I would like to gauge how many of us may be interested in using ACCESS models for large eddy simulations (LES).
LES with horizontal grid spacings around 100 m and vertical grid spacings around 10 m could be very useful to investigate individual physical processes, thought in a small domain (e.g. 200km by 200km by 5km in the vertical). It could be used for a range of topics such as air-sea interactions, (urban) boundary layer meterology, and cloud physics. The output could also be used for training machine learning models.
Currently, the highest vertical resolution achievable with ACCESS-rAM3 is 140 levels up to 40 km. I am wondering how much work it would be to adjust the model top to around 5 km and refine the vertical grid spacings to around 10 m. In addition, could we adjust horizontal grid spacings from 0.01 to 0.001 degree (~110 m) directly, or are there additional tweaks we need to consider?
Running the UM at 100m horizontal grid spacing is very common and has been done for more than 15 years for research, with that work supporting the development of operational models at 300m (such as the London Model that has been running since 2016).
The RAL3 science configuration has been tested and used at 100m (in fact down to 50m), you don’t need to change anything. The only limitation with the UM and what grid spacing you can use is that the timestep cannot be less than 1 sec.
You can define your own vertical level set, although reducing the model lid to 5km would be a major piece of work - think about the radiation for a start, and then there’s the wave reflection at the model top that would need to be dealt with.
The UM is not an LES. The Met Office LES is called MONC, and it is not installed at NCI.
Also note that MONC, by design of being an LES, is an idealised model and can’t be used for case study-type simulations. It also has limited physics (CASIM cloud microphysics and Edwards-Slingo radiation).
As Charmaine has said, the UM with RAL3 science has been tested down to high resolution.
I’d be very interested in helping to support LES modelling within ACCESS-NRI. My own opinion is that, for boundary layer and cloud work, LES does provide significant value (how much depends, of course, on the research question). There are obvious feedbacks between LES and "standard” weather models (ie. using LES to improve microphysics/turbulence schemes).
I use the Dutch Atmospheric Large Eddy Model (DALES) . It’s developed largely by cloud physicists so has sophisticated microphysics schemes, lots of chemical transport options (aerosols, smoke, dynmaic CCN, etc…). It’s open source, actively maintained and developed, with some new functionality coming very soon (open boundary conditions, variable SSTs and roughness lengths, dynamic wind turbine models, to name a few).
That said, if ACCESS-NRI were to provide support for an LES, I’d be somewhat ambivalent about which model. There are a lot out there. Given the expertise within the Australian community with the UM, it might make sense to stick with a Met-Office model if the functionality is what the Australian community needs.
Another option is breezy.jl that Madi uses, that’s very promising but not quite ready for research (yet!)
Concretely, things I help support:
Compare and contrast LES output with that from high resolution atmospheric simulations;
Compare DALES with another LES;
Try to wrangle the community to settle on a plan of action (herding cats?)
I’m very willing to help here, but only if there is strong community interest.
Inter-model comparison is appealing, but it is normally difficult to obtain concrete insights. I am quite interested in aerosol-microphysics interations and many individual processes within them. So DALES is quite intriguing for me. On the other hand, the use of UM/Lfric/MONC with RAL3.2+ could potentially hint at global climate simulations with UM and thus more direct implications. For Met Office models, we would definitely need support from ACCESS-NRI, especially for the upper model lid.
I will reflect further on the research questions I may answer with LES. In the meantime, I would be happy to discuss related issues.
I don’t think running the UM with a reduced lid is an option - it’s work that would be well beyond ACCESS-NRI’s resources. It’s not as simple a reducing the model top - lots of the other model physics and dynamics needs addressing too.
I think if you’re thinking of using the UM (not MONC) it would be better to do a very high resolution run, perhaps with a new vertical level set defined but keeping the model top where it is and putting a higher density of levels in the lower atmosphere.
The benefit of that solution is that it would allow you to use the full set of UM physics. From my background of work in aerosol-microphysics interactions, one thing I would strongly encourage in any study would be to use an ensemble of experiments. With full UM physics available, you could use the RP scheme to produce a perturbed parameter ensemble set.
I’m not sure if the RP scheme is available in rAM3, but if it isn’t then it’s a capability I would love to see available to the research community, and would be happy to be on board with any discussions and work around that.