top of page

Quantifying the past and future variability in the Bay of Bengal using statistical and deep learning methods

The ocean plays a leading role in shaping our climate and weather patterns like precipitation, extreme heat, and cold. Even though the ocean plays a role in everything from the air we breathe to daily weather and climate patterns, we know very little about our ocean. Ocean models are numerical models with a focus on the properties of oceans and their circulation.

Challenges in ocean modelling

Uncertainties 

  • IC; BC; Terms; Equations

  • Sparse and Gappy Data

  • Intermittent Features

  • Non-stationary, non-Gaussian statistics.

 

Nonlinear multiscale ocean dynamics

  • Tides, Internal waves, river inflow, fronts, eddies, upwelling, etc.

​​​

Drawbacks in GCM

Bias and Errors

  • System and Model bias and errors

 

Ocean dynamics

  • Major features are missing.

  • Accuracy of climate projections.

  • Impacts decision making related to climate

​​​

Some important Questions

  • What are the major salinity features in the Andaman Sea and the Bay of Bengal?

  • What is the intra-annual variability of Andaman Sea salinity?

  • Does it have any relationship with BoB variability? What are the causes and impacts?

  • Can we incorporate data assimilation to perform a synoptic forecast of the AnS?

  • Is CMIP6 good for future projections? If no, what are good correction methods?

  • Is it possible to train a deep neural network to correct the errors in CMIP6 using reanalysis data? If so, what architecture and training procedure should be used?

  • Is a deep learning-based method capable of achieving better error correction than classical statistical methods?

  • Do the corrected projections reveal dynamical implications that were not expected from the uncorrected projections?

Quantifying the past and future variability in the Bay of Bengal using statistical and deep learning methods

The ocean plays a leading role in shaping our climate and weather patterns like precipitation, extreme heat, and cold. Even though the ocean plays a role in everything from the air we breathe to daily weather and climate patterns, we know very little about our ocean. Ocean models are numerical models with a focus on the properties of oceans and their circulation.

Challenges in ocean modelling

Uncertainties 

  • IC; BC; Terms; Equations

  • Sparse and Gappy Data

  • Intermittent Features

  • Non-stationary, non-Gaussian statistics.

 

Nonlinear multiscale ocean dynamics

  • Tides, Internal waves, river inflow, fronts, eddies, upwelling, etc.

​​​

Drawbacks in GCM

Bias and Errors

  • System and Model bias and errors

 

Ocean dynamics

  • Major features are missing.

  • Accuracy of climate projections.

  • Impacts decision making related to climate

​​​

Some important Questions

  • What are the major salinity features in the Andaman Sea and the Bay of Bengal?

  • What is the intra-annual variability of Andaman Sea salinity?

  • Does it have any relationship with BoB variability? What are the causes and impacts?

  • Can we incorporate data assimilation to perform a synoptic forecast of the AnS?

  • Is CMIP6 good for future projections? If no, what are good correction methods?

  • Is it possible to train a deep neural network to correct the errors in CMIP6 using reanalysis data? If so, what architecture and training procedure should be used?

  • Is a deep learning-based method capable of achieving better error correction than classical statistical methods?

  • Do the corrected projections reveal dynamical implications that were not expected from the uncorrected projections?

Projects

Bias Correction (submitted)

Climate change affects ocean temperature, salinity and sea level, which in turn impacts monsoons and ocean productivity. Future projections by Global Climate Models (GCM) based on shared socioeconomic pathways (SSP) from the Coupled Model Intercomparison Project (CMIP) are widely used to understand the effects of climate change. However, CMIP models have significant bias compared to reanalysis in the Bay of Bengal for the time period when both projections and reanalysis are available. For example, there is a 1.5C root mean square error (RMSE) in the sea surface temperature (SST) projections of the climate model CNRM-CM6 compared to the Ocean Reanalysis System (ORAS5). We develop a suite of data-driven deep learning models for bias correction of climate model projections and apply it to correct SST projections of the Bay of Bengal. We found that the UNet architecture that is trained using a climatology-removed CNRM-CM6 projection as input and climatology-removed ORAS5 as output gives the best bias-corrected projections. Our novel deep learning-based method for correcting CNRM-CM6 data has a 15\% reduction in RMSE compared to the widely used statistical correction technique, the Equidistant Cumulative Distribution Function (EDCDF).

oras5_sst_june_hist_mean1.jpg

June climatology for the Bay of Bengal from 1958 to 2014 using (a) CMIP6 historical from CNRM-CM6 model and (b) ORAS5 reanalysis 

unet_des.jpg

Schematic of the UNet fully convolutional encoder-decoder architecture for bias correction of CNRM-CM6.

Ocean Modelling

Development and applications of stochastic and deterministic primitive equation ocean modeling systems. Currently, we are using the Regional Ocean Modeling System (ROMS) and Multidisciplinary Simulation, Estimation, and Assimilation Systems (MSEAS) for the synoptic-scale regional prediction for the different parts of the Northern Indian Ocean.

Crucially, regional ocean forecasting with primitive equations has several sources of uncertainty which affect the forecast and introduce erros. To minimize these errors an accurate estimation of initial and boundary conditions, parametrizations etc is essential.

Picture 1.png
Andaman Sea (published)

The Andaman Sea region is of major importance for India from a security and conservation viewpoint. Describing this region's salinity variability is fundamental for understanding its dynamics. We study the inter-annual salinity variability of the Andaman Sea during the Boreal summer using NEMO reanalysis data (1993-2018), focusing on its causal factors and impact. 

​

We found that significant salinity is being transported into the Andaman Sea by SMC using EOF analysis. Particle trajectories experiment is conducted to understand the path of high salinity water of SMC. 

​

bob_final_v4.jpg

Major features in the Bay of Bengal during the summer monsoon.

Data Assimilation (published)

Feature Oriented Regional Modelling System

Andaman Sea is inextricably linked to the Indian Ocean and plays a vital role in connecting the equatorial Indian Ocean to the Bay of Bengal. Figure shows the Andaman Sea region with the bathymetry. Very few studies exist in which data assimilative forecasts of this important region is attempted. Towards this end, first we create data-driven initial conditions of the region using feature models and Bayesian assimilation techniques. Thereafter we employ primitive equation ocean modelling systems  with data assimilation to forecast the ocean state in the region.

roms_as_domain.png

Publications

  1. Pasula, Abhishek, and Deepak N. Subramani. "Cause and impact of Andaman Sea's salinity variability: A modeling Study." Deep Sea Research Part II: Topical Studies in Oceanography (2023): 105291. https://doi.org/10.1016/j.dsr2.2023.105291 

  2. Pasula Abhishek, Deepak N Subramani. “4D-Var Data Assimilation of Sea Surface Temperature in a Regional Model of the Andaman Sea” IEEE Oceans 2022. https://doi.org/10.1109/OCEANS47191.2022.9977119.

  3. Pasula Abhishek, and Sourav Sil. "Validation of Multi-Scale Ultra-High Resolution (MUR) Sea Surface Temperature with Coastal Buoys Observations and Applications for Coastal Fronts in the Bay of Bengal." In 2019 URSI Asia-Pacific Radio Science Conference (AP-RASC), pp. 1-4. IEEE, 2019, 10.23919/URSIAP-RASC.2019.8738356. https://doi.org/10.23919/URSIAP-RASC.2019.8738356

  4. Pasula Abhishek, and Deepak N. Subramani. “Data Driven Deep Learning for Correcting Global Climate Model Projections of SST and DSL in the Bay of Bengal” (submitted in Climate Dynamics) https://arxiv.org/pdf/2504.20620

  5. Pasula Abhishek, and Deepak N. Subramani. “Bias Correction of Global Climate Model Salinity Projections in the Bay of Bengal” (submitted in Scientific Reports: Advances in Climate).

  6. Pasula Abhishek, and Deepak N. Subramani. “Global Climate Model Bias Correction using Deep Learning” (submitted in IOP Science Machine Learning: Earth) https://arxiv.org/pdf/2504.19145

Conferences

  1. Pasula Abhishek, and Deepak N. Subramani (2024), “A two-phase Neural Model for CMIP6 bias correction”, EGU General Assembly  2024, Vienna, Apr 14-19, 2024.

  2. Pasula Abhishek, Deepak N Subramani (2024),” AI-based correction of CMIP6 ocean projections”, Ocean Sciences Meeting 2024, Feb 16-23, 2024.

  3. Pasula Abhishek, Deepak N Subramani. “4D-Var Data Assimilation of Sea Surface Temperature in a Regional Model of the Andaman Sea” IEEE Oceans 2022.

  4. Pasula Abhishek, Deepak N Subramani (2021),” Cause and Impact of Andaman Sea’s Salinity Variability: A Modeling Study”, Ocean Sciences Meeting 2022, Feb 24- Mar 4, 2022.

  5. Deepak N. Subramani, Ratnakar Gadi, Abhishek P (2020), “Bayesian Estimation and Data Assimilation for Probabilistic Regional Forecasts in the northern Indian Ocean”, Ocean Sciences Meeting, San Diego, 16-21 Feb, 2020.

  6. Abhishek Pasula, and Sourav Sil (2019), “Validation of Multi-Scale Ultra-High Resolution (MUR) Sea Surface Temperature with Coastal Buoys Observations and Applications for Coastal Fronts in the Bay of Bengal”, URSI Asia-Pacific Radio Science Conference (AP-RASC 2019) New Delhi, 09-15 Mar, 2019.

  7. Abhishek Pasula, Samiran Mandal, Sourav Sil (2018), Observed Frontal eddies in the Coastal Bay of Bengal using HF Radar and High-Resolution SST data, Abstract [OS21D-1601] presented at 2018 Fall Meeting, American Geophysical Union, Washington, D.C., 10-14 Dec 2018.

  8. Abhishek Pasula, Samiran Mandal, Sourav Sil (2018), “East India Coastal Current and Frontal Eddies during Fall 2010: A study using HF Radar Currents and High-Resolution SST”, National Oceanography Workshop 2018, Indian National Centre for Ocean Information Services (INCOIS)-2018 at Hyderabad, 14-16 Nov 2018.

  9. Participated in the TROPICAL METEROLOGY-2018 by INDIAN METEROLOGICAL SOCIETY-National Symposium on understanding weather and climate variability: Research for Society at Banaras Hindu University- Varanasi. 24- 27 Oct 2018.

  10. Attended the Coastal Vulnerability Workshop by National Institute of Ocean Technology (NIOT) Chennai during December 2017.

bottom of page