Joint meeting between Forecasting & Prediction, and Atmosphere working groups
8 May 2024
Chaired by Kim Reid
Notes by Matt Woodhouse
27 participants from both WGs, full range of institutes
Debbie Hudson (BoM), F&P WG overview
Intro to F&P working group: data assim, seasonal prediciton, machine learning, for forecasting modelling - activities not currently supported by NRI
Key activities:
Identify how NRI could support DA and ML in the future
Use F&P to bring together community working on ML for modelling and prediction
Scope interest in community usage of ACCESS-S. Started running training on ACCESS-S
Keen to hear feedback comments on how NRI could support ML and AI
You can also provide input/comments by reply on the Hive document, or in track-changes in the google doc: How ACCESS NRI could support ML/AI - Google Docs
Charmaine (BoM), Atmos WG overview
Intro Atmos WG: goal is to establish and support a community, identifying areas for collaboration
Yi Huang (UniMelb), overview of uni research
Matt Woodhouse (CSIRO), overview of CSIRO research
Martin Dix (CSIRO, NRI), Single Column Model
SCM has full UM physics, not UKCA (UKCA require advective tendencies of all the species. Could perhaps be derived from the UM)
runs from rose/cylc like full model
easy to run debugger, fast, netcdf output
ideal platform for developing and testing code
Bethan White (BoM): development example using SCM to test a sea breeeze parameterisation
Charmaine: SCMs have also proven to be really useful for explaining different cloud feedbacks across climate models https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2013MS000246
Calum Knight (UTas)
Introduction to EMC2, an open-source ground-based lidar and radar instrument simulator and subcolumn generator for large scale models
now able to use with ACCESS / UM
Aiming to lodge back into repository so that it will be much easier to use with UM output
Discussion – find collaborations between groups
Craig Bishop – any interest in developing a TL or adjoint model?
Kim – could EMC2 be useful for estimating rainfall over Southern Ocean, where there are sparse observations? Sonya – yes, that’s one of the objectives. Yi – rainfall relatively easy, mixed phase is daunting
EMC2? How easy is it to learn how to use?
Took Calum a few months, creating model class was biggest task.
Bethan – COSP available in UM, very useful package
Julie – how does DA benefit non-direct users of DA? Outside of Bureau there is no capacity for using DA. Running models from observed initial conditions would be valuable
Matt: Newtonian relaxation available in AM2/CM2, and also RNS. Can be used to provided initial conditions to atmospheric simulations
Chun-Hsu has spectral nudging working in UM
Heidi: think about possibilities for sessions around NRI annual meeting in September
Charmaine: don’t forget to register for Momentum workshop