Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Data Driven Deep Learning for Correcting Global Climate Model Projections of SST and DSL in the Bay of Bengal

Abhishek Pasula, Deepak N. Subramani

TL;DR

This study addresses biases in CMIP6 projections for the Bay of Bengal by introducing a data-driven UNet-based bias-correction framework that learns a mapping from biased GCM outputs to ORAS5-like reanalysis fields. Trained on historical data (1958–2020) with climatology-removed inputs, the model corrects $SST$ and $DSL$ projections for 2015–2100, outperforming the traditional EDCDF method and reducing RMSE while increasing pattern correlations. The approach yields corrected 2021–2024 test results with RMSE reductions of 9–38% and PCC gains of 0.02–0.12, and provides detailed analyses of monthly, seasonal, and decadal changes across SSPs. The corrected projections reveal meaningful dynamical shifts in winter and pre-monsoon processes, stronger EICC variability, enhanced monsoon-related DSL features, and heightened cyclone risk, underscoring the practical importance of more accurate bias correction for climate impact assessments and policymaking.

Abstract

Climate change alters ocean conditions, notably temperature and sea level. In the Bay of Bengal, these changes influence monsoon precipitation and marine productivity, critical to the Indian economy. In Phase 6 of the Coupled Model Intercomparison Project (CMIP6), Global Climate Models (GCMs) use different shared socioeconomic pathways (SSPs) to obtain future climate projections. However, significant discrepancies are observed between these models and the reanalysis data in the Bay of Bengal for 2015-2024. Specifically, the root mean square error (RMSE) between the climate model output and the Ocean Reanalysis System (ORAS5) is 1.2C for the sea surface temperature (SST) and 1.1 m for the dynamic sea level (DSL). We introduce a new data-driven deep learning model to correct for this bias. The deep neural model for each variable is trained using pairs of climatology-removed monthly climate projections as input and the corresponding month's ORAS5 as output. This model is trained with historical data (1950 to 2014), validated with future projection data from 2015 to 2020, and tested with future projections from 2021 to 2023. Compared to the conventional EquiDistant Cumulative Distribution Function (EDCDF) statistical method for bias correction in climate models, our approach decreases RMSE by 0.15C for SST and 0.3 m for DSL. The trained model subsequently corrects the projections for 2024-2100. A detailed analysis of the monthly, seasonal, and decadal means and variability is performed to underscore the implications of the novel dynamics uncovered in our corrected projections.

Data Driven Deep Learning for Correcting Global Climate Model Projections of SST and DSL in the Bay of Bengal

TL;DR

This study addresses biases in CMIP6 projections for the Bay of Bengal by introducing a data-driven UNet-based bias-correction framework that learns a mapping from biased GCM outputs to ORAS5-like reanalysis fields. Trained on historical data (1958–2020) with climatology-removed inputs, the model corrects and projections for 2015–2100, outperforming the traditional EDCDF method and reducing RMSE while increasing pattern correlations. The approach yields corrected 2021–2024 test results with RMSE reductions of 9–38% and PCC gains of 0.02–0.12, and provides detailed analyses of monthly, seasonal, and decadal changes across SSPs. The corrected projections reveal meaningful dynamical shifts in winter and pre-monsoon processes, stronger EICC variability, enhanced monsoon-related DSL features, and heightened cyclone risk, underscoring the practical importance of more accurate bias correction for climate impact assessments and policymaking.

Abstract

Climate change alters ocean conditions, notably temperature and sea level. In the Bay of Bengal, these changes influence monsoon precipitation and marine productivity, critical to the Indian economy. In Phase 6 of the Coupled Model Intercomparison Project (CMIP6), Global Climate Models (GCMs) use different shared socioeconomic pathways (SSPs) to obtain future climate projections. However, significant discrepancies are observed between these models and the reanalysis data in the Bay of Bengal for 2015-2024. Specifically, the root mean square error (RMSE) between the climate model output and the Ocean Reanalysis System (ORAS5) is 1.2C for the sea surface temperature (SST) and 1.1 m for the dynamic sea level (DSL). We introduce a new data-driven deep learning model to correct for this bias. The deep neural model for each variable is trained using pairs of climatology-removed monthly climate projections as input and the corresponding month's ORAS5 as output. This model is trained with historical data (1950 to 2014), validated with future projection data from 2015 to 2020, and tested with future projections from 2021 to 2023. Compared to the conventional EquiDistant Cumulative Distribution Function (EDCDF) statistical method for bias correction in climate models, our approach decreases RMSE by 0.15C for SST and 0.3 m for DSL. The trained model subsequently corrects the projections for 2024-2100. A detailed analysis of the monthly, seasonal, and decadal means and variability is performed to underscore the implications of the novel dynamics uncovered in our corrected projections.
Paper Structure (40 sections, 10 figures, 3 tables)

This paper contains 40 sections, 10 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Prior work in bias correction of climate projections
  3. Model Development
  4. Problem Definition
  5. Data Sets
  6. UNet Model for SST and DSL Bias Correction
  7. Training Procedure
  8. Analysis of SST and DSL projections in the test years
  9. Evaluation of the corrected projections in the test period (2021-2024)
  10. Comparison of SST projections in 2022
  11. Winter
  12. Pre-monsoon
  13. Monsoon
  14. Post-monsoon
  15. Comparison of DSL projections in 2022
  16. ...and 25 more sections

Figures (10)

  • Figure 1: June climatology SST for the Bay of Bengal from 1958 to 2014 using (a) CNRM-CM6 historical (CMIP6) and (b) ORAS5 reanalysis.
  • Figure 2: Components of the neural network named UNet developed for correcting the bias in CNRM-CM6 projections of SST and DSL in the Bay of Bengal.
  • Figure 3: Comparison of Sea Surface Temperature (SST) and Dynamic Sea Level (DSL) during the test period in the Bay of Bengal. (a-d) SST for June 2022: ORAS5 observations and bias comparison with CNRM-ORAS5, EDCDF-ORAS5, and UNet-ORAS5. (e-h) DSL for March 2021: ORAS5 observations and bias comparison with CNRM-ORAS5, EDCDF-ORAS5, and UNet-ORAS5.
  • Figure 4: Monthly sea surface temperature (SST) in the BoB during 2022, comparing raw CNRM-CM6 model output (CNRM-CM6), UNet-corrected SST (UNet), and ORAS5 reanalysis data.
  • Figure 5: Monthly dynamic sea level (DSL) in the BoB during 2022, comparing raw CNRM-CM6 model output (CNRM-CM6), UNet-corrected DSL (UNet), and ORAS5 reanalysis data.
  • ...and 5 more figures