Greetings!
I am working to set up some large ensemble CESM runs with an aim to generate some realistic statistics around what might happen with different forcing scenarios. My working assumption so far is that I'll need ensembles with a wide enough variety of weather patterns to capture natural variations so my statistics are meaningful. I am currently considering using the F2010climo compset. Since this uses a prescribed SST, my assumption is that this won't capture multi-year cycles like ENSO, however, so this might not give me the type of variation I need.
I presume a fully coupled model would show ENSO-like behaviors. Are there less computationally expensive models that would also show this, though? Would CAM6 plus a slab ocean be expressive enough?
More generally, I'm curious how people evaluate whether or not their ensembles have a "reasonable" amount of variation. For example, I know I can use pertlim and pertseed in user_nl_cam to tweak the initial temperature fields at simulation startup and get variation that way. Presumably these perturbed simulations need to run for some time for the perturbations to result in different enough weather. Any advice on how to know that I've got large enough perturbations and long enough settling times?
Best,
-Rob
I am working to set up some large ensemble CESM runs with an aim to generate some realistic statistics around what might happen with different forcing scenarios. My working assumption so far is that I'll need ensembles with a wide enough variety of weather patterns to capture natural variations so my statistics are meaningful. I am currently considering using the F2010climo compset. Since this uses a prescribed SST, my assumption is that this won't capture multi-year cycles like ENSO, however, so this might not give me the type of variation I need.
I presume a fully coupled model would show ENSO-like behaviors. Are there less computationally expensive models that would also show this, though? Would CAM6 plus a slab ocean be expressive enough?
More generally, I'm curious how people evaluate whether or not their ensembles have a "reasonable" amount of variation. For example, I know I can use pertlim and pertseed in user_nl_cam to tweak the initial temperature fields at simulation startup and get variation that way. Presumably these perturbed simulations need to run for some time for the perturbations to result in different enough weather. Any advice on how to know that I've got large enough perturbations and long enough settling times?
Best,
-Rob