Sensitivity of Hybrid ENSO Models to Unresolved Atmospheric Variability

Abstract

Abstract A common practice in the design of forecast models for ENSO is to couple ocean general circulation models to simple atmospheric models. Therefore, by construction these models (known as hybrid ENSO models) do not resolve various kinds of atmospheric variability [e.g., the Madden–Julian oscillation (MJO) and westerly wind bursts] that are often regarded as “unwanted noise.” In this work the sensitivity of three hybrid ENSO models to this unresolved atmospheric variability is studied. The hybrid coupled models were tuned to be asymptotically stable and the magnitude, and spatial and temporal structure of the unresolved variability was extracted from observations. The results suggest that this neglected variability can add an important piece of realism and forecast skill to the hybrid models. The models were found to respond linearly to the low-frequency part of the neglected atmospheric variability, in agreement with previous findings with intermediate models. While the wind anomalies associated with the MJO typically explain a small fraction of the unresolved variability, a large fraction of the interannual variability can be excited by this forcing. A large correlation was found between interannual anomalies of Kelvin waves forced by the intraseasonal MJO and the Kelvin waves forced by the low-frequency part of the MJO. That is, in years when the MJO tends to be more active it also produces a larger low-frequency contribution, which can then resonate with the large-scale coupled system. Other kinds of atmospheric variability not related to the MJO can also produce interannual anomalies in the hybrid models. However, when projected on the characteristics of Kelvin waves, no clear correlation between its low-frequency content and its intraseasonal activity was found. This suggests that understanding the mechanisms by which the intraseasonal MJO interacts with the ocean to modulate its low-frequency content may help to better to predict ENSO variability.