Experiments

The objective of QBOi is to evaluate the sensitivity dependencies of tropical stratosphere variability within current global climate models. In so doing, we anticipate learning more about the modelled dynamics and variability of the tropical stratosphere. On a practical level, it is hoped this project will help improve QBO modelling in time for CMIP6.

It is commonly believed that to generate a QBO, GCMs require the right choice of: resolution (spatial and temporal), parameterisation (convection and GWs) and numerics (diffusion, numerical solvers). The choice of parameters is not completely unique: getting a QBO in one model, with one set of parameters, may not necessarily result in a QBO in another model. Having said this, one might suspect particular parameter choices being more likely to result in a QBO than others. We propose to explore these parameter settings, across a range of GCMs, to better identify robust effects.

We propose two levels of participation within QBOi. The first, QBOi-lite, assumes one or more sets of model runs with no changes in model formulation (but may include initial condition ensembles). The intention for these runs would be to compare across models i.e. an intermodel comparison. Although such an analysis is not without its problems, it is nothing different from other studies comparing different models e.g. CMIP5. A second level of participation, and more in line with the spirit of the project, is for separate runs using one model with structural changes to resolution, parameterisation etc. Differences would first be sought within a model, before then being checked across models for robustness.

It is suggested that runs should be AMIP styled (atmosphere only) and run over the recent past (1960 onwards). More specifically, it is suggested that boundary/ancillary fields used for CMIP5 should be employed. As many participating groups would have been involved in CMIP5, this would seem a natural choice.

The diagnostics which we suggest include those terms required to close the momentum budget, so include 4-daily instantaneous: u, v, Z, GW flux/tendencies. Presumably, there is flexibility for instantaneous or time-average output, but perhaps the former would be better(?). For completeness, temperature/heating terms (T, Q_sw, Q_lw) would be advantageous, but perhaps monthly mean output would be suitable. Obviously, the high frequency diagnostics would be used to derived terms in the TEM equations (e.g. EP flux) which your model may already log. However, it is recognised that model diagnostics should be openly discussed within the wider group.

But what do you think? Should experiments be more prescriptive? Should specific resolutions, parameterisations be suggested? What is the right balance of diagnostics which should be included? A number of these questions also require a discussion of the science people wish to do. If you think this is important do say so, otherwise science topics will be deferred to a separate discussion blog. Please feel free to contribute to this discussion, your thoughts are most welcome.

Update 22 March 2015 - Following the discussion coming out from the first QBOi Workshop held in Victoria, BC 16-18 March 2015, a number of experiments were endorsed. The details of these will be discussed in separate blogs found on this website.