heidi
(Heidi Nettelbeck, ACCESS-NRI Atmosphere Modelling Team Lead)
1
Notes from each meeting of the Atmosphere Working Group held in 2025 are added as a reply to this topic.
Make each reply a wiki post , that way other attendees can edit the notes. This can greatly reduce the burden of note-taking allowing that person to participate fully in discussions without worrying too much about missing important details.
Atmosphere WG Meeting 12.02.2025 minutes
12 Feb 2025
26 participants on Zoom
From: UNSW, BoM, Monash, ACCESS-NRI, CSIRO, UTas
Science talk – Free-running tests with the NRI-version of the Regional Nesting Suite by Matt Lipson
Mat mentioned that the talk would be more technical than scientific talk due to where we are at. Using Spare SU last year, Mat pulled together ancillaries and ran nested 1km simulations centered over Sydney.
Weather resources (off-shore wind energy generation)
Heat/breeze impacts (Negin)
Heavy Precipitation (Jason Evans) compared to his WRF runs. Domains were chosen to align with WRF runs.
Different setup to ACCESS-NRI Lismore tutorial.
Added an intermediate step at 5km.
Chermelle managed to get daily varying SST data working. Daily varying SST was thus included in the test. Mat found it easy to use.
Free-running, no cycling. Runs for 1 month.
Tested the two options: ERA5-land and BARRA-R2 land/surface initial conditions (ICs) in the ACCESS-NRI suite.
Results:
Mat compared how variables evolved when intialised with Era5-land or Barra-R2
Projections were compared to the BARRA-R2 reanalysis
Soil moisture: ERA5-land initial soil moisture wetter than BARRA-R2.
Projection initialised with BARRA-R2 remained drier than that initialised with ERA5-land (at 12km)
Within 1km inner nest, over the entire simulation period the soil moistures (top-layer) aligned with each other.
Latent Heat flux: More moisture going into the atmosphere from wetter ERA5-land IC than BARRA-R2 IC.
Air temperature: For 12km domain, all simulations tracked closely to observations (256 sites throughout QLD/NSW/TAS, but BARRA-R2 land/surface initialised product aligned better with observations). Similarly for 12km/5km/1km.
Animation of air-temp through the month at 1km
Urban effects: Urban heat island in the evening.
Sea-breeze effects.
Timing of things like change in cold fronts was ok. Sometimes an hour late.
Domain averaged air temperature: AWS locations were shown as coloured dots. The color difference between the AWS dot and the background indicate obs/projection differences.
Overall, from diurnal profile, difference between BARRA was 0.5 due to slightly drier initialisation.
Not clear which land/surface initialised run was better. Expected that BARRA land/surface ICs would perfom better. Need to do more testing.
Andrew Brown will analyse edge-effects.
Concluding remarks: Thinking about reducing size of outer domain or removing 5km domain. Suite is easy to run; hardest part is setting up STASH. There is no automatic conversion to netcdf, but there are ACCESS-NRI tools available (e.g., um2nc) which could work for the regional suite. Mat and Chermelle are talking about a GCM-driven nesting suite.
For details about setup see this GitHub page.
Questions/Answers
Qinggang: How much grid of the boundary do you remove? In previous work he removed 10% of the grid.
Mat: not removing any yet, but that will change.
Claire: interested in the lower boundary condition of SST versus land. The land boundary was initialised at the start and was left to run. Was the SST boundary changed every 6 hours?
Mat: the SST changed every day.
Claire: Did it change with a jump?
Mat: Yes, and the diurnal change during the day is missing.
Peter: is there any nudging going on?
Mat: no.
Peter: Could be interesting to further investigate for the energy fluxes at the surface and investigate the energy balance.
Chermelle: Surprised at how close the projects where when initialised with ERA5-land versus BARRA-R2. Is anyone else?
Mat: yes.
Peter: Mat probably needs to look in more detail.
Mat: yes
Tech talk – An introduction to ACCESS-Vis: Visualisation tools released to date and current work in progress by Owen Kaluza
Yi introduced Owen: “If you want to make your data come to life, Owen is a person who can show you the tricks”.
Owen is a senior Research Software Engineer working in the MedTeam at ACCESS-NRI, specialising in visualisation. His aim is to produce showpiece visualisations for ACCESS-NRI.
He tends to work in python and uses tools that researchers use; focus on open-source tools.
Also uses GPU acceleration.
There are three repositories:
ACCESS-Vis, which is available on Github and on the ACCESS-MED conda environment conda/access-med-0.12
ACCESS-Vis provides a canvas to render earth system data; based on the LavaVu python visualisation library.
Use on ARE with a gpuvolta queue using the conda/access-med-0.12 access-med conda module.
ACCESS-Visualisation-Recipes
A collection of notebooks; showed simple concepts of the library,. E.g., position of the sun for proper rendering on the blue marble (rather than having a light shining on a default position in the animation).
When you import the library you can select the resolution you want to run it at. You can start at a lower resolution when setting up then switch to a higher resolution for higher quality animations. There are examples for other options.
Can create images in a Jupyter Notebook; examples are given in the repository. Can also look at the source code to get better understanding.
Monthly textures available from Blue Marble, but also possible to use an interpolation between different months. Other notebooks demonstrate other features.
Owen plans to release more examples for visualisation and is open to feedback or feature requests.
AUS2200 visualisations.
The images are designed to look like satellite images.
A particle tracer is available that drops particles and advects them. Owen showed an example. The particles have random lifetime and are replaced when they go out of the boundary.
Owen also showed an example of clouds with wind vectors from AUS2200. There were jumps due to re-intialisations. Owen asked for other datasets or suggestions of data to plot.
Questions/Answers:
Mat: Are we able to access the notebook and run ourselves?
Owen: The last plot “will” be, it has not been added yet.
Charmaine: It is complicated to plot the 3D wind vectors? She gave a link to an example of work she was doing with Drew Whitehouse to produce visualisations of cloud-rain and 3D winds. She was happy to help.
Qinggang: when you position the sun right does involve some assumption that when the sun is in the domain is gets light?
Owen: we are positioning the sunlight so it looks like it does in satellite images. It will affect light falling on an image that is being produced, but it’s not a correct physical model.
Qinggang: At midnight can you still see data?
Owen: Yes, and you can adjust the ambient lighting and diffusion, etc.
2025 ACCESS Workshop:Heidi Nettelbeck (5-10mins)
This year the workshop will be held in Melbourne in early September. The workshop committee would like to schedule parallel science sessions. The Working Group would like to hear from the Atmosphere community what those sessions could look like. Any ideas?
You can also post comments on the forum.
Comments:
Peter: having single sesssions help sharing of ideas.
Heidi: Parallel sessions could focus on things relevant across other WGs (e.g., Atmosphere, Ocean, Land, etc.).
Charmaine: one thing to think about is a regional coupled model. We could do a joint session to discuss needs from both sides.
Yi: Interactions are important with land as well.
Heidi: any other thoughts? Like Owen’s talk, maybe a cross-tools session?
Yi: There are also new users working on atmospheric chemistry.