Climate scientists have long sought to understand how El Niño, the warm current in the tropical Pacific, affects the Indian monsoon. Typically, during El Niño the monsoon weakens, but in recent decades this connection has become less apparent. The problem is that traditional analysis methods rely on time averages, which are unreliable in a changing climate. This work stands apart: neighboring studies focus on El Niño prediction, reviving old weather records, and neural networks for drought forecasting, but none have applied an ensemble approach to this specific teleconnection. The authors fill that gap by using the concept of the instantaneous attractor to properly separate the forced signal from internal variability.
The authors used ensembles from two models: CESM (35 realizations) and MPI-ESM (up to 100 realizations). They selected two simple variables: the sea-level pressure difference between Tahiti and Darwin (an El Niño indicator) and precipitation over northern India (a monsoon indicator). To avoid errors from time averaging, they used the instantaneous attractor: an ensemble of trajectories that, after 40 years, forgets initial conditions and describes all possible climate states under a given forcing. They then computed the correlation coefficient between the variables at each time instant across the ensemble, rather than over time. To test stationarity, they applied Kolmogorov–Smirnov tests, t-tests, and the Mann–Kendall test.
In the MPI-ESM model with historical forcing (1890–2005), the correlation between El Niño and the monsoon strengthened, especially in the last two decades. Under the RCP8.5 scenario (2006–2099) and a 1% per year CO2 increase, the correlation remained stable. In the CESM model, no changes were detected. Correlation values for the summer season (June–August) in MPI-ESM are around 0.4, consistent with observations, while in CESM they are near zero. Mean sea-level pressure and precipitation shift linearly in phase space, but the slope depends strongly on season: precipitation increases mainly during the monsoon, while pressure rises toward the end of the year. This refutes the hypothesis of a universal link between them.
This work stands out methodologically: instead of time series, it uses ensembles converging to the instantaneous attractor. This allows, for the first time, a reliable separation of the forced response from internal variability for this teleconnection. The results show that the strengthening during the 20th century does not contradict observations when accounting for methodological errors in previous studies. The lack of change under future scenarios challenges the use of radiative forcing as the sole driver. The seasonality of the slope rules out emergent constraints between these variables.
The topic could advance by examining the role of individual forcing agents (aerosols, volcanoes, CO2) separately. It is also important to identify low-frequency modes of internal variability that might mask the forced signal. Larger ensembles would refine the temporal evolution of the correlation. The approach could be applied to other teleconnections, such as those between El Niño and rainfall in Africa or Australia.
The results will affect climate modeling and monsoon forecasting, as well as the interpretation of paleoclimate data.
Conduct a similar analysis for other models and scenarios, and investigate the mechanisms linking volcanic activity and the strengthening of the teleconnection.
This connects to a fundamental problem in climate science: how to separate the forced signal from internal variability in a nonstationary system. The work shows that time averages can produce spurious trends and proposes an ensemble approach as a solution.
The sea-level pressure difference between Tahiti and Darwin used in this study underlies the Southern Oscillation, an index that meteorologists have been calculating for over a century.