In previous studies, Kirtman and Shukla (2002) and Wu and Kirtman (2003) demonstrated the ability of the COLA interactive coupled ensemble to capture the ENSO-monsoon relationship significantly better than the uncoupled COLA AGCM. This study investigates two likely explanations for the improvement. First, in the coupled simulation, the atmospheric component is forced by daily SSTs, but monthly mean SSTs are used in the uncoupled simulation. The second argument is that the fluxes and SSTs are inconsistent in the uncoupled simulation.

Three model experiments are used to evaluate the two explanations. Each of the experiments is a 30-year integration with 5 ensemble members. Experiment 1 uses the COLA interactive coupled ensemble. Experiment 2 uses the COLA AGCM with daily SSTs from experiment 1 and perturbed initial conditions. Experiment 3 uses the COLA AGCM with the same initial conditions as Experiment 1 and SSTs from experiment 1 smoothed by a 30-day running mean.

The June- Sept, ensemble mean Indian monsoon rainfall anomalies are correlated with Dec-Feb SSTAs for each of the 3 model runs. Comparison of the correlation between these experiments shows that coupling, not daily SST forcing, is the main reason for the improved simulation of the ENSO-monsoon relationship by the COLA interactive coupled ensemble. Further investigation of the role of intraseasonal variability in this result will also be presented.