BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Fluid Mechanics (DAMTP)
SUMMARY:AI weather forecasting: past\, present and future 
 - Richard Turner\, University of Cambridge
DTSTART;TZID=Europe/London:20260227T160000
DTEND;TZID=Europe/London:20260227T170000
UID:TALK243397AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/243397
DESCRIPTION:Weather forecasting is critical for a range of hum
 an activities including transportation\, agricultu
 re\, industry\, as well as the safety of the gener
 al public. Over the last two years\, machine learn
 ing models have shown that they have\nthe potentia
 l to transform the complex weather prediction pipe
 line\, but current approaches still rely on numeri
 cal weather prediction (NWP) systems\, limiting fo
 recast speed and accuracy. In this talk\, some of 
 the background on these\ndevelopments will be give
 n. A new generation of machine learning model will
  then be introduced which can replace the entire o
 perational NWP pipeline. Aardvark Weather\, an end
 -to-end data-driven weather prediction system\, in
 gests raw observations and outputs global gridded 
 forecasts and local\nstation forecasts. Further\, 
 it can be optimised end-to-end to maximise perform
 ance over quantities of interest. It will be shown
  that the system outperforms an operational NWP ba
 seline for multiple variables and lead times for g
 ridded and station forecasts. Finally\, the talk w
 ill end by discussing how these ideas might develo
 p over the next few years\, including their applic
 ation to multiple parts of the Earth system on mul
 tiple time-scales\, and their potential impact on 
 climate modelling
LOCATION:MR2
CONTACT:Duncan Hewitt
END:VEVENT
END:VCALENDAR
