Forums

Back

RE: Raw vs. Bias Corrected NBM Violin Plots on the Component Point Visualization

RSS (Opens New Window)(Opens New Window)
PA
Patrick Ayd, modified 3 Years ago.

Raw vs. Bias Corrected NBM Violin Plots on the Component Point Visualization

Youngling Posts: 5 Join Date: 3/19/14 Recent Posts

For the Raw vs. Bias Corrected NBM Violin Plots on the Component Point Visualization, are you displaying an un-biased corrected NBM as the Raw (I did not know that existed or was sent)? Also, what does the BC Comp toggle do?

 

Thanks!

Attachments:

Attachment

DR
David Ruth, modified 3 Years ago.

RE: Raw vs. Bias Corrected NBM Violin Plots on the Component Point Visualization

Youngling Posts: 32 Join Date: 5/12/15 Recent Posts
Patrick, The WSUP prototype is being developed to investigate the feasibility of using the public cloud to visually convey whole story uncertainty and probabilistic forecast information to NWS forecast offices nationwide that is currently only available on supercomputers at the National Centers for Environmental Prediction (NCEP).   This includes point forecasts from raw NBM components (not bias-corrected) as well as bias-corrected NBM components that aren't disseminated with NBM grids.   The violin plot and BC Comp toggle enable you to view both. The thinking here is that in rapidly changing meteorological situations, bias-correction may not be as useful. DR On 3/21/2022 3:53 PM, Patrick Ayd wrote: > > For the Raw vs. Bias Corrected NBM Violin Plots on the Component Point > Visualization, are you displaying an un-biased corrected NBM as the > Raw (I did not know that existed or was sent)? Also, what does the BC > Comp toggle do? > > Thanks! > > > -- > Patrick Ayd Whole Story Uncertainty & Probabilities Viewer Virtual Lab > Forum > https://vlab.noaa.gov/web/wsup/forums/-/message_boards/view_message/21285914 > VLab.Notifications@noaa.gov
MS
Michael T Stavish, modified 3 Years ago.

RE: Raw vs. Bias Corrected NBM Violin Plots on the Component Point Visualization

Youngling Posts: 4 Join Date: 9/20/12 Recent Posts

Hi,

 

I was looking at this today and was curious about the BC Comp toggle as well.

- Is the time series data in the Point Vis (I'm viewing MaxT) always going to show the bias corrected suite regardless of the BC Comp toggle? In that case the quartile values should not change upon toggling on/off, correct? Or do they? It's hard to see, the scale appears to get tweaked a bit but I'm not sure the numbers do.

- And, when toggling the BC Comp button with MaxT  displayed, the violin plot seems to simply switch to a mirror image of what was previously shown. I.e Raw on one side and BC Comp on the other, then visa-versa. 

 

Apologies if I'm missing something obvious. Thanks for any help,

Mike

DR
David Ruth, modified 3 Years ago.

RE: Raw vs. Bias Corrected NBM Violin Plots on the Component Point Visualization

Youngling Posts: 32 Join Date: 5/12/15 Recent Posts

Mike,

NBM Quantiles are produced via NBM's Quantile Mapping and Dressing process, which is based on bias-corrected components only.   Therefore, quantiles displayed on the time series do not change when BC comp is toggled.

The violin plot always shows both bias-corrected and raw (non bias-corrected) components.   The toggle simply switches their position on the x axis of the violin.

DR

On 3/23/2022 1:58 PM, Michael T Stavish wrote:

Hi,

I was looking at this today and was curious about the BC Comp toggle as well.

- Is the time series data in the Point Vis (I'm viewing MaxT) always going to show the bias corrected suite regardless of the BC Comp toggle? In that case the quartile values should not change upon toggling on/off, correct? Or do they? It's hard to see, the scale appears to get tweaked a bit but I'm not sure the numbers do.

- And, when toggling the BC Comp button with MaxT  displayed, the violin plot seems to simply switch to a mirror image of what was previously shown. I.e Raw on one side and BC Comp on the other, then visa-versa. 

Apologies if I'm missing something obvious. Thanks for any help,

Mike

MS
Michael T Stavish, modified 3 Years ago.

RE: Raw vs. Bias Corrected NBM Violin Plots on the Component Point Visualization

Youngling Posts: 4 Join Date: 9/20/12 Recent Posts

Thanks Dave,

 

I get it now, I guess toggling between bc/non-bc, quantile vs component, and using data on QMD runs vs, not, I somehow  got confused!  I was thinking something was supposed to change in the violins with that toggle. BC and Raw - what else is there? I don't know what I was thinking there. 

 

On thing is, when looking at time series of the Max/MinT plots in the 1D viewer, we see adjustable quantiles on every run. That could be part of the blurriness here. After one of the recent SOO calls I was having a chat with someone about how they were identifying that the operational forecast was frequently falling below the 50th percentile even toward the 25th. I inquired how they were visualizing that, and they pointed out the time series in 1d and the black dots. Then I was sharing the point vis method of hovering with the "compare data on hover" toggled. I suppose any run could still show the quantiles/quartiles right? I'm curious of how the QMD's run quantiles would be of a different flavor versus quantiles of the BC data in non-qmd hours. 

 

Thanks for your help,

Mike

DR
David Ruth, modified 3 Years ago.

RE: Raw vs. Bias Corrected NBM Violin Plots on the Component Point Visualization

Youngling Posts: 32 Join Date: 5/12/15 Recent Posts

Mike,

1D displays the latest available quantile, which for a current forecast will come from a model run 6 or more hours earlier than deterministic forecast data.   This can explain apparent mismatches between the dots and the boxes in 1D.

In contrast, Point Vis always shows quantiles and deterministic data from the same model run times together.

DR

On 3/28/2022 8:01 PM, Michael T Stavish wrote:

Thanks Dave,

I get it now, I guess toggling between bc/non-bc, quantile vs component, and using data on QMD runs vs, not, I somehow  got confused!  I was thinking something was supposed to change in the violins with that toggle. BC and Raw - what else is there? I don't know what I was thinking there. 

On thing is, when looking at time series of the Max/MinT plots in the 1D viewer, we see adjustable quantiles on every run. That could be part of the blurriness here. After one of the recent SOO calls I was having a chat with someone about how they were identifying that the operational forecast was frequently falling below the 50th percentile even toward the 25th. I inquired how they were visualizing that, and they pointed out the time series in 1d and the black dots. Then I was sharing the point vis method of hovering with the "compare data on hover" toggled. I suppose any run could still show the quantiles/quartiles right? I'm curious of how the QMD's run quantiles would be of a different flavor versus quantiles of the BC data in non-qmd hours. 

Thanks for your help,

Mike

MS
Michael T Stavish, modified 3 Years ago.

Thank you!!RE: Raw vs. Bias Corrected NBM Violin Plots on the Component Point Visualization

Youngling Posts: 4 Join Date: 9/20/12 Recent Posts

Got it, thank you!!