Bayesian Convolutional Neural Network (BCNN)

Hyderabad-based Indian National Centre for Ocean Information Services (INCOIS) has created a novel tool 

Key Highlights On Bayesian Convolutional Neural Network

• The novel tool is Bayesian Convolutional Neural Network (BCNN).
• Objective: The tool forecasts the onset of El Niño and La Niña phases of the El Niño Southern Oscillation (ENSO).
  • It predicts these climate patterns up to 15 months in advance.
  • According to the bulletin issued on June 5, it is highly likely (70-90% probability) that La Niña conditions will develop from July to September and persist until February 2025.

About Bayesian Convolutional Neural Network (BCNN)

• It is a variant of the Convolutional Neural Network. 
  • It uses advanced technologies including Artificial Intelligence (AI), deep learning, and machine learning (ML).
• It forecasts the onset of El Niño and La Niña phases of the El Niño Southern Oscillation (ENSO).
  • It can predict these climate patterns up to 15 months in advance.

Enroll now for UPSC Online Course

Functionality of the Model

• Objective: BCNN aims to enhance forecasts related to El Niño and La Niña phases of the ENSO.
• The model’s predictive capabilities leverage the connection between these phases and gradual oceanic changes, coupled with atmospheric interactions.

Operational Details

• The model calculates predictions based on the Niño3.4 index value.
• This index averages sea surface temperature (SST) anomalies in the central equatorial Pacific region, spanning from 5°N to 5°S and 170°W to 120°W.

Significance

• Early Warning System: Provides early forecasts by analyzing oceanic variations and their atmospheric effects, offering valuable lead time for preparedness and planning.
• Advancement in ENSO:  The Bayesian Convolutional Neural Network represents a significant advancement in ENSO prediction technology.
  • It aids in better understanding and preparation for climate variability linked to ocean-atmosphere interactions.

What is ENSO?

• ENSO (El Niño Southern Oscillation) involves fluctuations in the temperature of waters in the central and eastern tropical Pacific Ocean, alongside changes in the overlying atmosphere.
• Phases of ENSO
  • ENSO operates in irregular cycles lasting 2-7 years and manifests in three main phases: warm (El Niño), cool (La Niña), and neutral.
    • Neutral Phase: During the neutral phase, the eastern Pacific near South America is cooler due to prevailing east-to-west winds displacing warmer waters towards Indonesia.
    • El Niño Phase
      • In El Niño, weakened wind systems reduce the displacement of warm waters.
        • Consequently, the eastern Pacific becomes warmer than usual.
    • La Niña Phase
      • Conversely, in La Niña, strengthened wind systems intensify the displacement of warm waters towards Indonesia. 
        • It leads to cooler-than-normal conditions in the eastern Pacific.
• Impact on India – Monsoon Effects
  • El Niño: Often results in a weak monsoon and heightened heat waves across India.
  • La Niña: Typically brings about a robust monsoon season in the region.

These phases of ENSO significantly influence global atmospheric circulation, thereby affecting weather patterns worldwide, including in India.

Comparison between Existing Models (Statistical & Dynamic) and BCNN Model

Challenges in Developing BCNN

• Limited Data: Weather forecasting models rely on historical data for training. While land-based data is plentiful, data for oceans is scarce, especially for long periods.
  • This scarcity of oceanic data especially impacts El Niño/La Niña prediction.
• Scarcity in Oceanic Data: Global oceanic temperature records have only been reliably accessible since 1871.
  • This results in fewer than 150 monthly samples available for training deep learning models like BCNN for El Niño and La Niña predictions.
    • It limits the training dataset for such predictions.

Check Out UPSC Modules From PW Store

Overcoming Data Challenges

• Incorporation of Historical Runs
  • Use of CMIP Data:
    • INCOIS addressed the data scarcity by integrating historical runs (1850-2014) from the Coupled Model Intercomparison Project phases 5 and 6 (CMIP5 and CMIP6).
    • This augmentation of the training dataset enriched the model’s capability to forecast ENSO phases.