ACCESS survey response submitted 18/12/2003.
Indicated interest in performing most experiments except DCPP.
Indicated likely to complete both concentration and emissions driven simulations but focus would be on emissions driven for building ensemble size.
Provided the following comments in the free text feedback section:
If we thought of the proposal (excluding the DECK) as three broad categories of experiments: (1) scenarios, (2) characterizing climate response/historical, (3) emerging issues, is the balance right? It seems like most of the model years contribute to group (2). We think there should be more new experiments targeting emerging issues and likely more emphasis on building ensembles for the scenario simulations.
What was well used for the last IPCC report may not be a good indication of what is needed this time. Perhaps we need to build more flexibility into the fast track to add experiments as the IPCC process develops, provided they are relatively simple to implement. Would it be useful to set up another task team to facilitate interaction with IPCC authors to be responsive to emerging needs, or is this part of the strategic ensemble design task team role?
To some extent CMIP7 will be defined by the Fast Track so we think it needs to focus on what is new. Partly that will be new models but it also needs to be new experiments. Emissions-driven can be an important part of what’s new and doesn’t seem to be highlighted enough in the current proposal. (One of the arguments we are using for trying to get additional local for CMIP7 funding is that it is different/adds to CMIP6 rather than just repeats).
While it is acknowledged that there is no expectation that all models will run all experiments, the desire to provide a manageable/realistic set of experiments still tends to imply that most models will attempt most experiments. Another approach would be to have an intentionally broader set of experiments with a target number of models to complete each one. Modelling centres could be encouraged to focus on experiments that best suit their model’s capability and the interest of the modelling centre. Would it even be worth considering a more coordinated approach where modelling centres are asked to prioritize certain experiments to spread the compute cost but still end up with a sufficient multi-model ensemble?