El Niño and the Southern Oscillation (ENSO) remain one of the great puzzles of climatology: the alternation of warm and cold phases in the equatorial Pacific is irregular, and its mechanism is not fully understood. Classic studies focus on interannual ocean temperature anomalies, but recent work has shown that El Niño is synchronized with the annual cycle: events tend to peak in winter and decay rapidly in spring. Along the same lines are studies on predicting El Niño one and a half years ahead, on revived historical weather records, and on the Sun's influence on ice ages: all seek hidden periodicities and connections in climate data. This paper stands somewhat apart: it analyzes not anomalies but the temperature oscillations themselves within a single year and links them to ENSO phases through its own micro-correlation method.
The authors used daily ERA-Interim reanalysis data for air temperature at 2 meters and wind at 10 meters from January 1, 1979, to December 31, 2016, over the tropical zone (30°S–30°N, all longitudes), yielding 3600 grid points. For each sliding window of 365 days (with a one-month step), they constructed a Pearson correlation matrix between grid points. Eigen-decomposition of this matrix gave eigenvalues and eigenvectors, interpreted as the principal modes of intraseasonal temperature variability. To quantitatively assess the coupling between the El Niño 3.4 region and the northern or southern hemisphere, the authors introduced "micro-correlation": the sum of contributions from positive or negative eigenvector components exceeding a threshold, normalized by the number of grid-point pairs. This technique allowed them to track how the region's coupling with the hemispheres changes over time.
The first eigenvalue λ1 correlates positively with the El Niño index (ONI) with a Pearson coefficient of 0.65, while the second λ2 correlates negatively with a coefficient of −0.62. λ1 and λ2 anticorrelate with each other (−0.75). The first mode shows a pronounced seasonal cycle with a maximum in mid-February and a minimum in mid-August; the second mode is phase-shifted: positive values from March to August, negative from August to March. The spatial pattern of the first mode divides the tropics into two large clusters: the Northern Hemisphere with negative components and the Southern Hemisphere with positive ones, with the boundary near the equator and shifting strongly over time. During neutral phases, the El Niño region lies on this boundary and is weakly coupled to the first mode. Before El Niño onset, the micro-correlation between the region and the Northern Hemisphere rises sharply, and at the event's end, coupling with the Southern Hemisphere strengthens abruptly. The second mode, in contrast, shows strong positive coupling within the El Niño region itself during cold and neutral phases, and weak coupling during warm phases.
The paper offers a fresh perspective on the ENSO mechanism: instead of examining ocean temperature anomalies, it focuses on intraseasonal air temperature oscillations and their spatial structure. Unlike neighboring studies that use climate networks or predictive models, here an eigen-decomposition method with the introduction of micro-correlation is applied, allowing a direct link between El Niño phases and the competition of influences from the Northern and Southern Hemispheres. This is the first work where such competition is quantitatively traced through the evolution of eigenvectors rather than statistical indices. The obtained correlations between eigenvalues and ONI are comparable in strength to known predictors, yet they provide a physically interpretable picture.
The proposed micro-correlation method can be applied to other climate modes, for example, in the Indian Ocean or the Atlantic, where seasonal synchronization is also observed. Extending the time series with newer reanalyses and CMIP6 model data would test the robustness of the found links. It is promising to combine this approach with climate network methods to refine spatial pathways of influence. Micro-correlation might prove useful for early warning of El Niño if its peaks indeed precede events.
The results could help improve seasonal forecasts of El Niño, especially in predicting the rapid decay of events. They also provide a new tool for analyzing tropical–extratropical interactions in climate models.
Test whether the found correlations hold in independent data, such as the ERA5 reanalysis or model ensembles. Investigate how micro-correlation behaves for different El Niño types (Eastern Pacific and Central Pacific).
The work touches on a fundamental problem in climatology: why El Niño is irregular and what role the seasonal cycle plays in its onset and decay. It links intraseasonal processes to interannual variability, which remains an open question in understanding ENSO complexity.
The first mode explains more than 40% of the total temperature variability in the tropics, and the second about 10%, so together they cover half of all oscillations.