[Go-essp-tech] RIP clarification

[Karl Taylor]

Hi Charlotte,

Congratulations! you managed to crack the intended use of "rip" in spite 
of what I thought was  a scheme so complicated, that no mere mortal 
could be expected to decipher it.

That being said, I think we should probably expect others (with lesser 
proclivity to meticulousness) to fail in attempting to follow the 
recommendations.  In particular, we may not be able to count on the "p" 
and "i" values being consistent across all simulations by a single 
model.   If that were the case, would it completely disrupt your plans?

One approach would be as follows:
The first time that someone records the information for an experiment 
(i.e., the first member of the "rip" ensemble), they would be asked to 
document the model and all the experimental conditions.  Then when 
subsequent members were entered, they should only have to reference the 
first member and say how the model physics (in some cases when p 
differs) or the experiment conditions (initial conditions when "i" 
differs and in some cases "forcing" when "p" differs).  In the case of a 
different value of "r", typically the "spawning" point from the run will 
differ from one member of the ensemble to another.  [Note, however, that 
for the TAMIP project, this identifies a different start time for the 
forecast within the season.]

Sorry it's so difficult.

Best regards,
Karl

On 2/1/11 6:09 AM, charlotte.pascoe at stfc.ac.uk wrote:
>
> Hi Karl and Bryan,
>
> It is important to know how the r<N>i<M>p<L> indices are to be 
> implemented in CMIP5 so that we can be sure that information about 
> them is captured correctly by the questionnaire. Here is my 
> understanding, please correct me if I’m wrong!
>
> The scope of the p<L> and i<M> indices are model wide – these indices 
> must mean the same thing across all simulations for a particular model.
>
> The scope of the r<N> indices are experiment wide – the r index is not 
> required to mean the same thing across all simulations for a 
> particular model, so the r indices can be reused by different experiments.
>
> Here is an example to show what this would mean in practice.
>
> If the meaning of the r<N> indices were confined to an experiment
>
> then if you had used r1, r2, and r3 to label the members of an amip 
> ensemble:
>
> tas_day_HADCM3_amip_r1i1p1_197901-200901.nc
>
> tas_day_HADCM3_amip_r2i1p1_197901-200901.nc
>
> tas_day_HADCM3_amip_r3i1p1_197901-200901.nc
>
> you could use r1,r2 and r3 again to label the members of a different 
> experiment, say decadal2000:
>
> tas_day_HADCM3_decadal2000_r1i2p1_200001-201001.nc
>
> tas_day_HADCM3_decadal2000_r2i2p1_200001-201001.nc
>
> tas_day_HADCM3_decadal2000_r3i2p1_200001-201001.nc
>
> rather than having to begin labelling from a new r index, say r4:
>
> tas_day_HADCM3_decadal2000_r4i2p1_200001-201001.nc
>
> Where
>
> p<L> - identify perturbations in physics (from the same base model)
>
> i<M> - identify perturbations in initialisation method
>
> r<N> - identify ensemble members which differ in other characteristics.
>
> Please shout if my interpretation is wrong!
>
> Kind regards,
>
> *Charlotte*
>
> ----------------------------------------------------------
>
> Dr Charlotte Pascoe
>
> NCAS British Atmospheric Data Centre
>
> STFC Rutherford Appleton Laboratory
>
> Phone +44 1235 445869; Fax ...5848
>
> e-mailcharlotte.pascoe at stfc.ac.uk <mailto:c.l.pascoe at rl.ac.uk>
>
> ----------------------------------------------------------
>
>
> -- 
> Scanned by iCritical.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/go-essp-tech/attachments/20110201/a21dd121/attachment-0001.html