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:20220812T074334Z
LOCATION:Sydney Room
DTSTART;TZID=Europe/Stockholm:20220627T170000
DTEND;TZID=Europe/Stockholm:20220627T173000
UID:submissions.pasc-conference.org_PASC22_sess127_msa221@linklings.com
SUMMARY:Statistical Downscaling of Surface Temperature and Precipitation w
 ith Deep Neural Networks
DESCRIPTION:Minisymposium\n\nStatistical Downscaling of Surface Temperatur
 e and Precipitation with Deep Neural Networks\n\nGong, Langguth, Ji, Mozaf
 fari, Mache...\n\nIn light of the success of superresolution (SR) applicat
 ions in computer vision, recent studies have started to develop statistica
 l downscaling methods for meteorological data based on deep neural network
 s (DNNs). DNNs are attractive, because they are computationally cheap, onc
 e they are trained.<br />In this study, deep neural networks are developed
  to downscale hourly 2 meter temperature and precipitation over the comple
 x terrain of Central Europe. Our approach is based on advanced generative 
 adversarial networks (GANs) and transformer networks. The merit of this ch
 oice is that GANs encourage the generator to preserve the strong spatial v
 ariability from the data, while the transformer can capture the temporal d
 ependencies. The experiments are designed to generate high-resolution temp
 erature (0.1°) from low resolution (0.8°), and time-evolving high-resoluti
 on precipitation (1 km) from low resolution (4 km/8 km). The DNNs are fed 
 with several relevant static and dynamic predictors and comprehensively ev
 aluated by grid point-level errors, and error metrics for spatial variabil
 ity and the generated probability distribution. Our results motivate the f
 urther development of DNNs that can be potentially leveraged to downscale 
 other challenging Earth system data such as cloud cover or wind in operati
 onal workflows.\n\nDomain: Climate, Weather and Earth Sciences, Computer S
 cience and Applied Mathematics, Engineering, Physics
END:VEVENT
END:VCALENDAR