Contributions of El Niño Southern Oscillation (ENSO) Diversity to Low-Frequency Changes in ENSO Variance

TL;DR

The paper tackles how ENSO diversity contributes to decadal changes in ENSO variance by introducing a fuzzy clustering approach in the PC1–PC2 space, producing five probabilistic ENSO categories. A Gaussian Mixture Model with five components is fitted to boreal-winter SSTA to yield non-binary category memberships, which are then used to decompose ENSO variance and its low-frequency modulation. The results show that Extreme EN and Strong LN dominate decadal variance after the 1970s, with shifts around 1976–77 and 2000 linked to increased extreme events. The findings offer a framework for evaluating ENSO diversity in climate models, highlighting biases such as CESM2 missing the Extreme EN category and underestimating ENSO nonlinearities, with implications for prediction and projection of ENSO impacts.

Abstract

El Niño Southern Oscillation (ENSO) diversity is characterized based on the longitudinal location of maximum sea surface temperature anomalies (SSTA) and amplitude in the tropical Pacific, as Central Pacific (CP) events are typically weaker than Eastern Pacific (EP) events. SSTA pattern and intensity undergo low-frequency modulations, affecting ENSO prediction skill and remote impacts. Yet, how different ENSO types contribute to these decadal variations and long-term variance trends remain uncertain. Here, we decompose the low-frequency changes of ENSO variance into contributions from ENSO diversity categories. We propose a fuzzy clustering of monthly SSTA to allow for non-binary event category memberships. Our approach identifies two La Niña and three El Niño categories and shows that the shift of ENSO variance in the mid-1970s is associated with an increasing likelihood of strong La Niña and extreme El Niño events.

Figures (4)

• Figure 1: El Niño and La Niña categories in PC1-PC2 space. Monthly boreal winter (DJF) SSTAs of El Niño and La Niña events of all reanalysis products (see SI Tab. S1) projected onto the PC1-PC2 space and fitted by a Gaussian Mixture Model (GMM). Each event (DJF averages are shown as black dots in A) has a probability of belonging to each of the five categories (colored Gaussians in A). Pacific SSTA composites for each category are obtained by using the category membership as weights for the averages, depicted in panels (B-F). We obtain three El Niño-like patterns: Extreme EN (B), Strong EN (C), and Weak EN (D), while La Niña events form two categories: Weak LN (E) and Strong LN (F).
• Figure 2: Probabilistic category membership. The GMM in Fig. \ref{['fig:pcgmm-latent']} allows us to estimate the likelihood, $p\left( c_{k}|z(t) \right)$, of each El Niño and La Niña winter month, $z(t)$, to belong to each of the categories, $c_{k}$ (SI Sec. S4). An event belongs only to one category when its probability is 1. However, many events have shared probabilities across several categories. The categories are sorted in the following order (top to bottom): Extreme EN (A), Strong EN (B), Weak EN (C), Weak LN (D), and Strong LN (E). For visual reasons, we average the monthly probabilities over each winter (DJF) and over reanalysis products. The dashed lines indicate the reported shifts in ENSO variability in 1976-77 and 2000.
• Figure 3: Extreme EN and Strong EN category. Hovmöller diagrams of SSTA (A, B), SSHA (C, D), high-frequency (HF) zonal wind anomalies (E, F), and low-frequency (LF) zonal wind (G, H) anomalies are obtained by meridional averages (5°S - 5°N) of each month in the year preceding and succeeding El Niño events. Each two-year period is weighted by the corresponding DJF average category membership probability (Fig. \ref{['fig:pcgmm-weights']}). The black line in (A) and (B) indicates the warm-pool edge, i.e., the 29°C SST isotherm (SI Sec. S8). Only values that are statistically significant above the 95th percentile are displayed (SI Sec. S9). SSTA and SSHA are taken from ORAS5 (1958--2022), while 10-meter zonal winds, with their HF- and LF components (Sec. S1) are computed from ERA5.
• Figure 4: Low-frequency changes of ENSO variance. For each category, the Niño3.4 index is multiplied by their category membership probabilities (A). The histogram of Niño3.4 intensities (B), highlights different SSTA amplitudes between categories. The 20-year running variance of Niño3.4 every 10 years (C), is used to normalize the 20-year running variance of each category (D). Extreme EN and Strong EN categories dominate the Niño3.4 variance in the early and late 20th century. The variance shift in 1976-77 and 2000 (dashed lines) highlight reported changes in ENSO variability. The Niño3.4 index is taken from HadISST rayner2003.