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_pos138@linklings.com
SUMMARY:P08 - Analysis and Application of CNN to Improve Deterministic Opt
 ical Flow Nowcasting at DWD
DESCRIPTION:Poster\n\nUlrich Friedrich (DWD)\n\nOptical flow based nowcast
 ing is essential for several operational productions at DWD, including tim
 e critical warnings. Precipitation and radar reflectivity nowcasts are pro
 duced every 5 minutes with a 5 minute stepping up to 2h lead time. The met
 hod assumes stationarity of the input data. It is a deterministic advectio
 n scheme without dynamic properties and does not take advantage of additio
 nal data sources.Recently, machine learning techniques were tested in nowc
 asting. Deterministic methods struggle to predict high-intensity values an
 d become blurry for larger lead times. In this presentation we explore the
  potential of deterministic convolutional neural networks (CNN) to improve
  the operational nowcasting at DWD. A two-year dataset consisting of radar
 , NWP and orography data is used for training modified UNet based neural n
 etworks. The goals are to understand the technically limitations of the ap
 proach as well as the impact of the additional input data. Clever data man
 ipulation and adaption of the network architecture to its properties are k
 ey. An impact study for the input data is performed. We explore combinatio
 n methods for several data encoders, additional computation blocks for mor
 e nonlinearity and loss functions with spatial context. Baselines for comp
 arison are the operational nowcasting at DWD and CNN approaches from liter
 ature.\n\nSession Chair: Elaine M. Raybourn (Sandia National Laboratories)
END:VEVENT
END:VCALENDAR