BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Forecasting of Photospheric Magnetograms using deep learning
DTSTART:20260416T053000Z
DTEND:20260416T063000Z
DTSTAMP:20260503T112400Z
UID:indico-event-9324@scitalks.tifr.res.in
DESCRIPTION:Speakers: Hemapriya Raju (DAA\, TIFR)\n\nSpace weather forecas
 ting is essential to protect power and space based infrastructure. Operati
 onal models rely on current nearside photospheric magnetograms to forecast
  the solar flares\, and to understand propagation of Coronal Mass Ejection
 s. In this current work\, we attempted to forecast the photospheric magnet
 ograms for next few days\, that can be used as inputs for space weather pr
 ediction models. We integrated farside observations to assist the model in
  learning the emergence of new active regions. In this work\, we propose a
  U-Net inspired deep learning model that takes a full solar rotation of ma
 gnetograms as input\, along with farside seismic maps\, to forecast full-d
 isk photospheric magnetograms over the next few days. Temporal evolution c
 an be accounted for by treating sequential magnetograms as additional inpu
 t channels\, enabling the network to learn the spatiotemporal evolution of
  the magnetic field. Since farside seismic maps are inherently noisy\, thr
 eshold based saturation to phase shift is applied to extract stable signal
 s of strong magnetic regions. Initial results demonstrate that the model i
 s capable of reproducing large scale transport processes\, such as differe
 ntial rotation\, and can preserve the polarity structure of magnetic field
 s.\n\nhttps://scitalks.tifr.res.in/event/9324/
LOCATION:A-269
URL:https://scitalks.tifr.res.in/event/9324/
END:VEVENT
END:VCALENDAR