The connection between the Atlantic multidecadal oscillation and the Indian summer monsoon in CMIP5 models

Abstract

A number of observational and modeling studies have shown a co-relationship between higher than normal sea surface temperatures (SSTs) in the North Atlantic and increased summer precipitation over India. However, discrepancies among the models make the robustness of the results debatable. This study examines the connections between the Atlantic multidecadal oscillation (AMO) and the Indian summer monsoon (ISM) in 66 “Historical” runs of 22 coupled models that were part of the fifth phase of the Coupled Model Intercomparison Project (CMIP5). Diverse results are obtained, and correlation coefficients between the AMO and ISM range from − 0.39 to 0.66. Only 10 out of 66 members (~ 15%) show a positive correlation statistically significant at the 90% level (> 0.42), close to the observation (0.5). The models with positive AMO–ISM correlations show an AMO-related atmospheric teleconnection that involves an extratropical–tropical SST gradient in the North Pacific, as well as a more regional temperature difference between the Indian subcontinent and the tropical Indian Ocean. In comparison, the models with negative correlations fail to capture these teleconnections. Moreover, the models with higher climatological precipitation over the tropical Atlantic and warmer climatological SST in the tropical Atlantic and the North Pacific relative to multi-member ensemble, as well as a weak westerly jet, perform better at reproducing the observed teleconnections.

Appendix: Statistical significance tests

For temporal correlation, the degrees of freedom are estimated by ${N_e}=\frac{N}{{1+2\mathop \sum \nolimits_{{i=1}}^{{10}} {a_i}{b_i}}}$, where N is the number of years, and a _i and b _i are the ith order autocorrelation for time series a and b, respectively.

For Pos. Corr. and Neg. Corr., the test statistic is defined as ${\text{t}}=\frac{{\bar {x} - \mu }}{{{\raise0.7ex\hbox{$s$} \!\mathord{\left/ {\vphantom {s {\sqrt n }}}\right.\kern-0pt}\!\lower0.7ex\hbox{${\sqrt n }$}}}}$, where ${{{\upmu}}}$ is the mean of all samples (all models), $\bar {x}$ represents the mean of samples (Pos. Corr. or Neg. Corr.), s is the standard deviation of samples, and n is the number of samples. Significance levels indicate that the difference between Pos. Corr. or Neg. Corr. and the all models is statistically significant at the 5% level, based on a two-tailed Student’s t test.