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:20230831T095747Z
LOCATION:Sertig
DTSTART;TZID=Europe/Stockholm:20230628T150000
DTEND;TZID=Europe/Stockholm:20230628T153000
UID:submissions.pasc-conference.org_PASC23_sess109_msa136@linklings.com
SUMMARY:Data-Driven Cloud Cover Parameterizations for the ICON Model
DESCRIPTION:Minisymposium\n\nArthur Grundner (DLR, Columbia University); T
 om Beucler (University of Lausanne); Pierre Gentine (Columbia University);
  Marco Giorgetta (Max Planck Institute for Meteorology); Fernando Iglesias
 -Suarez and Rémi Kazeroni (DLR); and Veronika Eyring (DLR, University of B
 remen)\n\nA promising approach to improving cloud parameterizations in cli
 mate models, and thus climate projections, is to train machine learning (M
 L) algorithms on coarse-grained output of global storm-resolving model sim
 ulations. The ICOsahedral Non-hydrostatic (ICON) modeling framework enable
 s simulations ranging from weather prediction to climate projections, maki
 ng it an ideal target for developing ML-based parameterizations. We develo
 p four different neural network (NN) types, distinguished by the degree of
  vertical locality they assume for diagnosing cloud cover. We enforce spar
 sity by sequentially selecting features based on the models' performance g
 ains for the NN type that offers the best tradeoff between complexity and 
 accuracy. In their training domain, the NNs already achieve excellent pred
 ictive performance (R^2 >0.9) with as few as three features. Using symboli
 c regression, we discover an interpretable equation with superior performa
 nce (R^2 =0.94) that satisfies physical constraints. These ML-based cloud 
 cover parameterizations are coupled to the atmospheric component of the IC
 ON Earth system model via the Fortran-Keras Bridge. To provide the most ap
 propriate measure of cloudiness for ICON's microphysics scheme, the NNs pr
 ovide cloud volume fraction in addition to cloud area fraction in each gri
 d cell. We evaluate the resulting coupled ICON-ML model using the Earth Sy
 stem Model Evaluation Tool (ESMValTool).\n\nDomain: Climate, Weather and E
 arth Sciences\n\nSession Chair: Tobias Weigel (German Climate Computing Ce
 ntre, DKRZ)
END:VEVENT
END:VCALENDAR