Arctic Oscillation Modulation of Winter Air-Sea Coupling in the East/Japan Sea: Persistence, Timescales, and Extremes

Gyuchang Lim; Jongjin Park

REVIEW 4 major objections 5 minor 1 cited by

Arctic Oscillation Modulation of Winter Air-Sea Coupling in the East/Japan Sea: Persistence, Timescales, and Extremes

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The Arctic Oscillation's phase determines both the spatial pattern of winter air-sea coupling in the East/Japan Sea and the effective memory timescales over which the ocean integrates atmospheric forcing, from roughly 2-3 weeks for wind-str

desk verdict Useful descriptive core, but the headline memory windows are partly censored boundary values and the DFA 'validation' is circular—worth refereeing, not desk-rejecting. read the letter →

arxiv 2509.09628 v1 pith:HLKFAWT7 submitted 2025-09-11 physics.ao-ph

Gyuchang Lim , JongJin Park This is my paper · ORCID

classification physics.ao-ph

keywords ArcticOscillationAir-seacouplingEast/JapanSeaSSTpersistenceMemorytimescalesOrnstein-UhlenbeckprocessMaximumCovarianceAnalysisMarineheatwaves

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks whether the Arctic Oscillation (AO) does more than shift mean winds over the East/Japan Sea: does it change how long the ocean 'remembers' atmospheric forcing in winter? Using 30 winters of daily SST and atmospheric reanalysis data, the authors show that the leading coupled pattern of SST and atmospheric variability explains 87% (positive AO) and 75% (negative AO) of the squared covariance, with hotspots in East Korea Bay and along the subpolar front. They then convert atmospheric principal components into exponentially weighted 'memory' series and find that the best-fitting memory scales cluster near 2-3 weeks for wind-stress curl, air temperature, and zonal wind, and 4-7 weeks for sea-level pressure and meridional wind, with longer memories under negative AO. The result matters because it turns the vague idea of air-sea coupling into concrete, phase-dependent predictor windows that could be used to forecast marine heatwaves and cold-surge-driven SST extremes at subseasonal lead times.

What carries the argument

The central object is an Ornstein-Uhlenbeck-like convolution: each atmospheric principal component is converted into an integrated response r_tau(t) by causal exponential smoothing with e-folding timescale tau, r_tau(t) = sum over past days of exp(-k/tau) times the atmospheric PC. The paper scans tau from 2 to 50 days, picks tau* that maximizes the zero-lag Pearson correlation with the SSTA principal component, and uses Detrended Fluctuation Analysis to compare the persistence of the integrated series with SSTA. This machinery turns the conceptual statement 'the ocean integrates weather' into a measured, phase-dependent number. It works together with rank-reduced multivariate Maximum Covaria

What would settle it

Split the 30 winters into training and held-out sets, estimate tau* for each AO phase on the training winters, then test whether the OU-integrated predictor with fixed tau* beats raw lead-lag correlations in predicting held-out SSTA. As a second check, smooth synthetic white noise with the same exponential kernel and run DFA: if pure white noise also yields H~1.3-1.4, the persistence validation is vacuous.

Watch

Extended reading notes

Core claim

The central claim is that AO polarity reorganizes the effective memory kernel of winter air-sea coupling in the East/Japan Sea, not just the mean forcing. Evidence: DFA shows SSTA persistence H~1.3-1.4 while atmospheric fields have H<1, consistent with a stochastic climate model in which the ocean mixed layer integrates fast atmospheric noise. Optimizing exponential memory timescales for the first MCA mode yields tau* ~18-25 days for wind-stress curl, 15-30 days for near-surface air temperature and zonal wind, and 30-50 days for sea-level pressure and meridional wind, with longer values under negative AO. The OU-integrated series acquire ocean-like persistence (H~1.2-1.4), which the authors

Load-bearing premise

The load-bearing premise is that a single-exponential Ornstein-Uhlenbeck memory kernel, with one fitted timescale tau, faithfully represents how the mixed layer accumulates atmospheric forcing; if the real coupling is multi-scale, nonlinear, or state-dependent, the reported tau* values and predictor windows are artifacts of the fitting procedure, and the subsequent DFA 'validation' does not test this assumption independently because exponential averaging itself creates ocean-

Editorial extensions

If this is right

Under positive AO, the most efficient predictors are 2-3 week OU-integrated wind-stress curl, air temperature, and zonal wind, matching the anticyclonic/downwelling warm response in the northwest basin and higher marine-heatwave propensity.
Under negative AO, sea-level pressure and meridional wind require 4-7 week integration, consistent with persistent north-westerly monsoon flow organizing front-parallel curl belts near 40-42N.
Raw lead-lag correlations are weak and sign-changing; only after mixed-layer integration do atmospheric PCs correlate strongly (r~0.4-0.7) with SSTA, so subseasonal prediction should use OU-integrated predictors rather than raw lags.
The leading coupled mode's high squared-covariance fractions (0.87 for +AO, 0.75 for -AO) and spatial loadings define phase-specific hot spots, East Korea Bay and the subpolar front, where extremes are most predictable.
DFA shows integrated atmospheric responses attain ocean-like persistence (H~1.2-1.4), supporting a stochastic-climate view of winter SSTA in the East/Japan Sea.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The DFA check may be weaker than it appears: exponentially smoothing any white-noise-like series tends to create reddened, ocean-like persistence by construction, so the reported H values for integrated atmospheric PCs do not independently confirm that the fitted tau* is the true coupling memory.
Because the tau grid stops at 50 days, the optimal integration times for sea-level pressure and meridional wind under negative AO may be truncated; if the correlation curves are still rising at tau=50, the '4-7 week' window could be a lower bound.
A natural extension is to test the predictor windows in cross-validated hindcasts: fit tau* on training winters, fix it, then see whether OU-integrated predictors beat raw lead-lag predictors on held-out winters for marine heatwave and cold-surge SST extremes.
The same OU-integration machinery could be transferred to other marginal seas or to summer seasons, where different forcing pathways (heat flux versus dynamical) would imply a different hierarchy of memory timescales.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper is worth a look if you care about regional air-sea coupling diagnostics, but read the quantitative claims with a skeptical eye. What is genuinely new: an AO-phase-stratified, multivariate MCA for the East/Japan Sea that maps the leading SSTA mode against five atmospheric fields, together with DFA persistence maps. The spatial patterns—East Korea Bay and subpolar-front hot spots, front-parallel curl belts, the SLP/wind/curl/SSH pathway—are consistent with Song et al. 2023 and extend it by quantifying phase-dependent persistence. That descriptive core is likely robust and useful for the regional community.

The soft spots are real, and they land on the central quantitative claims. The optimal integration time tau* is defined by argmax over tau in [2,50] days. Section 3.3.1 says SLP correlations 'peak near the end of the tested window' and reports tau* ~ 40-50 days; VA10 under -AO is tau* ~ 30-40 days, also near the upper end. When the maximum sits at the grid boundary, the true optimum is not identified—the claimed 4-7-week predictor window is a censored value, not a measured timescale. That is not a minor detail; it is the paper's headline. The DFA 'validation' in Section 3.3.2 is computed at a fixed tau=25 days rather than the field-specific tau*, and any exponential smoothing of atmospheric noise will raise the Hurst exponent toward the SSTA value by construction. It cannot independently confirm a Hasselmann memory kernel. Also, because the MCA is constructed to maximize covariance between a1 and b_j, the zero-lag correlations rho_j(tau) are in-sample fits; the paper offers no out-of-sample test, no autocorrelation-aware significance, and no sensitivity analysis for the AO threshold or saliency mask. The authors do acknowledge some of these limitations in Section 4.4, which is honest, but the acknowledgment does not fix the boundary issue.

The analysis is a regional statistical diagnostic, not a breakthrough. It deserves a serious referee because the descriptive material and the methodological framing are worth airing, but the memory timescales and predictor windows need either an extended tau grid, block-bootstrap confidence intervals, or out-of-sample validation before they should be cited as established. I would send this to peer review expecting major revision, and I would use it in a reading group as a case study in how fitted parameters and smoothing artifacts can masquerade as physical timescales.

Referee Report

4 major / 5 minor

Summary. The manuscript analyzes 30 winters (JFM 1993–2022) of daily OISST and ERA5 data over the East/Japan Sea to quantify how Arctic Oscillation polarity modulates the persistence, spatial coupling, and effective memory of winter sea-surface temperature anomalies. The analysis combines DFA-based Hurst exponents, rank-reduced multivariate MCA of SSTA against five concatenated atmospheric fields, and an OU-like exponential convolution of atmospheric principal components with e-folding time τ. For each atmospheric field and AO phase, τ* is selected by maximizing the absolute zero-lag correlation with the leading SSTA PC (Eq. 22), and the resulting integrated series are then characterized by DFA. The paper reports leading-mode squared covariance fractions of 0.87 (+AO) and 0.75 (−AO), SSTA hot spots in East Korea Bay and along the subpolar front, τ* values of roughly 18–25 d for curl, 15–30 d for ATMP/U10, and 30–50 d for SLP/V10 (longer under −AO), and DFA exponents H≈1.3–1.4 for SSTA with integrated atmospheric series reaching similar persistence. These diagnostics are used to propose AO-phase-specific predictor windows for subseasonal extremes such as marine heatwaves and cold-surge-impacted SST.

Significance. If the identification of τ* were robust, the paper would provide a useful, quantitative extension of Hasselmann’s stochastic climate framework to a marginal sea, with phase-dependent memory timescales that could inform forecast model construction. The rank-reduced multivariate MCA is an appropriate response to the T≪N sampling problem, and the authors are transparent about the linear, single-timescale nature of their model and its limitations. However, the central quantitative claims — the specific τ* windows and their validation by DFA — currently rest on a censored argmax at the boundary of the τ grid, an in-sample correlation fit, and a DFA check that is partly a smoothing artifact. These issues need to be addressed before the reported memory timescales and predictor windows can be considered established.

major comments (4)

[Section 3.3.1 / Eq. (22)] The identification of τ* is censored for the very fields that carry the strongest claims. For SLPA the text states correlations 'increase more gradually, peaking near the end of the tested window' with τ*≈40–50 d, and for VA10 under −AO τ*≈30–40 d, near the upper end of the grid τ∈[2,50]. If the maximum lies at τ=50 (or near it), the reported '4–7-week' predictor window is a boundary value, not an identified timescale. The paper should extend the grid well beyond 50 d (e.g., to 120 d), report whether the argmax is interior, and also report the plateau-onset statistic τ_s defined in §2.2.4.3, which is currently never used. Without this, the central quantitative claim is not supported.
[Section 3.3.2 / Figure 6] The DFA 'validation' does not validate the fitted τ*. The integrated series are computed at a fixed τ=25 d, not at the field-specific τ* values reported in §3.3.1, so the DFA exponents in Figure 6 are not tied to the memory windows being claimed. More importantly, any low-pass exponential smoothing of atmospheric noise raises the DFA exponent toward the SSTA value by construction; it is a property of the smoothing operation, not independent evidence for the Hasselmann/OU kernel. The authors should include a null comparison — e.g., DFA of exponentially filtered white noise with the same τ and sample length — and show that the observed integrated-atmosphere exponents are distinguishable from that null. As written, the sentence 'these DFA results confirm that our OU-like transform yields forcing series with SSTA-like memory' is circular.
[Section 2.2.4 / Eq. (22)] The correlation ρ_j(τ) is optimized in-sample over τ, and the atmospheric PC b_j and SSTA PC a_1 are obtained from the same MCA decomposition, so their zero-lag covariance is nonzero by construction. The reported τ* values therefore reflect an in-sample fit, not an out-of-sample coupling measure. No confidence intervals, block-bootstrap, or cross-validation are provided, and the AO-phase contrasts in τ* are not tested for significance given the short, autocorrelated winter segments. The authors should add uncertainty quantification (e.g., bootstrap over winters) and an out-of-sample or split-sample check before interpreting the τ* hierarchy as a physical memory scale.
[Section 2.1.4] The cut-and-stitch concatenation of JFM segments creates artificial discontinuities at each March–January boundary. These jumps affect DFA at scales comparable to or longer than the segment length (the analysis uses scales up to 90 d) and also contaminate the exponential convolution in Eq. (21) for lags crossing year boundaries. The paper should quantify the sensitivity of both H and τ* to this construction, for example by truncating cross-boundary lags, padding with climatology, or analyzing within-winter pieces separately. Without such a check, the persistence and memory estimates may be biased by the stitching procedure.

minor comments (5)

[Section 2.2.1] The DFA scale range is inconsistent: the text says segment scales 10 ≤ s ≤ 90 days, then later says the fit is over 5 ≤ s ≤ 45 days, while Figure captions refer to 10–90-day scales. Please clarify the exact fitting range and caption conventions.
[Section 3.3.1] The bullet results report r values and τ* ranges without any uncertainty estimates or significance tests. Adding confidence intervals (even from a simple block bootstrap) would materially strengthen the AO-phase comparisons.
[Section 2.2.4.1] The cross-references to equation numbers are off: the text says 'Eq.16 is the Green-function of Eq.15' and 'Eq.19' for the discrete implementation, but the relevant equations are numbered (20) and (21). Please renumber consistently.
[Section 2.2.4.1] The phrase 'OU-like' is appropriate, but the model with σ=0 is a deterministic exponential moving average, not an OU process. This should be stated more prominently so that the reader understands that the 'stochastic climate model' language is heuristic.
[General] The manuscript contains 'FIGURE 1 TO BE INSERTED' placeholders. These should be replaced with actual figures before submission; similarly, several figure captions state 'TO BE INSERTED' in the legend section.

Circularity Check

2 steps flagged · score 6.0 of 10

The τ* memory windows are fitted in-sample by Eq. 22 (with SLP peaking at the grid edge), and the DFA 'validation' (§3.3.2) re-measures the smoothing built into the OU kernel; the central Hasselmann claim is partially circular while MCA/DFA descriptive results remain independent.

self definitional [Section 3.3.2 (Persistence diagnostics from DFA), with Methods Eqs. 19–21 and §2.2.1]
"We applied DFA to the SSTA PC and to each integrated atmospheric PC (here with Δτ = 25 days). ... These DFA results confirm that our OU-like transform yields forcing series with SSTA-like memory, lending methodological validity to using τ* (estimated from Figure 5) as process-based predictors in subsequent statistical/dynamical prediction experiments."

The integrated series r_{τ,j}(t) is defined by Eqs. 19–21 as an exponential low-pass convolution of b_j with memory τ. A single-pole filter necessarily reddens its output (spectral roll-off toward f^{-2} at scales ≤ τ); with τ=25 d and DFA window ~5–45 d, H on the filtered series is a mathematical property of the kernel, not an independent ocean–atmosphere signal. The paper's own assumption is 'the oceanic field, SSTA, is an integration of white-noise like atmospheric fields' (§2.2.1), and the OU transform is exactly that integration. Thus DFA 'confirm[ing] that our OU-like transform yields forcing series with SSTA-like memory' is a tautology: it re-measures the smoothing imposed by Eq. 21 and cannot independently validate Hasselmann's framework. It also uses fixed τ=25 d, not the field-sp
fitted input called prediction [Section 2.2.4.3 (Eq. 22), §2.2.4.4, §3.3.1 (SLPA bullet), Abstract]
"ρj(τ) = corr(a1(t), rτ,j(t)), τj* = arg max τ∈[2,50] |ρj(τ)| ... τj* ... is interpreted as the effective memory over which the mixed layer accumulates forcing from field j to produce the leading SST anomaly mode. ... SLPA (sea-level pressure): Correlations increase more gradually, peaking near the end of the tested window; τ* ≈ 40–50 days."

τ* is defined (Eq. 22) as the τ that maximizes the in-sample zero-lag correlation between the SSTA PC a1 and the filtered atmospheric PC r_{τ,j}. The reported 'characteristic memory timescales' are therefore fitted parameters, chosen to match the SSTA PC; the Abstract's 'Zero-lag correlations ... reveal characteristic memory timescales (τ*)' renames the fitted argmax as a physical memory. The high correlation values (r~0.4–0.7) are in-sample and double-optimized (MCA mode + τ scan), not independent coupling measurements. For SLP the argmax sits at the grid edge ('peaking near the end of the tested window'), so the headline 4–7-week 'predictor window' is a censored boundary value whose true optimum may lie beyond 50 days; the paper still reports this as a memory timescale and builds predict

full rationale

The paper's descriptive core — MCA SSTA/atmospheric loading maps, DFA persistence maps (H>1), and AO-phase stratification — is a self-contained analysis of OISST/ERA5 data and is cross-checked against externally published composites (Song et al. 2023), so that portion is not circular. The circularity is concentrated in the quantitative memory-timescale claim. Step 1: the DFA 'validation' (§3.3.2) is not an independent test because the OU-integrated PCs are defined (Eqs. 19–21) as exponential low-pass convolutions; any such single-pole filter reddens its output, and measuring H≈1.2–1.4 on the filtered series merely re-measures the smoothing built into the transform (at fixed τ=25 d). This cannot confirm Hasselmann's framework, which the paper already assumed in §2.2.1. Step 2: τ* is fitted as the argmax of the zero-lag correlation between the SSTA PC and the filtered atmospheric PC (Eq. 22); the paper then presents these fitted values as 'revealed' characteristic memory timescales and as 'process-based predictor windows' with no out-of-sample validation, and for SLP the fitted argmax lies at the edge of the 2–50-day grid, making the headline 4–7-week window a censored boundary value. The self-citations to Lim & Park 2024a/b are contextual (full-year DFA crossover) and are not load-bearing; no uniqueness theorems are invoked. The §4.4 limitation statement ('linear, single-time-scale OU kernel ... does not resolve potential multi-time-scale kernels, nonlinearity ...') corroborates that the 'validation' operates inside the model's own ansatz. The ChatGPT acknowledgment (§8) asserts writing-only use and does not affect the statistical derivations. Overall: the τ*/DFA validation chain partially reduces by construction (score 6); the spatial and persistence descriptive results remain independent.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The quantitative claims rest on several chosen parameters and domain assumptions. The optimal tau is inherently fitted to the target SSTA PC; the AO threshold and EOF fraction are user choices; the DFA range is inconsistently reported. The Hasselmann interpretation requires assuming a single-exponential integration process with no stochastic component.

free parameters (5)

optimal integration time tau_j* = 18-25 days (CurlTau), 15-30 days (ATMP/UA10), 30-50 days (SLP/VA10)
Chosen to maximize zero-lag correlation with SSTA PC (Eq. 22); not independently derived or out-of-sample tested.
AO phase threshold = +/-0.8 standard deviations of JFM AO index
Arbitrary threshold; no sensitivity analysis is provided to test whether results depend on the chosen cutoff.
EOF truncation fraction gamma = 0.90
Retains 90% variance in each field before MCA; affects which modes enter the coupled analysis.
DFA scale range = 10-90 days (text) and 5-45 days (Eq. 4 context), inconsistent
The fitted Hurst exponent depends on the chosen range; the paper does not reconcile the two ranges.
Saliency threshold = Area-weighted mean absolute loading
Post hoc effect-size filter for MCA loadings; not a formal significance test, and it shapes the reported spatial patterns.

assumptions (4)

domain assumption Sea surface temperature anomalies behave as an integrated response to atmospheric noise (Hasselmann 1976).
Adopted as the guiding framework in the Introduction; the entire OU model and persistence interpretation rely on it.
ad hoc to paper A single-exponential (OU) kernel with zero stochastic forcing describes the air-sea integration.
Equation 19 sets sigma=0 and uses one exponential time constant; no justification that this kernel captures multiple timescales or nonlinearity, as acknowledged in Section 4.4.
domain assumption Concatenating JFM segments from different years yields a continuous time series for DFA and correlations.
Section 2.1.4 stitches winter segments together; this creates artificial year-boundary jumps that can inflate persistence estimates at scales near the segment length.
domain assumption Daily samples within a winter are treated as statistically independent for correlation and MCA.
Autocorrelation in daily atmospheric and oceanic fields inflates the effective sample size; no significance testing accounts for this.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Arctic Oscillation Modulation of Winter Air-Sea Coupling in the East/Japan Sea: Persistence, Timescales, and Extremes." pith.science (2026). https://pith.science/paper/HLKFAWT7

@misc{pith2026250909628,
  author       = {Pith},
  title        = {Pith review of: Arctic Oscillation Modulation of Winter Air-Sea Coupling in the East/Japan Sea: Persistence, Timescales, and Extremes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HLKFAWT7}},
  note         = {Machine review of arXiv:2509.09628}
}

abstract

The winter climate of the East/Japan Sea (EJS) is strongly affected by the Arctic Oscillation (AO), yet how AO polarity reshapes the memory, coupling patterns, and predictability of sea-surface temperature anomalies (SSTA) remains poorly quantified. Using 30 winters (1993--2022) of daily OISST and ERA5 fields, we combine multivariate Maximum Covariance Analysis (MCA) with an Ornstein--Uhlenbeck (OU)-like integration of atmospheric principal components (PCs). The leading coupled mode explains 87% (+AO) and 75% (-AO) of squared covariance, with SSTA hot spots in East Korea Bay and along the subpolar front. Zero-lag correlations between the SSTA PC and OU-integrated atmospheric PCs reveal characteristic memory timescales ($\tau$) of $\sim$18--25 days for wind-stress curl (CurlTau), $\sim$15--30 days for near-surface air temperature (ATMP) and zonal winds, and $\sim$30--50 days for sea-level pressure (SLP) and meridional winds -- longer under -AO. Detrended Fluctuation Analysis (DFA) shows SSTA persistence $H \approx 1.3$--$1.4$ and that integrated atmospheric responses acquire ocean-like persistence, validating Hasselmann's stochastic framework for winter EJS. AO-phase contrasts align with a curl$\rightarrow$Ekman pumping$\rightarrow$eddy/SSH$\rightarrow$SST pathway: +AO favors anticyclonic/downwelling responses and warmer SSTA, whereas -AO favors cyclonic/upwelling and cooler SSTA. These diagnostics identify phase-specific predictor windows (e.g., 3-week OU-integrated CurlTau/ATMP; 4--7-week SLP/V-wind under -AO) to initialize subseasonal extremes prediction (marine heatwaves and cold-surge-impacted SST). The approach quantifies memory scales and spatial coupling that were not explicitly resolved by previous composite analyses, offering a tractable foundation for probabilistic forecast models.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Wintertime Cross-correlational Structures between Sea Surface Temperature Anomaly and Atmospheric-and-Oceanic Fields in the East/Japan Sea Under Arctic Oscillation
physics.ao-ph 2025-09 conditional novelty 5.0 of 10

Over 30 winters, persistent warmth in the East/Japan Sea lines up with sea-surface height and along-front currents, while winds and heat fluxes act only as short-lived, non-persistent forcing.

Reference graph

Works this paper leans on

2 extracted references · cited by 1 Pith paper

[913]

Interannual to interdecadal variability in the Japan Sea based on a new gridded upper water temperature dataset

https://doi.org/10.1002/cpa.1014 Minobe, S., Sako, A., Nakamura, M. Interannual to interdecadal variability in the Japan Sea based on a new gridded upper water temperature dataset. J. Phys. Oceanogr. 34, 2382–2397 (2004). NOAA CPC (2023). Arctic Oscillation (AO) Index. National Oceanic and Atmospheric Administration Climate Prediction Center. [Online] Ava...

doi:10.1002/cpa.1014 2004
[2022]

cut-and-stitch

of daily OISST and ERA5 fields, we combine multivariate Maximum Covariance Analysis (MCA) with an Ornstein–Uhlenbeck (OU)-like integration of atmospheric principal components (PCs). The leading coupled mode explains 87% (+AO) and 75% (−AO) of squared covariance, with SSTA hot spots in East Korea Bay and along the subpolar front. Zero-lag correlations betw...

1999

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [913]

Interannual to interdecadal variability in the Japan Sea based on a new gridded upper water temperature dataset

https://doi.org/10.1002/cpa.1014 Minobe, S., Sako, A., Nakamura, M. Interannual to interdecadal variability in the Japan Sea based on a new gridded upper water temperature dataset. J. Phys. Oceanogr. 34, 2382–2397 (2004). NOAA CPC (2023). Arctic Oscillation (AO) Index. National Oceanic and Atmospheric Administration Climate Prediction Center. [Online] Ava...

doi:10.1002/cpa.1014 2004

[2] [2022]

cut-and-stitch

of daily OISST and ERA5 fields, we combine multivariate Maximum Covariance Analysis (MCA) with an Ornstein–Uhlenbeck (OU)-like integration of atmospheric principal components (PCs). The leading coupled mode explains 87% (+AO) and 75% (−AO) of squared covariance, with SSTA hot spots in East Korea Bay and along the subpolar front. Zero-lag correlations betw...

1999