El Niño Southern Oscillation (ENSO) is unrealistically weak in the latest ACME coupled model runs. This is manifest in both atmosphere and ocean variables; for example, sea surface temperatures (SSTs) and precipitation both show too little inter-annual variability in the central equatorial Pacific. With the aim of attributing the ENSO simulation deficiencies to the relevant model components – atmosphere, ocean, or coupler – we look at surface fluxes and associated atmosphere and ocean variables in the tropical Pacific from various coupled and un-coupled ACME runs. To gain additional insights, comparisons are made with ERA-Interim and CESM.

In the atmosphere, the main finding is a positive bias in the west Pacific net shortwave, even in runs where the total cloud amount is close to ERA-Interim. This issue is also present in CESM, pointing to possible problems with CAM5’s cloud and radiation schemes. Mean wind stress and latent heat flux are biased in the central and east Pacific, even in an AMIP-like ACME atmosphere-only model run. The result of these biases in ACME coupled runs is that net surface heat flux is too large in the west Pacific. 

In the ocean, we find that SSTs in coupled runs are too low across the tropical Pacific, despite the positive net heat flux biases mentioned above. Ocean mixed layer depth is too shallow in the east Pacific. In the west Pacific, near-isopycnal conditions exist in a deeper layer than in CESM, but the diagnosed mixed layer depth is much smaller, as strictly isopycnal conditions occur only in a very shallow (~10 m) layer.

Based on these analyses using coupled and un-coupled ACME runs, our hypothesis is that the most likely source of the weak ENSO in the ACME coupled runs is the unrealistic representation of vertical mixing in the ocean model caused by very high vertical resolution near the ocean surface.