[gradsusr] Cluster Ensemble members

[Bill Bua - NOAA Affiliate]

Thomas --

Correlation of each member to the others is part of the first step, and
represents the distance of the members from each other. Another part is
choosing the forecast variable most relevant for your clustering .... 500
hPa heights (this is a common one), sea level pressure, or some other
variable. While a global domain can be used, you may even want to choose a
smaller domain over which to cluster, say Europe or North America.

Once those decisions are made, things get a little fuzzy for me, but I can
recommend you read Tracton and Kalnay 1993 (Tracton, M. S., and E. Kalnay,
1993: Operational ensemble prediction at the National Meteorological
Center: Practical aspects.* Wea. Forecasting*, *8*, 379–398) or.Toth et al.
(Toth, Z., E. Kalnay, S. M. Tracton, R. Wobus, and J. Irwin, 1997: A
synoptic evaluation of the NCEP ensemble. Wea. Forecasting,12, 140–153
<http://journals.ametsoc.org/doi/pdf/10.1175/MWR-D-11-00016.1>.) for
details of using *anomaly* correlation as a distance metric in a clustering
algorithm.

Good luck!

Dr. William R. (Bill) Bua
UCAR/COMET/NCEP/EMC
5830 University Research Court #2784
College Park, MD 20740


On Wed, Feb 17, 2016 at 5:32 PM, Thomas Davies <tsd.davies at gmail.com> wrote:

> Thanks for your reply.
>
> I looped over each GEFS ensemble member, calculating the correlation with
> every other member and stored these values like this:
>
> *p1p2 0.625844*
> *p1p3 0.778631*
> *p1p4 0.718804*
> *p1p5 0.597726*
> *p1p6 0.73876*
> *p1p7 0.744276*
> *p1p8 0.683437*
> *p1p9 0.548653*
> *p1p10 0.479272*
> *p1p11 0.668351*
> *p1p12 0.600638*
> *p1p13 0.698219*
> *p1p14 0.606607*
> *p1p15 0.536851*
> *p1p16 0.736465*
> *p1p17 0.564005*
> *p1p18 0.557531*
> *p1p19 0.641763*
> *p1p20 0.5872*
> *p1p21 0.57592*
> *...*
> p21p1 0.57592
> p21p2 0.623274
> p21p3 0.630096
> p21p4 0.551386
> p21p5 0.561231
> p21p6 0.60267
> p21p7 0.638492
> p21p8 0.646961
> p21p9 0.5132
> p21p10 0.621424
> p21p11 0.684271
> p21p12 0.541928
> p21p13 0.502403
> p21p14 0.671402
> p21p15 0.637574
> p21p16 0.76546
> p21p17 0.564914
> p21p18 0.580842
> p21p19 0.644114
> p21p20 0.599859
>
> However I am not sure how I should progress. I have read through the link
> you provided but couldn't find anything useful, all I know is that I need
> to firstly correlate between members
>
> On Wed, Feb 17, 2016 at 6:02 PM, Stephen McMillan <
> smcmillan at planalytics.com> wrote:
>
>> Thomas,
>>
>> In the GrADS Documentation Index (
>> http://cola.gmu.edu/grads/gadoc/gadocindex.html), under the letter "E"
>> you can find several relevant topics, such as "ensemble dimension" and
>> "eloop."  One or more may answer your questions.
>>
>> Stephen McMillan
>>
>> On Wed, Feb 17, 2016 at 12:27 PM, Thomas Davies <tsd.davies at gmail.com>
>> wrote:
>>
>>> Hi,
>>>
>>> Does anybody know how to cluster ensemble members showing similar
>>> solutions and display? Is there a quick command to do this
>>>
>>> Thanks in advance
>>>
>>> _______________________________________________
>>> gradsusr mailing list
>>> gradsusr at gradsusr.org
>>> http://gradsusr.org/mailman/listinfo/gradsusr
>>>
>>>
>>
>> _______________________________________________
>> gradsusr mailing list
>> gradsusr at gradsusr.org
>> http://gradsusr.org/mailman/listinfo/gradsusr
>>
>>
>
> _______________________________________________
> gradsusr mailing list
> gradsusr at gradsusr.org
> http://gradsusr.org/mailman/listinfo/gradsusr
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://gradsusr.org/pipermail/gradsusr/attachments/20160218/a2c2504a/attachment.html