BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Rethinking Weather and Climate Model Parametrisations with Reinfor
 cement Learning - Pritthijit Nath (University of Cambridge)
DTSTART:20260421T140000Z
DTEND:20260421T150000Z
UID:TALK246067@talks.cam.ac.uk
CONTACT:Kerstin Enright
DESCRIPTION:Weather and climate models rely on parametrisation schemes to 
 represent sub-grid processes that cannot be explicitly resolved on the com
 putational grid. Many such traditional schemes depend on fixed coefficient
 s that are only weakly constrained and tuned offline\, often locking in pe
 rsistent biases and limiting adaptability across regimes\, resolutions\, a
 nd climates. \n\nIn this talk\, I present a strategy that reframes part of
  parametrisation design as a sequential control problem by embedding a rei
 nforcement learning (RL) agent within the model\, allowing it to observe t
 he evolving state and update selected tunable components online during int
 egration. We evaluate this approach across a hierarchy of idealised enviro
 nments\, from a simple single-parameter bias-correction setting to multi-p
 arameter zonal energy balance models (EBMs)\, exploring both single-agent 
 and federated multi-agent configurations\, the latter mirroring the spatia
 l decomposition used in general circulation models. \n\nAcross these setti
 ngs\, we find RL-assisted parameter updates consistently reduce area-weigh
 ted RMSE relative to static tuning\, with the largest gains emerging in tr
 opical and mid-latitude bands\, while federated training accelerates conve
 rgence and enables geographically specialised control without sacrificing 
 physically meaningful parameter adjustments. Overall\, results from these 
 idealised setups suggest that RL provides a viable pathway toward regime-a
 ware\, state-dependent parametrisations and a scalable framework for onlin
 e learning within numerical weather and climate models.\n\n\n*NOTE - ZOOM 
 ONLINE TALK*\nMore information including the Zoom link can be found on the
  "ICCS website":https://iccs.cam.ac.uk/events/journal-club-rethinking-weat
 her-and-climate-model-parametrisations-reinforcement-learning\n\n
LOCATION:Zoom (Online only)
END:VEVENT
END:VCALENDAR