RE: new component/member pie plot histogram and component weights Q - Forums - Whole Story Uncertainty & Probabilities Viewer

Ed,

Thanks for contacting us with these questions.

There are light vertical bars across the y-scale of the histograms that reveal the counts for each value (1 bar = 1 model component).   I put together a table (link below) showing the raw and bias-corrected values for each model component from your Point Vis display.  The counts for raw and bias-corrected match the histogram except the bias-corrected histogram shows 3 at 104 and 1 at 105 instead of 4 at 104 in Point Vis.   I will ask our developer to look into this, but I suspect it is rounding.

https://docs.google.com/presentation/d/11UxM-e35KlNXm72XmAu4AxQFwnDDYz5-VgSHL8i8BsM/edit?usp=sharing

Unfortunately, NBM only provides its component weights for 12z and 00z inits (not 17z).   If you look at 12z and 00z for KHRI, you will see that gefs, ecwmwfe, and ecwmwfd are highly weighted models.   This is consistent with your interpretation that NBM is favoring the warmer model components at KHRI.

DR

Hey,

I want to be sure I am interpreting the new component/member pie plot histograms and the component weights correctly.

The NBM deterministic forecast at KHRI (eastern OR) for Thursday, Valid Fri 7/21 06Z, is 103 F. 

The pie chart shows the majority of components BC and raw are under the dataset value of 103 based on the pie chart and histogram. The median of all the raw guidance is 101 and for the bias corrected components is 102.5 F.

Qs:

1) Is my conclusion correct that I'm drawing in this example: the deterministic MaxT final blended value that the NBM is coming up with must be due to one or two of the warmer components and their larger weight (considering most of the components are less than the NBM dataset value of 103 F). A spreadsheet showing the screenshots is here too.

2) is there a default y-scale for the number of counts when viewing the histogram plots?

Thanks!

--
Edward Townsend Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum https://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/30814559VLab.Notifications@noaa.gov

I meant to say light horizontal bars (not vertical).

DR

Ed,

Thanks for contacting us with these questions.

There are light vertical bars across the y-scale of the histograms that reveal the counts for each value (1 bar = 1 model component).   I put together a table (link below) showing the raw and bias-corrected values for each model component from your Point Vis display.  The counts for raw and bias-corrected match the histogram except the bias-corrected histogram shows 3 at 104 and 1 at 105 instead of 4 at 104 in Point Vis.   I will ask our developer to look into this, but I suspect it is rounding.

https://docs.google.com/presentation/d/11UxM-e35KlNXm72XmAu4AxQFwnDDYz5-VgSHL8i8BsM/edit?usp=sharing

As you can see, the Bias-corrected models with the highest forecast values are gefs, ecmwfe, ecmwfd, and gfs.

Unfortunately, NBM only provides its component weights for 12z and 00z inits (not 17z).   If you look at 12z and 00z for KHRI, you will see that gefs, ecwmwfe, and ecwmwfd are highly weighted models.   This is consistent with your interpretation that NBM is favoring the warmer model components at KHRI.

DR

On 7/17/2023 3:02 PM, Edward Townsend wrote:

Hey,

I want to be sure I am interpreting the new component/member pie plot histograms and the component weights correctly.

The NBM deterministic forecast at KHRI (eastern OR) for Thursday, Valid Fri 7/21 06Z, is 103 F. 

The pie chart shows the majority of components BC and raw are under the dataset value of 103 based on the pie chart and histogram. The median of all the raw guidance is 101 and for the bias corrected components is 102.5 F.

Qs:

1) Is my conclusion correct that I'm drawing in this example: the deterministic MaxT final blended value that the NBM is coming up with must be due to one or two of the warmer components and their larger weight (considering most of the components are less than the NBM dataset value of 103 F). A spreadsheet showing the screenshots is here too.

2) is there a default y-scale for the number of counts when viewing the histogram plots?

Thanks!

--
Edward Townsend Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum https://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/30814559VLab.Notifications@noaa.gov

Ed,

Dave Miller has looked into the apparent discrepancy in the histogram counts and point vis.    As it turns out, the histogram you sent was actually for the 16z NBM v4.1 init.   The histogram for the 17z NBM v4.1 init has the correct number of 104 and 105 component values plotted.

https://docs.google.com/presentation/d/1-SBLIPIgkmo93ZET-Wz17fZJRe9-kxYC5x45ds-uCz4/edit?usp=sharing

Please let me know if you have other questions.

DR

I meant to say light horizontal bars (not vertical).

DR

On 7/17/2023 5:58 PM, David.Ruth wrote:

Ed,

Thanks for contacting us with these questions.

There are light vertical bars across the y-scale of the histograms that reveal the counts for each value (1 bar = 1 model component).   I put together a table (link below) showing the raw and bias-corrected values for each model component from your Point Vis display.  The counts for raw and bias-corrected match the histogram except the bias-corrected histogram shows 3 at 104 and 1 at 105 instead of 4 at 104 in Point Vis.   I will ask our developer to look into this, but I suspect it is rounding.

https://docs.google.com/presentation/d/11UxM-e35KlNXm72XmAu4AxQFwnDDYz5-VgSHL8i8BsM/edit?usp=sharing

As you can see, the Bias-corrected models with the highest forecast values are gefs, ecmwfe, ecmwfd, and gfs.

Unfortunately, NBM only provides its component weights for 12z and 00z inits (not 17z).   If you look at 12z and 00z for KHRI, you will see that gefs, ecwmwfe, and ecwmwfd are highly weighted models.   This is consistent with your interpretation that NBM is favoring the warmer model components at KHRI.

DR

On 7/17/2023 3:02 PM, Edward Townsend wrote:

Hey,

I want to be sure I am interpreting the new component/member pie plot histograms and the component weights correctly.

The NBM deterministic forecast at KHRI (eastern OR) for Thursday, Valid Fri 7/21 06Z, is 103 F. 

The pie chart shows the majority of components BC and raw are under the dataset value of 103 based on the pie chart and histogram. The median of all the raw guidance is 101 and for the bias corrected components is 102.5 F.

Qs:

1) Is my conclusion correct that I'm drawing in this example: the deterministic MaxT final blended value that the NBM is coming up with must be due to one or two of the warmer components and their larger weight (considering most of the components are less than the NBM dataset value of 103 F). A spreadsheet showing the screenshots is here too.

2) is there a default y-scale for the number of counts when viewing the histogram plots?

Thanks!

--
Edward Townsend Whole Story Uncertainty & Probabilities Viewer Virtual Lab Forum https://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/30814559VLab.Notifications@noaa.gov