Stochastic Aspects of Convection

There are many conceptual problems with standard, deterministic, convection schemes. Such schemes start from a basic, but really rather dubious, assumption that the "average properties" of the unresolved convection within each grid box can somehow be determined as functions of the large-scale, resolved variables. Such an assumption can only be reasonable if the parameterised convective behaviour in the grid box is considered to be obtained by averaging over some a large number of convective clouds within that grid box. But, there are many situations where that assumption will break down, particularly for grid box sizes that are around the transition region between parameterised and explicitly-resolved convection.

We have built a stochastic convection scheme. Rather than releasing a single convective mass having some total flux, the scheme considers instead the release of a number of convective air masses with randomly-chosen updraft fluxes that sum approximately to a known total flux. Thus, at any timestep there are a variable number of clouds with distinct properties. One nice feature about this scheme is that it should be stochastic at short grid lengths, with a small number of clouds per grid box, but will tend towards being an old-fashioned deterministic scheme for large grid boxes with many clouds.

Some links for this work:

Papers:

1. A paper on tests within the Met Office regional ensemble
2. A paper on single-column testing in an ensemble framework
3. PhD thesis by Michael Whitall on high-frequency variability
4. A paper on averaging lengthscales for the stochastic parameterization
5. A paper on single-column testing of stochastic methods
6. A paper on the stochastic parameterization

Book chapters:

1. A chapter reviewing stochastic parameterizations
2. A chapter on stochastic parameterization testing

Talks:

1. A talk on implementation issues with stochastic parameterizations presented at the Future of Cumulus Parameterization conference, Delft, 10-14 July 2017.
2. A talk on stochastic aspects of convection-permitting models presented at the Joint SRNWP Workshop on Model Physics and Ensemble Prediction Systems, Berlin, 27-29 March 2017.
3. A talk on the PC scheme in MOGREPS presented by Richard Keane at the Joint SRNWP Workshop on Model Physics and Ensemble Prediction Systems, Madrid, 18-20 June 2013.
4. A talk on stochastic aspects of boundary-layer parameterization presented at an NCAS workshop.
5. A talk presented at the ECMWF workshop on Representing Model Uncertainty and Error in Weather and Climate Prediction. The write up for the proceedings can be found here.
6. A talk, presented by Michael Whitall at the EGU Assembly 2010 in Vienna.
7. A talk, presented by Richard Keane as a seminar at LMU, Munich.
8. A talk, presented by Richard Keane, on the development of the stochastic convection parameterisation scheme, given as a departmental seminar in Reading.
9. A talk presented by Richard Keane at the 9th European Conference on Applications of Meteorology, Toulouse, France, 28 September-2 October 2009.
10. A talk presented by Richard Keane at the 4th SRNWP Workshop on Short-range Ensemble Prediction Systems, Exeter, 23-25 June 2009.
11. An invited talk on near grid-scale noise, given as at a Workshop on Mathematical challenges in climate science, 9-13 March, Lorentz Centre, Leiden, Netherlands.
12. A talk, presented by Michael Ball, comparing stochastic parameterisation approaches in a single-column model, given as a departmental seminar in Reading.
13. A talk presented at the EGU Assembly 2006 in Vienna.
14. A talk presented at the RMetS conference 2005 in Exeter.
15. A talk presented at a meeting of the RMS Dynamical Problems Group (Also presented at DLR-Institut fuer Physik der Atmosphaere, Germany)
16. A talk (given by George Craig) presented at the ECMWF workshop on Representation of sub-grid processes using stochastic-dynamic models. The write up for the proceedings can be found here.
17. A talk on observational requirements of stochastic convection schemes, presented at the Planning Workshop for the CSIP campaign

Posters:

1. A poster presented by Richard Keane at the 5th WGNE workshop on systematic errors in weather and climate models, 2017.
2. A poster presented by Richard Keane at the Royal Meteorological Society Conference 2009.
3. A poster (presented by George Craig) on the physical basis for the stochastic convection scheme, presented in Hamburg, 2005.
4. Extended abstract for the poster presented at the 14th conference on clouds and precipitation, organized by ICCP.
5. A poster on the stochastic convection scheme, presented at the Royal Meteorological Society Conference 2003

Others:

1. Working group report from the ECMWF/WWRP workshop on Model Uncertainty.
2. Working group reports from the ECMWF workshop on Representing Model Uncertainty and Error in Weather and Climate Prediction.
3. Working group reports from the ECMWF workshop on Representation of sub-grid processes using stochastic-dynamic models.