BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Machine learning seminar series - Causal Networks as a framework f
 or climate science to improve process understanding
DTSTART:20201027T160000Z
DTEND:20201027T170000Z
DTSTAMP:20260415T002400Z
UID:indico-event-196@events.ecmwf.int
CONTACT:events@ecmwf.int
DESCRIPTION:\n\n\n\n\n\nHost\n\nInna Polichtchouk (ECMWF)\n\nSpeaker\n\nMa
 rlene Kretschmer is a post doctorate researcher at the University of Readi
 ng. Before that she worked at the Potsdam Institute for Climate Impact Res
 earch in Germany where she received her PhD in climate physics. Her resear
 ch focuses on the dynamical Stratosphere-Troposphere coupling and its impa
 cts for winter circulation and especially for extreme weather events. To a
 ddress these issues\, she is particularly interested in applying novel sta
 tistical approaches from machine learning such as causal discovery algori
 thms.  Moreover\, she is keen on applying these new techniques to evaluat
 e teleconnection processes in climate models and to improve sub-seasonal t
 o seasonal (S2S) forecasts.\n\nAbstract\n\nIn the light of ongoing anthrop
 ogenic climate change and associated risks\, supporting regional decision 
 making should be a guiding principle of climate research. However\, season
 al forecast models only have low skill and climate models often give incon
 clusive results about regional aspects of climate change. One major source
  of uncertainty are dynamical drivers in the climate system\, such as stor
 m tracks or blocking\, which are not well understood theoretically and whe
 re models show diverse responses.\n\nThe recent hype of machine learning p
 romises data-driven solutions to these issues. While data-centric methods 
 such as deep learning have and certainly will make notable contributions t
 o the earth sciences\, their power lies in their ability to efficiently de
 scribe complex relationships present in the data. There is reason to doubt
  whether these methods can\, on their own\, deal with the sort of epistemi
 c uncertainty described above. Moreover\, machine learners and climate sci
 entists often lack a common language\, making successful collaboration sti
 ll difficult. In particular\, climate scientists are trained to think in t
 erms of causal relationships\, whereas machine learning is mostly descript
 ive (i.e. correlational) and does not explicitly incorporate domain knowle
 dge.\n\nHere we call for the use of causal networks in climate science as 
 a framework to overcome some of these challenges. We argue that causal net
 works are a simple yet powerful tool to translate qualitative expert knowl
 edge about physical processes into mathematical objects\, to gain quantita
 tive information about the role of these processes through applying the ru
 les of causal inference.\n\n\n\n\n\nhttps://events.ecmwf.int/event/196/
IMAGE;VALUE=URI:https://events.ecmwf.int/event/196/logo-528950275.png
LOCATION:16:00 GMT
URL:https://events.ecmwf.int/event/196/
END:VEVENT
END:VCALENDAR