Jump to content
  • Member Statistics

    17,514
    Total Members
    7,904
    Most Online
    CHSVol
    Newest Member
    CHSVol
    Joined

December 19-21 Storm Thread


Deck Pic

Recommended Posts

  • Replies 914
  • Created
  • Last Reply

thx for that

I wonder if they keep stats on the performance of each member like they do with the operational....and if so, which ones are better

one would think they would but I don't know if it would be published

The members use stochastic components (pseudo-random perturbations during the integration) to address model uncertainty.  Also, one of the main tenets of a well-designed ensemble prediction systems is that any member should be as equally skillful as any other member (and the true state should fit in the ensemble envelope just as a member).  I suspect a large enough sample would show that there are no members that are more skillful.   I bet that one batch is slightly more skillful than the other.  The ECMWF ensemble is well designed and well calibrated, and quantitative statistics bears this out.

 

The only problem with the aforementioned for the ECMWF EPS is that some of the perturbations are cycled in time, and they use two classes of initial perturbations:  1) initial singular vectors, and 2) analysis based from their ensemble of 4DVARs. 

 

Regarding the earlier bits....their ensemble is 50 perturbed members (IC and stochastic model perturbations) + 1 control run (non perturbed) = 51 total members.

Link to comment
Share on other sites

The members use stochastic components (pseudo-random perturbations during the integration) to address model uncertainty.  Also, one of the main tenets of a well-designed ensemble prediction systems is that any member should be as equally skillful as any other member (and the true state should fit in the ensemble envelope just as a member).  I suspect a large enough sample would show that there are no members that are more skillful.   I bet that one batch is slightly more skillful than the other.  The ECMWF ensemble is well designed and well calibrated, and quantitative statistics bears this out.

 

The only problem with the aforementioned for the ECMWF EPS is that some of the perturbations are cycled in time, and they use two classes of initial perturbations:  1) initial singular vectors, and 2) analysis based from their ensemble of 4DVARs. 

 

Regarding the earlier bits....their ensemble is 50 perturbed members (IC and stochastic model perturbations) + 1 control run (non perturbed) = 51 total members.

thx for that explanation

Link to comment
Share on other sites

So, should we consider the control run just "another ensemble member", or is there something more specific to be gained by looking at it?

It is sort of like a "baseline" run for the perturbed members.  A few things:

1) It is like the operational run except that it is run at lower (ensemble) resolution (T639 instead of T1279).

2) It does not have it's own data assimilation, but uses a truncated analysis from the OP

3) Since the ensemble is coupled to a dynamic ocean model, I am assuming that the control forecast is as well (the deterministic OP run is not coupled to an ocean model, only a wind-wave model).

 

The control will exhibit somewhat different behavior than the members since it does not include the stochastic physics parameterizations (SKEB, SPPT).  There is more reading on their system available here:

http://old.ecmwf.int/research/ifsdocs/CY40r1/IFSPart5.pdf

 

In a sense, yes, you can consider it like "another member of the ensemble".  I would not look it on its own like we do the operational deterministic (high resolution) integration since it lacks assimilation, etc. 

Link to comment
Share on other sites

Stochastic.

Is that anythign like fantastic???

Obviously....

 

sto·chas·tic
stəˈkastik/
adjective
 
  1. randomly determined; having a random probability distribution or pattern that may be analyzed statistically but may not be predicted precisely.
Link to comment
Share on other sites

12z Euro Ens still very much on point... expecting Euro Op to bounce back soon.

In general, it seems like the Euro Op and the ensembles don't often diverge very much compared to the GEFS and GFS Op.  The differences on the 12z runs between the Euro Op and ensembles, particularly when other model systems now agree quite well with the Euro ensembles are a red flag IMO. 

Link to comment
Share on other sites

dtk, I've got a question,

 

Instead of using a control run for the ensambles, why don't they just used the OP run as a conrol since it's running on a higher resolution. The truncation analysis limits the ens control runs max and mins anyways anyways, it seems like your basically trying to keep the ens control run as close in line with the OP.

 

It seems weird, but would it save computing time just to simplify the system and only used the OP as your control run for the ens?

Link to comment
Share on other sites

dtk, I've got a question,

 

Instead of using a control run for the ensambles, why don't they just used the OP run as a conrol since it's running on a higher resolution. The truncation analysis limits the ens control runs max and mins anyways anyways, it seems like your basically trying to keep the ens control run as close in line with the OP.

 

It seems weird, but would it save computing time just to simplify the system and only used the OP as your control run for the ens?

Except that it is not then representative of the other members.  The control member will include the same physics as the other members (which is different than the OP), and resolution can make a big difference.  Plus, it is dynamically coupled to an ocean model with makes things quite a bit different.  Having a control member to compare to the OP helps to quantify where/why/how the ensemble may be different.  It also helps understand and correct for systematic biases between the high resolution configuration and the lower resolution ensemble configuration.

Link to comment
Share on other sites

Except that it is not then representative of the other members.  The control member will include the same physics as the other members (which is different than the OP), and resolution can make a big difference.  Plus, it is dynamically coupled to an ocean model with makes things quite a bit different.  Having a control member to compare to the OP helps to quantify where/why/how the ensemble may be different.  It also helps understand and correct for systematic biases between the high resolution configuration and the lower resolution ensemble configuration.

thanks, that makes a lot sense

Link to comment
Share on other sites

The difference between the strength and orientation of the shortwaves between the GFS Op and Euro Op are night and day. The GFS is weaker with both pieces and it basically lags the northern s/w behind the southern system causing it to be a weak wave that just comes and goes. Interested to see the PGFS compared to its last solution.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

  • Recently Browsing   0 members

    • No registered users viewing this page.

×
×
  • Create New...