Workshop Program for Connecting Machine Learning to Earth System Science 2026

The ACCESS Community Workshop Connecting Machine Learning to Earth System Science runs across three days from Wednesday 19 to Friday 21 August. It contains a mixture of plenary sessions that occur in a single location, parallel sessions, training and informal discussion sessions.

The program is subject to amendments—please revisit this webpage for the most up-to-date version of the program.

On this page:

Link to access the live streaming of all sessions, with the exception of training, group discussions and parallel sessions.

Day 1: Wednesday 19 August

TimeSessionSession ChairPresenter Title (and link to abstract)
9:00 – 9:30Arrival Registration and tea/coffee
9:30 – 10:00Welcome to Country and IntroductionVassili KitsiosWurundjeri Elder Jason Tweedy

Andy Hogg
10:00 – 10:30Session 1: Plenary (1 speaker)Sanaa HobeichiTerry O'KaneNovel neural operator and transformer architectures for climate prediction
10:30 – 11:00Morning Tea
11:00 – 11:30Session 2: Evaluation methods (2 speakers)Yiyi GuoHung LuuCan we test machine learning models without ground truth?
Yuan Zhuang*Evaluating Geo-Foundation Model Embeddings for Wildfire Risk Assessment
11:30–12:00Lightning Talks (2 min per speaker including change over)Will Hobbs (#1)Data-driven polar research: M/L activities at the Australian Antarctic Program Partnership
Chandana Nagenahally Manjunath* (#2)Dynamic Network Analysis of ENSO-driven Global Commodity Price Volatility: Implications for Food and Energy Security
Wei Ji Leong (#3)CuPy-Xarray: towards direct to GPU decoding of raster file formats
Christian Stassen (#4)Machine learning for regional climate downscaling in the Australian Climate Service
Abhnil Amtesh Prasad (#5)WRF-LES training data generation framework for AI surrogates of parameterized convection
Pearse Buchanan (#6)Optimization of Australia’s ocean biogeochemical model with a machine learning surrogate.
Lynn Zhou (#7)Rapid Forecast‑Based Attribution via a Climate Emulator
Haoran Li, Shixue Li (#8)Coarse-to-Fine: Multi-source Fusion for Bias-Corrected Precipitation Downscaling
Jakob Gradl* (#9)A hybrid variational physics informed/physics encoded neural network for strictly mass-conserving joint inversion of mass and momentum balance in ice flow modelling
Irina Rudeva (#10)When Observations Are Sparse: Shared Reanalysis Biases in Southern Ocean MSLP and Insights from Machine Learning
Peter Oke (#11)When Machine Learning Breaks the Ocean: Why Dynamics Still Matter
Taimoor Sohail (#12)Towards a Machine Learning Ocean Emulation Framework for Decadal Climate Prediction
12:00 - 13:30Poster viewing (Dedicated first 30 minutes, with last hour overlapping with lunch)
12:30 - 13:30Lunch
13:30 – 15:00Training Session: Autoencoders Part 1 - IntroductionTennessee LeeuwenburgAuto-encoders training session Part 1
15:00 - 15:30Afternoon Tea
15:30 - 16:15Session 3: Physics-Informed / Hybrid Machine Learning Part I (3 speakers)Micael OliveiraLachlan AstfalckFlowGP: Physically Coherent ML for Probabilistic Field Reconstruction
David LeeData driven thermodynamically consistent moist phase exchanges for atmospheric convection
David FuchsPlugging Pytorch to an atmospheric model the TorchClim way
16:15 - 17:00Breakout session: Putting GenAI to Work in Earth System Science Part 1Yue Sun, Micael Oliveira, Edward Yang, Taimoor Sohail & Sanaa HobeichiGenAI Breakout session part 1
17:00 - 18:00Free time
18:00 - onwardsDinner. Venue: Henry and the Fox (Fox's Den area)
Address: 525 Little Collins Street, Melbourne
Time: Arrival 6pm, Service 6.30pm
Food: Vegan

Day 2: Thursday 20 August

TimeSessionSession ChairPresenterTitle (with link to Abstract)
8:30 – 8:50Arrival
8:50 – 9:00Housekeeping and Acknowledgement of Country
9:00 – 10:30Session 4: Forecasting, nowcasting & data assimilation (6 speakers)Steve SherwoodYiyi Guo*Distributional Bias Correction for Madden-Julian Oscillation Forecast
Carl Doedens*Solar nowcasting with latent diffusion and XGBoost: assessing feature importance and evaluation in different environments
Chen LiAssessment of Global Data-Driven Models for Weather and Sub‑Seasonal Prediction over Australia
Husnain Asif*Shared Latent Representations for Kilometer-Scale Multivariate Climate Fields
Ajitha CyriacValidating satellite-derived data for sea surface temperature downscaling using machine learning
Peter OkeA data-driven approach for mesoscale ocean forecasting
10:30 – 11:00Morning Tea
11:00 – 11:45Session 5: Atmosphere applications (3 speakers)Heidi NettelbeckEun-Pa LimOzone – an untapped source of subseasonal to seasonal forecast predictability
Hongyan ZhuTropical Cycle simulations: Machine learning versus physical models
Harun RashidAssessing Australian rainfall teleconnections using linear and machine learning models
11:45 – 12:05Lightning Talks (2 min per speaker including change over)Spencer Patrick Clark* (#13)Pattern scaling in an overshoot world
Kevin Horner / Daimon Byl / Benjamin Cottrell * (#14)Practical Gray-Box Universal Differential Equation Methods for Chaotic Climate-Relevant Dynamics
Tishampati Dhar (#15)A review of Geospatial Foundation models, Embeddings and Practical Applications
Li Wang (#16)AI-Assisted Porting and Performance Optimization of LFRic Kernels
Ryan Holmes (#17)Coastal downscaling of Australian seasonal sea level and sea surface temperature forecasts using neural networks
Jasmine Peterson* (#18)Machine Learning and Australia’s shape
Krish Singh* (#19)Using Machine learning to calibrate complex process-based models for carbon stock projection in NSW forests
Zhihao Deng* (#20)Climatological Anomalies Projection using Neural Ordinary Differential Equations
William Edge (#21)Gaussian Process Regression for multi-scale oceanographic observations
12:05 - 13:30Poster viewing (Dedicated first 25 minutes, with last hour overlapping with lunch)
12:30 - 13:30Lunch
Potential for community led breakout sessions run in parallel during lunch
13:30 – 15:00Training Session: Auto-encoders Part 2Tennessee LeeuwenburgAuto-encoder training session 2
15:00 - 15:30Afternoon Tea
15:30 - 16:15Session 6: Ocean applications (3 speakers)

Taimoor SohailRick de Kreij*Statistical inversion of surface tracers to infer fine-scale near-surface ocean currents
Stephanie ContardoA machine-learning and process-based hybrid workflow for shoreline position prediction
Chaojiao SunDeep learning projections of coral bleaching risk: will the 2024–25 Ningaloo marine heatwave become the new normal?
In Parallel: Atmosphere Working Group Meeting
16:15 - 17:00Breakout session: Putting GenAI to Work in Earth System Science Part 2Yue Sun, Micael Oliveira, Edward Yang, Taimoor Sohail & Sanaa HobeichiGenAI Breakout Session Part 2

Day 3: Friday 21 August

TimeSessionChairPresenterTitle
8:30 – 8:50Arrival
8:50 – 9:00Housekeeping and Acknowledgement of Country
9:00 – 10:30Session 7: Infrastructure & Physics-informed / Hybrid Machine Learning Part II(6 speakers)Paul LeopardiMengmeng Han
Infusing deep-learning model with domain knowledge for better surface wind nowcasting
Samuel GreenTRACE: Transformer Reanalysis Atmospheric Compression Engine for MERRA2
Yue SunScaling AI for Earth Science: Deployment, Extension and Evaluation of Zephyrus on Gadi
Maruf AhmedEvaluation of NVIDIA H200 GPUs for training Large-scale AI/ML models at NCI
Andrew Zammit-MangionPhysics-based machine learning models for spatio-temporal forecasting
Vassili KitsiosPhysics-constrained climate emulator of the coupled global atmosphere, land, ocean and sea-ice Earth system
10:30 – 11:00Morning Tea
11:00 – 12:30Training Session: Neural Ordinary Differential EquationsAndrew OoiNeural ODEs training session
12:30 – 14:00Lunch
13:00 – 14:00Breakout session: Physics-informed / Hybrid Machine Learning Steve SherwoodPhysics informed hybrid machine learning
14:00 - 15:00Session 8: High Resolution & Land Applications (4 speakers)Edward YangHuidong (Warren) JinMachine Learning for High-Resolution Long-Range Daily Rainfall Forecasts from ACCESS-S2
Reza NosratpourAnalysis of Extreme Precipitation Pathways using Climate Networks and Tail Dependence Coefficient
Steve PetrieMapping carbon uptake with high accuracy across different agricultural land cover types
Ying-Ping WangApplications of machine learning to improve global soil carbon predictions for Earth system models
15:00 - 15:30Closing