BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:A Bayesian Neural Network approach to study dissolved oxygen in So
 uthern Ocean water masses - Gian Giacomo Navarra\, Princeton University
DTSTART:20250326T153000Z
DTEND:20250326T163000Z
UID:TALK228058@talks.cam.ac.uk
CONTACT:129418
DESCRIPTION:Oxygen plays a critical role in the health of marine ecosystem
 s. As oceanic O2 concentration decreases to hypoxic levels\, marine organi
 sms’ habitability decreases rapidly. However\, identifying the physical 
 patterns driving this reduction in dissolved oxygen remains challenging. T
 his study employs a Bayesian Neural Network (BNN) to analyze the uncertain
 ty in dissolved oxygen forecasts. The method’s significance lies in its 
 ability to assess oxygen forecasts’ uncertainty with evolving physical d
 ynamics. The BNN model outperforms traditional linear regression and persi
 stence methods\, particularly under changing climate conditions. Our appro
 ach leverages three Explainable AI (XAI) techniques—Integrated Gradients
 \, Gradient SHAP\, and DeepLIFT—to provide meaningful interpretations of
  2- and 8-year forecasts. The XAI analysis reveals that buoyancy frequency
  and eddy kinetic energy is a critical predictor for short-term forecasts 
 across the North Atlantic Deep Water (NADW)\, Upper Circumpolar Deep Water
  (UCDW)\, masses. While the LCDW variability emphasizes also a role played
  by advection processes\, such as salinity\, over short and long timescale
 s.\n
LOCATION:BAS Seminar Room 330b
END:VEVENT
END:VCALENDAR