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:20260624T171341Z
LOCATION:Bldg. 6 - 002
DTSTART;TZID=Europe/Stockholm:20260701T140000
DTEND;TZID=Europe/Stockholm:20260701T143000
UID:submissions.pasc-conference.org_PASC26_sess122_msa245@linklings.com
SUMMARY:Computational and Verification Challenges in Data-Driven Atmospher
 ic Downscaling
DESCRIPTION:Mary McGlohon and Petar Stamenkovic (MeteoSwiss, ETH Zurich); 
 David Leutwyler, Xavier Lapillonne, and Oliver Fuhrer (MeteoSwiss); Fabian
  Bösch, Lukas Drescher, and Henrique Mendonça (ETH Zurich / CSCS); Sebasti
 an Schemm (University of Cambridge); and Siddhartha Mishra (ETH Zurich)\n\
 nUsing generative machine learning for performing atmospheric downscaling 
 (super-resolution for meteorological data) is of growing interest, as the 
 methods used for data-driven downscaling are computationally inexpensive c
 ompared with statistical downscaling methods or with data-driven models fo
 r forecasting. While previous research in data-driven atmospheric downscal
 ing has shown promising results using standard ML model quality scores, th
 ere are open questions regarding how well these methods perform when it co
 mes to real-world tasks such as climate modeling or severe weather alertin
 g.\nIn this talk, we introduce some challenges and limitations that we enc
 ounter when using ML for atmospheric downscaling. We report about computat
 ional efficiency of different methods (grid-based diffusion vs graph-based
  neural nets) and discuss verification methods for the data-driven downsca
 ling models against forecast models. Finally, we discuss potential use cas
 es and propose a path forward.\n\nDomain: Climate, Weather, and Earth Scie
 nces, Physics, Computational Methods and Applied Mathematics\n\nSession Ch
 air: Oliver Fuhrer (MeteoSwiss, ETH Zurich)\n\n
END:VEVENT
END:VCALENDAR