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:20210916T132456Z
LOCATION:
DTSTART;TZID=Europe/Stockholm:20210706T173000
DTEND;TZID=Europe/Stockholm:20210706T190000
UID:submissions.pasc-conference.org_PASC21_sess182_post143@linklings.com
SUMMARY:P21 - AI for Earth System Sciences: Bottlenecks, Pitfalls and Reco
 mmendations
DESCRIPTION:Poster\n\nP21 - AI for Earth System Sciences: Bottlenecks, Pit
 falls and Recommendations\n\nWeigel, Lüttgau, Kadow, Greenberg, Bouwer...\
 n\nDKRZ and Hereon are prototyping new services to support scientists from
  Earth system sciences who seek to apply AI methods. AI tools promise a nu
 mber of major benefits for Earth system science, but wider uptake and succ
 essful applications require radically new infrastructure and support servi
 ces. This requirement is especially pressing for large scale modelling eff
 orts in HPC environments, such as climate and coastal modelling. This pres
 entation highlights technical and organizational tasks set up in the conte
 xt of the Helmholtz AI project and shares initial insights from service re
 ndering over the past year. Technical tasks include optimizing AI applicat
 ions for HPC and storage, including multi-GPU and multi-GPU-node setups, b
 enchmarking of typical ML workflows on HPC, automating hyperparameter tuni
 ng, and maintaining software environments. Organizational tasks include co
 nceptual guidance in choosing suitable AI methods, understanding feasibili
 ty, limitations and uncertainties, and building and curating a knowledge b
 ase of methods, AI frameworks, and example studies including both successe
 s and failures. An accompanying training and counselling programme builds 
 capacities for applying AI in Earth system sciences and beyond. This prese
 ntation will share first insights and future evolution paths of the protot
 yped service infrastructure and explore opportunities for community collab
 oration.
END:VEVENT
END:VCALENDAR