Predicting El Niño and La Niña with Bayesian CNNs (BCNN)
The Indian National Centre for Ocean Information Services (INCOIS) in Hyderabad has introduced a new forecasting product using advanced technologies to predict El Niño and La Niña conditions up to 15 months in advance. This product, known as the Bayesian Convolutional Neural Network (BCNN), leverages Artificial Intelligence (AI), deep learning, and machine learning (ML) to enhance the accuracy of forecasts related to the El Niño Southern Oscillation (ENSO) phases.
Understanding ENSO
ENSO is a climate phenomenon involving changes in the temperature of waters in the central and eastern tropical Pacific Ocean, coupled with atmospheric fluctuations. It significantly influences global weather patterns. ENSO occurs in irregular cycles of 2-7 years, comprising three phases:
• El Niño (warm phase): Weakened wind systems lead to warmer waters on the eastern side of the Pacific.
• La Niña (cool phase): Strengthened wind systems cause cooler waters on the eastern side.
• Neutral phase: The eastern Pacific remains cooler than the western side due to prevailing wind systems moving warm waters towards Indonesia.
In India, El Niño conditions usually result in a weak monsoon and intense heatwaves, whereas La Niña conditions bring a strong monsoon.
BCNN: The New Forecasting Model
BCNN combines dynamic models with AI to forecast the emergence of El Niño and La Niña conditions more accurately than traditional models. It calculates the Niño3.4 index value, which averages the sea surface temperature anomaly in the central equatorial Pacific, to make predictions. This model can provide a 15-month lead time for forecasts, significantly longer than the 6-9 months lead time of other models.
Challenges and Innovations
Developing the BCNN model involved overcoming significant challenges, primarily the scarcity of historical oceanic temperature data. While land data is abundant, ocean data is limited. To address this, the INCOIS team used historical runs from the Coupled Model Intercomparison Project phases 5 and 6 (CMIP5 and CMIP6), which provided a more extensive dataset for training the model. It took eight months to develop and test the BCNN model thoroughly.
Current Forecast
According to the June 5 bulletin, La Niña conditions are expected to emerge between July and September 2024, with a high probability (70-90%) and continue until February 2025.
SRIRAM’s
Share:
Get a call back
Fill the below form to get free counselling for UPSC Civil Services exam preparation
