Atmosphere WG Meeting 23.07.2025 minutes**
23 participants on zoom
1. Updates from the co-chairs:
-
Charmaine Franklin (BoM) :
- Detail about 2025 BoM workshop which encompasses the complete warning chain from observations, modelling and forecast to communications and response.
- The workshop will include lots of attendees and presenters from outside the meteorological community: Emergency agencies, social science etc.
-
Yi Huang (Uni Melb) :
- 21stCenturyWeather Modelling Science update. Our flagship experimental configuration is progressing well with good peformance against AUS2200.
- Other members are using rAM3 to invstigate cloud physics propoerties and biases.
- A new postdoc (Nasim) has started at Monash and is comparing early flagship runs against Cape Grim campaign data.
- Addition from Bethan : we are moving forward with rAM3 with customised executables built with modified source.
-
Heidi Nettelbeck (ACCESS-NRI) :
- Marcus has joined the community representing a CORDEX-Australasian project focusing on sub-hourly extreme precipitation.
- They are looking initially at case studies in Australian and NZ cities.
- The next phase will then look at limited climate runs.
- They are starting with ACCESS and working with Emma Howard at BoM.
- If 21stCenturyWeather want to participate let them know.
- Data sharing will be a challenge. Could NRI support data hosting, interpretation and sharing of results?
2. New introductions:
- Vinod Kumer - works with BoM in Charmaine’s team. Has recently been working on sub-km scales and working with LFRic.
- Luke Burgess - PhD student at UniMelb working with Andrew Dowdy on PyroCb.
3. Science talks.
3.1 Charmaine - regional LFRric simulations. A BoM team presentation including Belinda, Vinod and Ian.
- RAL3-LFRic assessment. BoM is testing regional LFRic with UK Met Office over the Australian region. A quick review of LRFic/UM differences (e.g. Psyclone, finite element geometry). GungHo is the new dynamical core of LFRic.
- This project finishes in September. The aim is deliver a community research suite.
- RAL4-LFRic will be used to replace all BoM regional modelling suites (including forecasts and reanalysis)
- We are comparing RAL3-LFRic to RAL3-UM.
- There will be no RAL4-UM!
- Case study of TC Tiffany from 2022. Examination of radar reflectivity
- Diurnally forced convection (DFC) study form 2023. Some differences - LFRic generally rains less.
- TC shows similar rain totals and rate rates
- DFC case - LFRic reduced rain total by < 50%. A smaller diurnal cycle and different rain rate distribution with fewer occurrences of < 100 mm/hr. A reduction in light/medium rain rates in LFRic.
- Cloud cover differences. LFRic has less medium and high cloud, particularly over the ocean and along the coast.
- High cloud initially spins down in LFRic. Medium cloud spins down in LFRic after T+3.
- UM has more and stronger updrafts c/f LFRic. Largest differences occurs in upper troposphere for TC case and mid-upper troposphere for DFC.
- Looking at convective outflows and waves. LFRic often has waves at boundaries, and waves propagating ahead of convection at all heights.
- These convective outflows look to be inhibiting convection initiation, and could be another cause of rainfall differences over the ocean and coastal region as convection propogates off-shore.
- Melbourne case studies have examined fog, extreme cold and heatwaves.
- For Jan 2020 heatwave example, LFRric delayed a cool change c/f UM, resulting in large Tmax discrepancy.
- Fog case study of December 2023. LFRic has slightly more fog but overall is quite similar. Similar visibility histograms.
- Mid-latitude precip rainfall amounts are similar, LFRic produces more snow. LFRic has lower cloud amounts.
- In the winter extreme case study of June 2021, LFRic accumulated more snow, especially in the first 24 hours, but less rainfall than the UM. Higher max lightning flashes in LFRic, but larger lightning mean in the UM.
- Next steps :
- Further case studies.
- Verification and evaluation against observations.
- RAL3-LFRic RAS and RNS are already ported to NCI.
- Testing RAL3-LFRic for ACCESS-C/CE and A/AE domains.
- Q&A:
- Q : Paul Gregory. Where are the differences coming from?
- A: Ian/Charmaine. Science is unchanged. All differences come from discretisation changes (new finite element mesh) and new dynamical core (i.e the underlying numerical methods).
- Q : Peter Stienle. LFRic seems more ‘organised’ in convection than UM. Is it an issue of small and large scale noise? Do we know which one is better yet?
- A : Charmaine. LFRic is less diffusive which is good. But the reduction in rain rates look worse.
- Q : Peter - peaks in high cloud seem to have been moved by a few hours?
- A : Charmaine. How fields are initialised and ice clouds needs further examination. These factors changes the radiative fluxes which could impact cloud spin-up. Could be a technical problem, not a model problem.
- Q : Peter - initialised from same initial conditions?
- A : All initialised from UM. We are not yet up to nesting regional LFRic in global LFRic.
3.2 Andrew Brown - coastal wind / sea breeze climatology.
- Sea breezes can relate to convection initiation, coastal wind energy resources, and human health in urban areas
- Previous studies have used point-based observations or hi-rest numerical model data to produce climatologies
- Exploring a diagnostic which can be applied across all model data for the Australian coastline.
- We are going to process data from 2.2 to 25 km. Hi-res data requires some smoothing to remove small scale fronts and gradients that aren’t sea breezes.
- Compute three diagnostics used in previous studies.
- The diagnostics then create a mask based on a threshold to find candidates for sea breeze objects.
- Diagnostic #1 - Sea Breeze Index (SBI) - circulation diagnostic which evaluates 3d winds over land and water.
- #2 - Horizontal kinematic frontogenesis (F) - looking at gradients in 2d surface data.
- #3 Hourly rates of change - analysing timeseries changes of temperature for example.
- Case study in Perth compares three indices w/AUS2200 and ERA5. The H diagnostic doesn’t show up in ERA5, the other two are similar.
- Challenge to remove spurious ‘F’ in inland convection.
- Spatial counts b/w SBI an F are similar. H tends to count objects further inland.
- Challenges in object detection remain, e.g. removing topographic flows, cold fronts.
- How to verifiy? Against what kinds of observations?
- SBI is computationally difficult as it requires model level data.
- As you resolve more in high resolution models, things get messy!
- Want to apply F to BARRA-C for more climatological analysis. Explore relationship b/w sea breezes and SSTs. Might investigate further w/ rAM3.
4. ACCESS 2025 Workshop - Heidi Nettelbeck
- Reminder that ACCESS 2025 workshop registration closes on Aug 15. Please register!
- Annual working group day is September 12. What do we want to discuss? What agenda items?
- Have to finalise the program over the next few days.
- LFRic questions from Paul and Chermelle, resources/runtime, driving with different ICs
- Suggestion from Mat Lipson re: collaborating on STASH outputs?
- Please post further suggestions to forum, preferably a new topic in the Atmosphere WG forum.
5. ACCESS-NRI Visualisation Project - Call for EOI - Heidi Nettelbeck
- Call for expressions of interest. We have a visualisation expert (Owen). Let us know if you want to be involved in a 3-month project. EOI closes in next week 30.07.!
6. Compute and storage - Heidi Nettelbeck
- We have 875 kSU until end of September. See here for how to apply for Atmosphere WG Resources.
7. Paul Gregory:
- Request for any existing AUS2200 input data for a small case study.
EDIT : No longer required. Paul has AUS2200 up and running!