QPF GRIB files contain the  member and average values in one file, which is different from the raw as they're in separate files.  So the scripts that process the GEFS BC QPF GRIB files didn't adjust to the increased number of members as 20 were still expected.  Since the processing scripts for the raw GEFS member values adjusted with the increase in members without issue, I think we were just unaware that the processing scripts for the BC GEFS QPF data had not adjusted.  Note that, and somewhat in our defense, NCEP NCO online documentation still lists these GEFS BC QPF GRIB files as only having 20 members along with the controls, average/mean, and standard deviation.

So, to answer your first question, our data processing script hasn't been processing the correct GRIB message in the GEFS BC QPF GRIB v7 files due to the increased number of members in the files as the average/mean value is now in a different GRIB message number.

These issues should be fixed with today's GEFS 12Z run, which should start processing around 18-19Z.  

Again, thanks for letting us know!

Dave M

On Mon, Apr 22, 2024 at 5:57 PM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Happened to notice this afternoon that the plot line (dark blue) for the BC GEFS Average looks to be basically a max of the members plotted here, whereas the Raw GEFS looks more correctly like the Average of the plotted members.  

Is that due to a time-lagged approach to bias-correcting, at that far out forecast time step or something?  Also, with the GEFS membership now up to ~32, is it by choice just 20 are plotted here, or for readability?




--
Jack Settelmaier
Digital Techniques Meteorologist
NOAA/NWS, Southern Region HQ
Fort Worth, TX

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36670061VLab.Notifications@noaa.gov


--

David T. Miller

Research Associate, Programmer/Analyst

National Weather Service, Meteorological Development Laboratory

Cooperative Institute for Research in the Atmosphere/Colorado State University

1325 East West Highway |  Silver Spring, MD 20910  |  Phone (GV): (202) 936-5659  |  Fax: 301-713-9316


--
David T Miller Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36692620VLab.Notifications@noaa.gov
DS
Dana Strom, modified 1 Year ago.

RE: GEFS Raw vs BC

Youngling Posts: 87 Join Date: 12/8/14 Recent Posts
Jack,

We do not do our GEFS processing utilizing the NODD archive.

Unfortunately, there is no good way around downloading the full grid. The data stored in NODD is in GRIB format, and there's not a great way to split out singular records or geographic regions without pulling the full file and doing the band and domain processing locally. We have no quick solution for this.

Dana

On Tue, Apr 23, 2024 at 11:23 AM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Thanks Dave, for the detailed response.    I look forward to seeing the changes, as I peruse the site for the next rainfall event.

Hey, while I have your ear, do you do your processing of these GEFS files on AWS?  There was a question that came in via NODD circles (see text below), asking about the ability to subset S3 data stores (by variable, and by bounded area) so that one needn't download TBs of data across an archive just to get one variable in one small area.  You have any sample code that may get a user closer to that?

"I am encountering a problem where I need five months of just one "b" variable (SST) in a small bounded area from GEFS v11.0 archived data. I cannot find any way to do this without downloading an estimated 18 TB of data and then processing it which seems very wasteful of computational resources given the final subset size. Is there any way to directly download a subset of the "b" variable files?"

On Tue, Apr 23, 2024 at 10:03 AM David T Miller <VLab.Notifications@noaa.gov> wrote:
Jack,

Hey thanks very much for pointing this out to us!  It's good to know folks are looking at our displays and keeping us on our toes when there seem to be issues ;)

In investigating your questions, it turns out that the two issues are related.  Getting to your second question first, we discovered you're correct in that there are now 30 BC QPF06 perturbation members along with two controls, the average, and the standard deviation in the GRIB files.  The number of GEFS raw and BC members were increased when NAEFS v7 became operational on Dec 5.  As the average and member values are in different files for the raw GEFS QPF GRIB values, our data processing scripts adjusted for the increased number of members in those files.  However, the GEFS BC QPF GRIB files contain the  member and average values in one file, which is different from the raw as they're in separate files.  So the scripts that process the GEFS BC QPF GRIB files didn't adjust to the increased number of members as 20 were still expected.  Since the processing scripts for the raw GEFS member values adjusted with the increase in members without issue, I think we were just unaware that the processing scripts for the BC GEFS QPF data had not adjusted.  Note that, and somewhat in our defense, NCEP NCO online documentation still lists these GEFS BC QPF GRIB files as only having 20 members along with the controls, average/mean, and standard deviation.

So, to answer your first question, our data processing script hasn't been processing the correct GRIB message in the GEFS BC QPF GRIB v7 files due to the increased number of members in the files as the average/mean value is now in a different GRIB message number.

These issues should be fixed with today's GEFS 12Z run, which should start processing around 18-19Z.  

Again, thanks for letting us know!

Dave M

On Mon, Apr 22, 2024 at 5:57 PM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Happened to notice this afternoon that the plot line (dark blue) for the BC GEFS Average looks to be basically a max of the members plotted here, whereas the Raw GEFS looks more correctly like the Average of the plotted members.  

Is that due to a time-lagged approach to bias-correcting, at that far out forecast time step or something?  Also, with the GEFS membership now up to ~32, is it by choice just 20 are plotted here, or for readability?




--
Jack Settelmaier
Digital Techniques Meteorologist
NOAA/NWS, Southern Region HQ
Fort Worth, TX

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36670061VLab.Notifications@noaa.gov


--

David T. Miller

Research Associate, Programmer/Analyst

National Weather Service, Meteorological Development Laboratory

Cooperative Institute for Research in the Atmosphere/Colorado State University

1325 East West Highway |  Silver Spring, MD 20910  |  Phone (GV): (202) 936-5659  |  Fax: 301-713-9316


--
David T Miller Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36692620VLab.Notifications@noaa.gov

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36693148VLab.Notifications@noaa.gov


--
Dana Strom
Visualization Task Lead
Digital Forecast Services Division
Meteorological Development Lab
NOAA/National Weather Service

301-427-9451

JS
Jack Settelmaier, modified 1 Year ago.

RE: GEFS Raw vs BC

Youngling Posts: 67 Join Date: 11/2/12 Recent Posts
Thanks Dana.  I can imagine there is nothing easy when it comes to GRIB, ;), but I thought I'd ask to see if any snippets exist out there, in the subsetting space. I see the Herbie code make such subsetting claims, but have not gone as far as trying it myself to see if it might be something worthy of promoting for subsetting NODD stores of GRIB files.

On Tue, Apr 23, 2024 at 10:44 AM Dana Strom <VLab.Notifications@noaa.gov> wrote:
Jack,

We do not do our GEFS processing utilizing the NODD archive.

Unfortunately, there is no good way around downloading the full grid. The data stored in NODD is in GRIB format, and there's not a great way to split out singular records or geographic regions without pulling the full file and doing the band and domain processing locally. We have no quick solution for this.

Dana

On Tue, Apr 23, 2024 at 11:23 AM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Thanks Dave, for the detailed response.    I look forward to seeing the changes, as I peruse the site for the next rainfall event.

Hey, while I have your ear, do you do your processing of these GEFS files on AWS?  There was a question that came in via NODD circles (see text below), asking about the ability to subset S3 data stores (by variable, and by bounded area) so that one needn't download TBs of data across an archive just to get one variable in one small area.  You have any sample code that may get a user closer to that?

"I am encountering a problem where I need five months of just one "b" variable (SST) in a small bounded area from GEFS v11.0 archived data. I cannot find any way to do this without downloading an estimated 18 TB of data and then processing it which seems very wasteful of computational resources given the final subset size. Is there any way to directly download a subset of the "b" variable files?"

On Tue, Apr 23, 2024 at 10:03 AM David T Miller <VLab.Notifications@noaa.gov> wrote:
Jack,

Hey thanks very much for pointing this out to us!  It's good to know folks are looking at our displays and keeping us on our toes when there seem to be issues ;)

In investigating your questions, it turns out that the two issues are related.  Getting to your second question first, we discovered you're correct in that there are now 30 BC QPF06 perturbation members along with two controls, the average, and the standard deviation in the GRIB files.  The number of GEFS raw and BC members were increased when NAEFS v7 became operational on Dec 5.  As the average and member values are in different files for the raw GEFS QPF GRIB values, our data processing scripts adjusted for the increased number of members in those files.  However, the GEFS BC QPF GRIB files contain the  member and average values in one file, which is different from the raw as they're in separate files.  So the scripts that process the GEFS BC QPF GRIB files didn't adjust to the increased number of members as 20 were still expected.  Since the processing scripts for the raw GEFS member values adjusted with the increase in members without issue, I think we were just unaware that the processing scripts for the BC GEFS QPF data had not adjusted.  Note that, and somewhat in our defense, NCEP NCO online documentation still lists these GEFS BC QPF GRIB files as only having 20 members along with the controls, average/mean, and standard deviation.

So, to answer your first question, our data processing script hasn't been processing the correct GRIB message in the GEFS BC QPF GRIB v7 files due to the increased number of members in the files as the average/mean value is now in a different GRIB message number.

These issues should be fixed with today's GEFS 12Z run, which should start processing around 18-19Z.  

Again, thanks for letting us know!

Dave M

On Mon, Apr 22, 2024 at 5:57 PM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Happened to notice this afternoon that the plot line (dark blue) for the BC GEFS Average looks to be basically a max of the members plotted here, whereas the Raw GEFS looks more correctly like the Average of the plotted members.  

Is that due to a time-lagged approach to bias-correcting, at that far out forecast time step or something?  Also, with the GEFS membership now up to ~32, is it by choice just 20 are plotted here, or for readability?




--
Jack Settelmaier
Digital Techniques Meteorologist
NOAA/NWS, Southern Region HQ
Fort Worth, TX

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36670061VLab.Notifications@noaa.gov


--

David T. Miller

Research Associate, Programmer/Analyst

National Weather Service, Meteorological Development Laboratory

Cooperative Institute for Research in the Atmosphere/Colorado State University

1325 East West Highway |  Silver Spring, MD 20910  |  Phone (GV): (202) 936-5659  |  Fax: 301-713-9316


--
David T Miller Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36692620VLab.Notifications@noaa.gov

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36693148VLab.Notifications@noaa.gov


--
Dana Strom
Visualization Task Lead
Digital Forecast Services Division
Meteorological Development Lab
NOAA/National Weather Service

301-427-9451


--
Dana Strom Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36693203VLab.Notifications@noaa.gov
JS
Jack Settelmaier, modified 1 Year ago.

RE: GEFS Raw vs BC

Youngling Posts: 67 Join Date: 11/2/12 Recent Posts
Hey Dave,

Took a look here late in the day, at some QP:F member plumes for KDFW from the 4/23 12z run.

I'm seeing the extra members in this readout, BUT when I hover of the circled point, which has a value of about 0.08"  I do not see any members (p15) that have values that low. Perhaps the setting for that "rollover" is still limited to 20 values?

Also, I'm trying to understand what goes into the Raw-->BC calculation, that makes data for this same point, go up when BCd for some members, but down for others.  I guess it has to do with both the surrounding gridpoints forecasts/observed QPF/X-day QPE?



On Tue, Apr 23, 2024 at 10:03 AM David T Miller <VLab.Notifications@noaa.gov> wrote:
Jack,

Hey thanks very much for pointing this out to us!  It's good to know folks are looking at our displays and keeping us on our toes when there seem to be issues ;)

In investigating your questions, it turns out that the two issues are related.  Getting to your second question first, we discovered you're correct in that there are now 30 BC QPF06 perturbation members along with two controls, the average, and the standard deviation in the GRIB files.  The number of GEFS raw and BC members were increased when NAEFS v7 became operational on Dec 5.  As the average and member values are in different files for the raw GEFS QPF GRIB values, our data processing scripts adjusted for the increased number of members in those files.  However, the GEFS BC QPF GRIB files contain the  member and average values in one file, which is different from the raw as they're in separate files.  So the scripts that process the GEFS BC QPF GRIB files didn't adjust to the increased number of members as 20 were still expected.  Since the processing scripts for the raw GEFS member values adjusted with the increase in members without issue, I think we were just unaware that the processing scripts for the BC GEFS QPF data had not adjusted.  Note that, and somewhat in our defense, NCEP NCO online documentation still lists these GEFS BC QPF GRIB files as only having 20 members along with the controls, average/mean, and standard deviation.

So, to answer your first question, our data processing script hasn't been processing the correct GRIB message in the GEFS BC QPF GRIB v7 files due to the increased number of members in the files as the average/mean value is now in a different GRIB message number.

These issues should be fixed with today's GEFS 12Z run, which should start processing around 18-19Z.  

Again, thanks for letting us know!

Dave M

On Mon, Apr 22, 2024 at 5:57 PM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Happened to notice this afternoon that the plot line (dark blue) for the BC GEFS Average looks to be basically a max of the members plotted here, whereas the Raw GEFS looks more correctly like the Average of the plotted members.  

Is that due to a time-lagged approach to bias-correcting, at that far out forecast time step or something?  Also, with the GEFS membership now up to ~32, is it by choice just 20 are plotted here, or for readability?




--
Jack Settelmaier
Digital Techniques Meteorologist
NOAA/NWS, Southern Region HQ
Fort Worth, TX

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36670061VLab.Notifications@noaa.gov


--

David T. Miller

Research Associate, Programmer/Analyst

National Weather Service, Meteorological Development Laboratory

Cooperative Institute for Research in the Atmosphere/Colorado State University

1325 East West Highway |  Silver Spring, MD 20910  |  Phone (GV): (202) 936-5659  |  Fax: 301-713-9316


--
David T Miller Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36692620VLab.Notifications@noaa.gov

Attachments:

Attachment

DM
David T Miller, modified 1 Year ago.

RE: GEFS Raw vs BC

Youngling Posts: 10 Join Date: 11/7/13 Recent Posts
Jack,

Unfortunately, you have to switch to individual point hover rather than the comparison hover when there are more than a certain number of values.  That's a limitation of the display software and we haven't found a good way around this.  The individual or closest hover vs compare hover options can be selected by the options that appear when hovering the mouse in the upper right of the display.  You can zoom in to see if you can isolate the members you want to see values for.  We haven't found another way yet to get around this limitation.

As for your other question, you'd have to ask the GEFS folks.  Don't know what goes into BC calculations.

Dave M

On Tue, Apr 23, 2024 at 5:34 PM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Hey Dave,

Took a look here late in the day, at some QP:F member plumes for KDFW from the 4/23 12z run.

I'm seeing the extra members in this readout, BUT when I hover of the circled point, which has a value of about 0.08"  I do not see any members (p15) that have values that low. Perhaps the setting for that "rollover" is still limited to 20 values?

Also, I'm trying to understand what goes into the Raw-->BC calculation, that makes data for this same point, go up when BCd for some members, but down for others.  I guess it has to do with both the surrounding gridpoints forecasts/observed QPF/X-day QPE?



On Tue, Apr 23, 2024 at 10:03 AM David T Miller <VLab.Notifications@noaa.gov> wrote:
Jack,

Hey thanks very much for pointing this out to us!  It's good to know folks are looking at our displays and keeping us on our toes when there seem to be issues ;)

In investigating your questions, it turns out that the two issues are related.  Getting to your second question first, we discovered you're correct in that there are now 30 BC QPF06 perturbation members along with two controls, the average, and the standard deviation in the GRIB files.  The number of GEFS raw and BC members were increased when NAEFS v7 became operational on Dec 5.  As the average and member values are in different files for the raw GEFS QPF GRIB values, our data processing scripts adjusted for the increased number of members in those files.  However, the GEFS BC QPF GRIB files contain the  member and average values in one file, which is different from the raw as they're in separate files.  So the scripts that process the GEFS BC QPF GRIB files didn't adjust to the increased number of members as 20 were still expected.  Since the processing scripts for the raw GEFS member values adjusted with the increase in members without issue, I think we were just unaware that the processing scripts for the BC GEFS QPF data had not adjusted.  Note that, and somewhat in our defense, NCEP NCO online documentation still lists these GEFS BC QPF GRIB files as only having 20 members along with the controls, average/mean, and standard deviation.

So, to answer your first question, our data processing script hasn't been processing the correct GRIB message in the GEFS BC QPF GRIB v7 files due to the increased number of members in the files as the average/mean value is now in a different GRIB message number.

These issues should be fixed with today's GEFS 12Z run, which should start processing around 18-19Z.  

Again, thanks for letting us know!

Dave M

On Mon, Apr 22, 2024 at 5:57 PM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Happened to notice this afternoon that the plot line (dark blue) for the BC GEFS Average looks to be basically a max of the members plotted here, whereas the Raw GEFS looks more correctly like the Average of the plotted members.  

Is that due to a time-lagged approach to bias-correcting, at that far out forecast time step or something?  Also, with the GEFS membership now up to ~32, is it by choice just 20 are plotted here, or for readability?




--
Jack Settelmaier
Digital Techniques Meteorologist
NOAA/NWS, Southern Region HQ
Fort Worth, TX

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36670061VLab.Notifications@noaa.gov


--

David T. Miller

Research Associate, Programmer/Analyst

National Weather Service, Meteorological Development Laboratory

Cooperative Institute for Research in the Atmosphere/Colorado State University

1325 East West Highway |  Silver Spring, MD 20910  |  Phone (GV): (202) 936-5659  |  Fax: 301-713-9316


--
David T Miller Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36692620VLab.Notifications@noaa.gov

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36699287VLab.Notifications@noaa.gov


--

David T. Miller

Research Associate, Programmer/Analyst

National Weather Service, Meteorological Development Laboratory

Cooperative Institute for Research in the Atmosphere/Colorado State University

1325 East West Highway |  Silver Spring, MD 20910  |  Phone (GV): (202) 936-5659  |  Fax: 301-713-9316

JS
Jack Settelmaier, modified 1 Year ago.

RE: GEFS Raw vs BC

Youngling Posts: 67 Join Date: 11/2/12 Recent Posts
Thanks Dave.

Using the Compare Data on Hover, and hovering over one of the dots seemingly with a value of about 0.08", I can see that limit of ~24 members show up in that view, and it looks like it starts the count of 24 using members with the highest forecast QPF value, os that the lower values do not make the view?



Switching to the Individual/Closest Hover view, and again hovering over a dot with a seeming value of 0.27", I can peruse the 0.27" value from member p29, and can see about 7 more members with lesser values below.  However, when hovering over each, there are STILL several member's values that do not reveal.  In this particular case, I cannot see the p15 value, which, by tracing the plot from other time steps, I can see likely also has a value of 0.08"  So, it appears that, at least when zoomed in, one cannot get to some member's values, perhaps for being "under" the plot of another member with the same value?

Minutiae, for sure, but as I was trying to compare Raw and BC values from the members, to perhaps understand why some inflate the PQF significantly, while others get deflated in the BC process, I wanted to get the values for each, and could not.  In manually reviewing this date/time, I am unable to get the values from p06 and p15 to show up.  I'm guessing for them being "under" other dots with the same value.



I'll look for GEFS BC answers elsewhere, thanks.

On Tue, Apr 23, 2024 at 4:58 PM David T Miller <VLab.Notifications@noaa.gov> wrote:
Jack,

Unfortunately, you have to switch to individual point hover rather than the comparison hover when there are more than a certain number of values.  That's a limitation of the display software and we haven't found a good way around this.  The individual or closest hover vs compare hover options can be selected by the options that appear when hovering the mouse in the upper right of the display.  You can zoom in to see if you can isolate the members you want to see values for.  We haven't found another way yet to get around this limitation.

As for your other question, you'd have to ask the GEFS folks.  Don't know what goes into BC calculations.

Dave M

On Tue, Apr 23, 2024 at 5:34 PM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Hey Dave,

Took a look here late in the day, at some QP:F member plumes for KDFW from the 4/23 12z run.

I'm seeing the extra members in this readout, BUT when I hover of the circled point, which has a value of about 0.08"  I do not see any members (p15) that have values that low. Perhaps the setting for that "rollover" is still limited to 20 values?

Also, I'm trying to understand what goes into the Raw-->BC calculation, that makes data for this same point, go up when BCd for some members, but down for others.  I guess it has to do with both the surrounding gridpoints forecasts/observed QPF/X-day QPE?



On Tue, Apr 23, 2024 at 10:03 AM David T Miller <VLab.Notifications@noaa.gov> wrote:
Jack,

Hey thanks very much for pointing this out to us!  It's good to know folks are looking at our displays and keeping us on our toes when there seem to be issues ;)

In investigating your questions, it turns out that the two issues are related.  Getting to your second question first, we discovered you're correct in that there are now 30 BC QPF06 perturbation members along with two controls, the average, and the standard deviation in the GRIB files.  The number of GEFS raw and BC members were increased when NAEFS v7 became operational on Dec 5.  As the average and member values are in different files for the raw GEFS QPF GRIB values, our data processing scripts adjusted for the increased number of members in those files.  However, the GEFS BC QPF GRIB files contain the  member and average values in one file, which is different from the raw as they're in separate files.  So the scripts that process the GEFS BC QPF GRIB files didn't adjust to the increased number of members as 20 were still expected.  Since the processing scripts for the raw GEFS member values adjusted with the increase in members without issue, I think we were just unaware that the processing scripts for the BC GEFS QPF data had not adjusted.  Note that, and somewhat in our defense, NCEP NCO online documentation still lists these GEFS BC QPF GRIB files as only having 20 members along with the controls, average/mean, and standard deviation.

So, to answer your first question, our data processing script hasn't been processing the correct GRIB message in the GEFS BC QPF GRIB v7 files due to the increased number of members in the files as the average/mean value is now in a different GRIB message number.

These issues should be fixed with today's GEFS 12Z run, which should start processing around 18-19Z.  

Again, thanks for letting us know!

Dave M

On Mon, Apr 22, 2024 at 5:57 PM Jack Settelmaier <VLab.Notifications@noaa.gov> wrote:
Happened to notice this afternoon that the plot line (dark blue) for the BC GEFS Average looks to be basically a max of the members plotted here, whereas the Raw GEFS looks more correctly like the Average of the plotted members.  

Is that due to a time-lagged approach to bias-correcting, at that far out forecast time step or something?  Also, with the GEFS membership now up to ~32, is it by choice just 20 are plotted here, or for readability?




--
Jack Settelmaier
Digital Techniques Meteorologist
NOAA/NWS, Southern Region HQ
Fort Worth, TX

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36670061VLab.Notifications@noaa.gov


--

David T. Miller

Research Associate, Programmer/Analyst

National Weather Service, Meteorological Development Laboratory

Cooperative Institute for Research in the Atmosphere/Colorado State University

1325 East West Highway |  Silver Spring, MD 20910  |  Phone (GV): (202) 936-5659  |  Fax: 301-713-9316


--
David T Miller Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36692620VLab.Notifications@noaa.gov

--
Jack Settelmaier Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36699287VLab.Notifications@noaa.gov


--

David T. Miller

Research Associate, Programmer/Analyst

National Weather Service, Meteorological Development Laboratory

Cooperative Institute for Research in the Atmosphere/Colorado State University

1325 East West Highway |  Silver Spring, MD 20910  |  Phone (GV): (202) 936-5659  |  Fax: 301-713-9316


--
David T Miller Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum http://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/36699685VLab.Notifications@noaa.gov