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:20230831T095745Z
LOCATION:Davos
DTSTART;TZID=Europe/Stockholm:20230626T112000
DTEND;TZID=Europe/Stockholm:20230626T115000
UID:submissions.pasc-conference.org_PASC23_sess104_pos114@linklings.com
SUMMARY:P15 - Compressing Multidimensional Weather and Climate Data into N
 eural Networks
DESCRIPTION:Poster\n\nLangwen Huang and Torsten Hoefler (ETH Zurich)\n\nWe
 ather and climate simulations produce petabytes of high-resolution data th
 at are later analyzed by researchers in order to understand climate change
  or severe weather. We propose a new method of compressing this multidimen
 sional weather and climate data: a coordinate-based neural network is trai
 ned to overfit the data, and the resulting parameters are taken as a compa
 ct representation of the original grid-based data. While compression ratio
 s range from 300x to more than 3,000x, our method outperforms the state-of
 -the-art compressor SZ3 in terms of weighted RMSE, MAE. It can faithfully 
 preserve important large scale atmosphere structures and does not introduc
 e artifacts. When using the resulting neural network as a 790x compressed 
 dataloader to train the WeatherBench forecasting model, its RMSE increases
  by less than 2%. The three orders of magnitude compression democratizes a
 ccess to high-resolution climate data and enables numerous new research di
 rections.\n\nSession Chair: Elaine M. Raybourn (Sandia National Laboratori
 es)
END:VEVENT
END:VCALENDAR