Tropical Atlantic sea surface temperature is a key node in the global climate system, influencing precipitation, circulation, and hurricanes. Traditionally it has been described with linear methods, but the ocean and atmosphere exhibit nonlinearity, long memory, and intermittency, requiring multifractal analysis. Nearby in the archive are studies of gravitational waves from cosmic cracks, magnetic fields after a solar storm, and weather forecasting, as well as analyses of DNA multifractality and protein evolution. This paper stands apart: the nearest neighbor is at distance 0.478, with no close works, and the thematic group links it to LIGO, baryon acoustic oscillations, and gravitational lensing. This underscores that the work is about scaling of time series, not astrophysics, and fills a gap in comparative analysis of regional indices and the interhemispheric gradient.
The authors used second-order multifractal detrended fluctuation analysis (MFDFA-2) on three weekly SST indices: SAT, TSA, and TASI (the difference in anomalies between the northern and southern tropical Atlantic). A key choice is the scale range from 18 to 100 weeks: the lower bound is set by the condition smin ≳ 6(m+1) for a second-order polynomial trend, and the upper bound stops before the scaling break observed beyond 100 weeks. Moments q ranged from −4 to 4, since for large |q| the fluctuation function is dominated by a handful of extreme segments. To test the significance of the multifractal spectrum width, two types of surrogates were generated: randomly shuffled series (destroying correlations while preserving the distribution) and IAAFT surrogates (preserving linear autocorrelation but randomizing phases). Additionally, a sliding window analysis (window 5 years, step 3 months) was applied to estimate the local Hurst exponent h2(t) and local width with segment-wise surrogate tests. The link to El Niño was tested via lagged detrended cross-correlation ρDCCA with a family-wise null distribution accounting for the search over lags.
In the primary scale range, none of the three indices showed significant multifractality. The spectral widths of SAT (Δh = 0.077) and TASI (Δh = 0.054) lie within both surrogate null hypotheses. The width of TSA (Δh = 0.217) exceeds both nulls (p = 0.010 against shuffled, p = 0.040 against IAAFT), but is excluded for structural reasons: the generalized Hurst exponent h(q) increases monotonically with q, which through the relation f(α) = 1 + q²h′(q) gives f(α) > 1, impossible for a one-dimensional representation. The increase in h(q) indicates fitting a single power law across a change in scaling regime, not a multiplicative cascade. When the range is extended to 230 weeks, the picture changes completely: the width of TASI triples to 0.239 and becomes significant, while the width of TSA drops to 0.010, demonstrating an artifact. The ordering of indices by width changes with the range, moments, and detrending order, so no ranking survives scrutiny. Only two properties are robust: strong persistence (h(2) from 1.19 to 1.42, exceeding unity for all detrending orders) and moderate leptokurtosis of increments (excess kurtosis less than 1). Pairwise relationships are described by a single cross-scaling exponent λ(q), constant in q within uncertainty, confirming monofractality. The previously reported link between the TASI gradient and El Niño does not survive a family-wise surrogate test: the peak at lag 2–3 months does not exceed the threshold ±0.285 (p = 0.16).
This work differs from its archive neighbors: they investigate gravitational waves, magnetic fields, or genomic acoustics, whereas here the focus is on scaling of climate time series. Unlike previous studies that applied MFDFA to individual indices or SST fields, this work is the first to systematically compare regional indices and the interhemispheric gradient with surrogate tests and a check of the singularity spectrum's admissibility. It shows that previously reported estimates of multifractal strength in the tropical Atlantic are substantially inflated due to artifacts in the choice of scale range and moments. The methodological conclusion is universal: the width of a multifractal spectrum cannot be interpreted without checking the sign of the derivative h′(q) and the stability of the fitting range.
The topic could evolve toward longer series or high-resolution data that would allow distinguishing weak multifractality, if it exists. It is also promising to apply the same tests to other climate indices and fields to separate genuine cascade processes from artifacts. Further development of surrogate testing methods that account for multiple comparisons and window analysis will help avoid false conclusions about intermittency.
The results will affect climatology, oceanography, and the analysis of complex systems, where multifractal analysis is used to diagnose nonlinear dynamics. They are also important for the statistical physics of time series, as they set a standard for testing the significance of scaling exponents.
The next step is to apply the same criteria to longer SST series or to data with higher temporal resolution, to see whether significant multifractality emerges with increased data volume. It is also worth investigating how the choice of scale range affects conclusions in other geophysical series.
In climatology, the question remains open as to which processes generate long memory and intermittency in ocean temperature. This work shows that for the tropical Atlantic, on available scales, monofractal persistence dominates rather than a multiplicative cascade. This limits the applicability of multifractal models to this region and underscores the need for careful selection of the scale range.
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
The authors re-examined the scaling properties of three weekly sea surface temperature indices for the tropical Atlantic (SAT, TSA, TASI) over 1981–2025 using MFDFA-2 and MFCCA, choosing a scale range of 18–100 weeks and moments |q| ≤ 4. They showed that, with a proper choice of parameters, none of the indices exhibits significant multifractality: the spectral widths of SAT and TASI are consistent with surrogate null hypotheses, while the width of TSA is ruled out for structural reasons (f(α) > 1). All series are strongly persistent (h(2) = 1.32–1.35) and locally nonstationary, and pairwise links are described by a single cross-scaling exponent, indicating a monofractal nature.
Where its strength is
The strength of the work lies in the careful handling of methodological subtleties: the choice of scale range, the restriction on moments, the use of two types of surrogates, and structural diagnostics of artifacts. This allows distinguishing true multifractality from spurious multifractality arising from nonstationarity or extreme fluctuations.
Where it could go next
Worth trying
Why this may matter
The work demonstrates how a rigorous statistical approach can revise previous conclusions about multifractality in climate series, which is important for accurate modeling and forecasting of the tropical Atlantic and its influence on global climate.
Nearby works in our archive
The archive is sparse around this work: the closest review is only 0.48, against a usual 0.64. It looks like few have drilled nearby.
The width of the TSA multifractal spectrum collapsed by a factor of 20 when the scale range was extended from 100 to 230 weeks, dropping from 0.217 to 0.010. This is a vivid example of how a single fitting parameter can completely change the conclusion about a system's complexity.