Simulations and Projections of the South Atlantic Ocean Dipole Impacts on Regional Rainfall in CMIP6 Models

Abstract

This study investigates the impacts of the South Atlantic Ocean Dipole (SAOD) on rainfall over land areas of the South Atlantic region during austral winter (June-July-August), using the latest Coupled Model Intercomparison Project phase 6 (CMIP6). We consider a historical reference period (1950-2014) and, future changes under the Shared Socioeconomic Pathway 5-8.5 (SSP585) from 2015 to 2079. Our historical analyses of observations revealed four regions with spatially-coherent correlations of grid-point rainfall with the SAOD namely; Northern Amazon, Guinea Coast, Central Africa and South East Brazil – a result generally consistent with similar analyses in earlier literature. The ensemble mean of 44 CMIP6 models’ historical simulations largely underestimated the SAOD-rainfall correlation in these regions, while individual model performance revealed a spread in model behavior in each region. Model performance was best in Central Africa with 52% of the models simulating statistically significant positive correlations, similar to observations. The worst performance was for South East Brazil with only 7% of the models performing well, while over the Guinea Coast and South East Brazil, more than 40% of the models simulated a negative sign correlation in opposition to observations. Observations showed that SAOD influence on rainfall varied between the regions, being strongest over the Guinea Coast and weakest over South East Brazil. Future simulations of an ensemble of the best-performing models in each region indicated a decrease in SAOD influence on rainfall variability, in all the respective regions, under the SSP585 scenario. Our results underscore the significant impacts of SAOD on regional rainfall variability and highlights the need to enhance CMIP6 models' ability to simulate the SAOD-rainfall relationship. Furthermore, a future with unabated greenhouse gas emissions could cause significant changes in rainfall patterns, leading to unpredictable impacts on the affected regions. Overall, these results could be a useful first step in improving the prediction of regional climate variability and planning adaptation of theregional ecosystems and human socio-economic activities to climate change.