Change in Equatorial SST Variability in CMIP6Climate Models

Characterising climate variability and evaluating the effects of global warming require an understanding of the spatiotemporal evolution of sea surface temperature (SST) variability. The ability of many current methods to capture changes in variability under a changing climate is limited because they assume stationary covariance structures. The covariance regression framework of Hoff and Niu (2012) is used in this study to enable the smooth evolution of SST covariance structures over time.

We analyse equatorial SST anomalies from CMIP6 climate model simulations from 1850 to 2100. The analysis is carried out in a reduced-dimensional space using principal components, and the results' sensitivity to the number of retained modes is systematically assessed. Model evaluation is based on variance explained and diagnostics based on likelihood.
The proposed framework provides insight into changes in the structure of equatorial SST variability by visualising evolving SST variance, covariance, and correlation patterns using Hovmöller representations. The results demonstrate how time-varying covariance models can be used to identify coherent large-scale patterns of variability, particularly over the tropical Pacific, and diagnose climate-driven changes in SST structure.

The results demonstrate that equatorial SST variability is consistently increased by both the CESM2 and MRI-ESM2 models. The largest increases are seen in the eastern and central Pacific regions. Variance is steadily increasing in the twenty-first century, according to kernel-based estimates. However, this event is better described by the Hoff covariance regression, which incorporates a coherent large-scale temporal structure. Because the signals are strong and consistent throughout the entire Pacific basin and weaker and less regular in the Indian and Atlantic sectors, the covariance and association patterns remain constant over time. While both models exhibit similar geographic patterns, CESM2 exhibits a larger and more consistent increase in variance than MRI-ESM2, which exhibits smaller and more erratic changes. The study's primary finding is that SST variation increases with global mean temperature, despite correlation patterns remaining largely consistent across models and measurement techniques.Work is still ongoing on other CMIP6 models.