Anatomy of the Indian Summer Monsoon and ENSO relationships in state-of-the-art CGCMs: role of the tropical Indian Ocean

Abstract

Indian Summer Monsoon (ISM) rainfall and El Niño-Southern Oscillation (ENSO) exhibit an inverse relationship during boreal summer, which is one of the roots of ISM interannual variability and its seasonal predictability. Here we document how current climate and seasonal prediction models simulate the timing and amplitude of this ISM-ENSO teleconnection. Many Coupled General Circulation Models (CGCMs) do simulate a simultaneous inverse relationship between ENSO and ISM, though with a large spread. However, most of them show significant negative correlations before ISM, which are at odd with observations. Consistent with this systematic error, simulated Niño-3.4 Sea Surface Temperature (SST) variability has erroneous high amplitude during boreal spring and ISM rainfall variability is also too strong during the first part of ISM. The role of the Indian Ocean (IO) in modulating the ISM-ENSO relationships is further investigated using dedicated experiments with the SINTEX-F2 CGCM. Decoupled tropical Pacific and IO experiments are conducted to assess the direct relationship between ISM and IO SSTs on one hand, and the specific role of IO feedback on ENSO on the other hand. The direct effect of IO SSTs on ISM is weak and insignificant at the interannual time scale in the Pacific decoupled experiment. On the other hand, IO decoupled experiments demonstrate that El Niño shifts rapidly to La Niña when ocean–atmosphere coupling is active in the whole IO or only in its western part. This IO negative feedback is mostly active during the decaying phase of El Niño, which is accompanied by a basin-wide warming in the IO, and significantly modulates the length of ENSO events in our simulations. This IO feedback operates through a modulation of the Walker circulation over the IO, which strengthens and shifts eastward an anomalous anticyclone centered on the Philippine Sea and associated easterly wind anomalies in the equatorial western Pacific during boreal winter. In turn, these atmospheric anomalies lead to a fast ENSO turnabout via oceanic adjustement processes mediated by eastward propagating upwelling Kelvin waves. An experiment in which only the SouthEast Indian Ocean (SEIO) is decoupled, demonstrates that the equatorial SST gradient in the IO during boreal winter plays a fundamental role in the efficiency of IO feedback. In this experiment, simulated ISM-ENSO lead-lag correlations match closely the observations. This success is associated with removal of erroneous SEIO SST variability during boreal winter in the SEIO decoupled experiment. Finally, it is illustrated that most CMIP5 CGCMs exhibit similar SST errors in the SEIO during boreal winter in addition to an exagerated SEIO SST variability during boreal fall.

Appendix

The SST nudging is performed inside the SST equation of the ocean model. We suppressed the SST variability in a specific domain by applying a strong nudging of the SST toward a SST climatology computed from a control experiment or observations (see Table 1 and Sect. 2.3 for details). More precisely, this is done through a modification of the non-solar heat flux provided by the atmosphere to the ocean by adding a correction term that is proportional with the SST difference with the target climatology at each time step:

$${\text{Q}}_{{{\text{ns}}}} \;{ = }\;{\text{Q}}_{{{\text{ns}}}} \;{ + }\;{\text{dq/dt}}\times\left( {{\text{SST}} - {\text{SST}}_{{{\text{clim}}}} } \right)$$

Q_ns, the nonsolar heat flux received from atmosphere. dq/dt, − 2400 W/m²/K (corresponds to the heat flux needed to warm a 50 m thick ocean layer of 1 K during 1 day). SST_clim, the target SST daily climatology, after a linear interpolation at each time step of the day.

The very strong dq/dt constant used here implies that the SST variability is almost suppressed in the selected domain. A Gaussian smoothing is finally applied in a transition zone in both longitude and latitude at the limits of the SST restoring domains (see Table 1).

This approach is interesting because the ocean dynamics (e.g., also thermocline variations) will be consistent with the SST in the ocean model, which is not the case when the SST is changed in the coupling interface of the coupled model as it is commonly done in many past/recent studies.