About this Event
2026 SGEAS Seminar Series – School of Geography, Earth and Atmospheric Sciences, University of Melbourne. Free and hybrid: join us in person or on Zoom.
GCMagicc is a new generative climate emulator, useful for multivariate climate impact analysis. It uses an hierarchical machine-learning generator trained on 32 CMIP6 Earth system models and ERA5 reanalysis. The generator is trained with engression, a distributional-learning method built on the energy score. It produces physically coherent monthly fields of ten climate variables at about 1° resolution, including every input needed for Penman–Monteith evaporation and drought indices.
The seminar has three parts
- Context – Dr Zebedee Nicholls. Why IPCC assessments and impact modellers rely on emulators, and what the CMIP7 process and its new scenario set change.
- Main talk – Prof Nicolai Meinshausen. How GCMagicc works: a statistical and machine-learning view of turning a global temperature path into whole distributions of regional, multivariate climate.
- Application – Prof Malte Meinshausen. Did human influence drive Iran’s severe 2025 drought? GCMagicc puts essentially zero chance of the current drought conditions in a natural-only world. The talk also covers what current-policy and Paris-pledge pathways mean for droughts around the globe ahead.
Followed by Q&A and discussion.
You can explore 5.6 TB of open ensembles at GCMagicc.org. Preprint (under review at GMD): egusphere-2026-5176
Speakers
Prof Nicolai Meinshausen, University of Washington, Seattle (Honorary) and University of Melbourne (Honorary). Statistician working on high-dimensional statistics, causal inference and machine learning. He developed stability selection, quantile regression forests, invariant causal prediction and engression, the generative method at the core of GCMagicc. Lead developer of GCMagicc. He was previously Professor of Statistics at ETH Zürich.
Dr Zebedee Nicholls, University of Melbourne. ARC DECRA Fellow from 1 December 2026, and a member of the WCRP CMIP forcing panel that prepares the forcing datasets for CMIP7. In IPCC AR6 he led Cross-Chapter Box 7.1 on reduced-complexity models and contributed to Working Groups I and III. Lead developer of MAGICC, the climate emulator also used for IPCC scenario assessments (live.magicc.org). He founded the Reduced Complexity Model Intercomparison Project (RCMIP).
Prof Malte Meinshausen, University of Melbourne. He was an IPCC AR6 WG I Lead Author and a Core Writing Team member for the AR6 Synthesis Report, and he produced the CMIP6 greenhouse-gas concentration projections. Co-developer of MAGICC. Former advisor to the German Ministry of Environment and part of German and EU UNFCCC climate negotiation delegations for over 10 years.
ZN and MM are also affiliated with Climate Resource.
Registration – in person or on Zoom
Please choose a ticket: an in-person seat (limited to the room capacity) or a Zoom place. Zoom link: tinyurl.com/GCMagicc-seminar (Meeting ID 810 9382 1616), also included in your confirmation email.
Getting there
Skeats Room (Room 209), Level 2, McCoy Building (Building 200), Parkville. The McCoy Building is the one with the footpath bridge and ramp on the corner of Swanston and Elgin Streets. Come into the foyer, make your way around the climate globe to the left, and you'll find the 209 seminar room.
The session may be recorded (speakers and slides only). Enquiries: [email protected]
Event venue & nearby stays
Skeats Room 209, McCoy Building, University of Melbourne, 253-283 Elgin Street, Carlton, Australia
Concerts, fests, parties, meetups — all the happenings, one place.