BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20241120T082409Z
LOCATION:HG F 1
DTSTART;TZID=Europe/Stockholm:20240603T153000
DTEND;TZID=Europe/Stockholm:20240603T160000
UID:submissions.pasc-conference.org_PASC24_sess155_msa237@linklings.com
SUMMARY:The AI2 Climate Emulator (ACE): A fast, Skillful Learned Global At
 mospheric Model for Climate Prediction
DESCRIPTION:Minisymposium\n\nJeremy McGibbon, Spencer K. Clark, Gideon Dre
 sdner, James Duncan, Brian Henn, and Oliver Watt-Meyer (Allen Institute fo
 r Artificial Intelligence); Boris Bonev, Noah D. Brenowitz, Karthik Kashin
 ath, and Michael S. Pritchard (NVIDIA Inc.); and Matthew E. Peters and Chr
 istopher S. Bretherton (Allen Institute for Artificial Intelligence)\n\nTh
 e AI2 Climate Emulator (ACE) marks a significant leap in climate modeling,
  employing a deep learning framework to replicate the comprehensive dynami
 cs of the FV3GFS atmospheric model efficiently. ACE incorporates a Spheric
 al Fourier Neural Operator (SFNO) with approximately 200M parameters. Usin
 g the previous weather state and externally prescribed forcings, this mode
 l forecasts the atmospheric state 6 hours ahead, alongside diagnostics suc
 h as surface precipitation rate, and turbulent and radiative fluxes. This 
 variable set facilitates a robust assessment of the moisture and dry air m
 ass budgets, and allows us to incorporate constraints to conserve dry air 
 mass and ensure a closed moisture budget. Trained on a dataset with 100 ye
 ars of atmospheric states simulated by a physics-based global atmosphere m
 odel, coarsened to a resolution of 1° and eight vertical levels, ACE demon
 strates the ability to conduct stable multi-decadal simulations that maint
 ain accurate weather dynamics and seasonal cycles, closely mirroring the r
 eference model's precipitation and temperature patterns. Notably, ACE uses
  60 times less energy than the FV3GFS model, leveraging modern GPU technol
 ogy for efficient inference. This study underscores the potential of machi
 ne learning in climate prediction, offering a path towards fast and access
 ible climate models.\n\nDomain: Climate, Weather, and Earth Sciences, Comp
 utational Methods and Applied Mathematics\n\nSession Chair: Karthik Kashin
 ath (NVIDIA Inc., Lawrence Berkeley National Laboratory)
END:VEVENT
END:VCALENDAR