Dear all - can someone please inform me on the available stochastic perturbation schemes available in oifs?

Are the Stochastically Perturbed Parametrization Tendency scheme (SPPT, Palmer et al., 2009; Buizza et al., 1999) and the Stochastic Kinetic Energy Backscatter scheme (SKEB, Berner et al., 2009; Shutts, 2005) available in openifsr43r3 ?

And if so -is there any documentation aside from the scientific description in 3.8 OpenIFS: Modelling Uncertainty ?

Many thanks ! Etienne

4 Comments

  1. Hi Etienne,

    Yes, in short, both SPPT and SKEB are available in OpenIFS 43r3. You'll need a small patch for 43r3v1 as there's some code (inadvertently) missing to enable SKEB.   There is also mention of the 'SPP' scheme in the code. This is not a typo, but a replacement for SPPT which, at 43r3, is in the early stages of development and should not be used.  Operationally, at 43r3, SKEB was 'turned' right down and had only a small effect on the spread. The advice I've had from the stochastic experts here is that for OpenIFS 43r3 it's best to use only SPPT.

    Documentation.

    The scientific detail for the ensemble schemes is contained in the IFS CY43r3 manual which you can find here: https://www.ecmwf.int/en/elibrary/17737-part-v-ensemble-prediction-system

    The user guide is lacking some practical examples of using ensembles with OpenIFS which I'm aware of and on my list to do as soon as possible. In the meantime, there is some helpful information from the OpenIFS workshops which have used ensembles. See OpenIFS 2017: Creating a new OpenIFS experiment#Runningtheforecastexperiment . Much of this is still valid for 43r3.

    It's important to be bear in mind that the parameters for the SPPT scheme operationally are chosen to give a forecast spread that matches the error based on many simulations of the 10 day forecast. The seasonal prediction system also uses SPPT but a different set of SPPT parameters for that reason.

    Hope that helps.

    Best regards,  Glenn


  2. Thanks for such a prompt and detailed answer Glenn!  So if I understand SPP would be used in place of SPPT (with a small patch), and that SKEB is not needed, great!

    The application we potentially envisage for SPP(T) would be mainly for seasonal forecasting (and perhaps climate projections???) sometime in the future of EC-Earth4, so it's good to know of our options!

  3. Hi Etienne,

    No, don't use SPP for now, only SPPT. I only mentioned it because it confuses some people that the names are so similar. They are quite different schemes and SPPT is the one to use.

    Just realised I typed the wrong information above. SPPT works fine openifs 43r3v1, it's the SKEB scheme which has a subroutine missing if you choose to enable it.  This bug has been fixed in version 2 of openifs 43r3 due early next year.

    If you just want to 'get a spread of results' that easy enough with SPPT, tuning the spread to be a realistic representation of errors is where it gets more involved.

    Cheers,  Glenn

  4. Hi again - ok I misread your first explanation on SPP.  Thanks for the clarifications.