Bayesian Convolutional Neural Network



Published on 09 Mar 2025

• Indian National Centre for Ocean Information Services (INCOIS) has developed a new product known as Bayesian Convolutional Neural Network (BCNN).

• BCNN predicts El Niño and La Niña conditions up to 15 months in advance using AI, deep learning, and machine learning to improve ENSO phase forecasts.

• The model relies on the connection between El Niño/La Niña and slow oceanic variations with atmospheric coupling, providing sufficient lead time for early forecasts.

• It calculates the Nino 3.4 index value, which is the average sea surface temperature (SST) anomaly in the central equatorial Pacific (5°N to 5°S, 170°W to 120°W), to make predictions.

Keywords:

Bayesian Convolutional Neural Network el nino la nina enso el nino southern oscillation