Impacts of Extratropical-Cyclone Extreme Events on SST and Mixed-Layer Depth over the Kuroshio Extension

Abstract

Wintertime extratropical cyclones frequently traverse the Kuroshio–Oyashio Extension frontal system. However, their net impacts on synoptic sea-surface temperature (SST) variability and mixed-layer structure remain uncertain in the presence of strong fronts and intrinsic ocean variability. Using reanalysis data, we classify extreme events into cyclone cold-sector and warm-sector types based on synoptic air–sea flux anomalies. With ensembles of single-column model experiments, we decompose the upper-ocean response into surface heat-flux forcing, wind-driven mechanical mixing, Ekman temperature advection, wave-breaking mixing, and freshwater effects. Cold-sector events amplify synoptic SST variability and deepen the mixed layer, whereas warm-sector events suppress SST variability and shoal the mixed layer. Surface heat flux is the primary driver of both responses. Ekman advection provides crucial modulation within the frontal zone. Wave-breaking mixing generally damps temperature perturbations. Freshwater forcing exerts a pronounced regional influence southeast of the subarctic front. The combined effects yield an asymmetric spatial fingerprint on SST variability and mixed-layer depth across the frontal system. Comparison between forced variability and total reanalysis variability indicates that within the frontal zone, atmospheric impacts can be redistributed or partly offset by intrinsic ocean processes, while outside the frontal zone, the behavior is closer to an externally forced response.

1. Introduction

The Kuroshio Extension and the Oyashio Extension converge east of Japan, forming the Kuroshio–Oyashio Extension (KOE) region and several intense oceanic fronts, including the Kuroshio Extension Front (KEF) and the Subarctic Front (SAF). This region lies near the entrance of the North Pacific storm track, where transient baroclinic storms are frequent in winter [1,2]. It is also characterized by a strong western boundary current background, a front–jet structure, and vigorous mesoscale eddy activity [3,4]. The strong SST gradients maintained by the frontal zone enhance near-surface atmospheric baroclinicity and can shift the storm track [2,5,6], making this region one of the most prominent hotspots for coupling between midlatitude atmospheric transients and intrinsic ocean dynamics [7]. Two-way air–sea interaction over the KOE plays a key role in local and remote climate variability, forming quasi-decadal oscillations [8,9] that strengthen under global warming [10] and may contribute to the decadal variability of northern-hemispheric average temperature [11].

Midlatitude air–sea interaction is typically time-scale dependent: from synoptic to interannual scales, the atmosphere mainly forces the ocean through surface heat fluxes and wind stress; whereas on longer time scales the ocean feeds back via SST anomalies and modulates atmospheric circulation [12,13,14]. This framework is supported in many midlatitude regions [15,16]. However, western boundary current frontal zones may deviate from this canonical picture. Wu et al. [17] showed that winter sub-seasonal SST variability over the Kuroshio Extension is largely controlled by intrinsic ocean advection and eddy processes, with a much smaller contribution from direct atmospheric forcing than outside the frontal zone. Bishop et al. [18] further reported distinctive, scale-dependent heat-flux feedback characteristics in frontal regions. These findings indicate that under strong fronts, the relative roles of atmospheric forcing and intrinsic ocean processes, and how they couple, remain to be clarified.

On sub-seasonal variability, synoptic (2–10-day) extremes are an important component. A small number of extreme turbulent heat-flux events can account for most of the winter turbulent heat flux and its variance [19,20,21], and they are often associated with extratropical cyclones and cold-air outbreaks [22,23]. Observations and reanalysis also show that strong winds and intense heat exchange can markedly alter mixed-layer structure and modulate SST within a few days [24,25]. Yet existing work has largely emphasized flux statistics or composite responses, lacking a systematic, event-scale decomposition of the relative contributions from heat fluxes, wind-driven advection, and vertical mixing processes, especially in midlatitude western boundary current frontal zones where intrinsic ocean variability is strong.

By contrast, the upper-ocean response to tropical cyclones has been studied more systematically. Strong winds can rapidly adjust the mixed layer through turbulent mixing, inertial responses, and upwelling/entrainment, accompanied by pronounced surface cooling and restructuring of temperature–salinity profiles [26]. Because tropical cyclones are compact and their wind fields are approximately axisymmetric, the ocean response often appears as a relatively localized cold wake.

Extratropical cyclones draw energy primarily from baroclinic instability; they have larger horizontal scales, cold and warm fronts, and more asymmetric structures. The wind fields, cloud–radiation, and turbulent heat-flux distributions differ markedly between the cold and warm sectors, so their oceanic impacts need not mirror those of tropical cyclones. Compared with tropical cyclones, systematic studies of extratropical cyclone impacts on the upper ocean remain limited. Kobashi et al. [27] reported that extratropical cyclones over a subtropical North Pacific frontal region can cool the sea surface through reduced shortwave radiation and enhanced latent heat loss, but they did not distinguish cold- and warm-sector forcing nor quantitatively separate contributions from local heat fluxes, wind-driven advection, and vertical mixing. For regions like the Kuroshio Extension, where fronts are strong and intrinsic ocean variability is already pronounced, it remains unclear to what extent extreme cyclone forcing alters synoptic SST variability and whether the effect is modulated by background advection and eddy activity.

More importantly, in strong western boundary current frontal zones, atmospheric events do not act on a stationary background SST field but on a field superposed with high-frequency intrinsic ocean variability dominated by mesoscale eddies and advection. When strong background variability coexists with strong forcing, the net impact of extremes on local SST variability may no longer be a simple superposition; instead, it may be partly offset or redistributed by intrinsic ocean processes. Clarifying this interaction is essential for understanding synoptic air–sea coupling in midlatitude frontal zones.

Against this backdrop, we focus on the wintertime Kuroshio Extension–Oyashio Extension frontal system and adjacent waters during 1992–2018. We identify extreme events corresponding to cyclone cold and warm sectors, characterize the upper-ocean response using ECCO2 reanalysis, and use a 1-D turbulence model (GOTM) to separate the relative contributions from surface heat fluxes, wind-driven Ekman advection, mechanical mixing, wave-enhanced mixing, and freshwater flux. By “upper-ocean response”, we primarily mean SST (or mixed-layer-mean temperature) synoptic variance and mixed-layer depth changing rate. Through an event-scale dynamical–thermodynamical decomposition and regional comparisons, we aim to elucidate the primary pathways and regional differences through which extratropical-cyclone extreme forcing affects the upper ocean, and to further discuss the coupling between atmospheric forcing and intrinsic ocean processes.

The remainder of this paper is organized as follows. Section 2 describes data, model, and experimental design. Section 3 presents the results, including the characteristics of cold- and warm-sector extreme events, the upper-ocean responses, the contributions of different physical processes, the contrasts and net effects between the two event types, and the differences and relation between forced and total responses, including intrinsic variability. Section 4 summarizes the main conclusions, and Section 5 discusses implications and limitations.

2. Data and Model

2.1. Reanalysis Data

All oceanic variables and atmospheric forcing fields used here are taken from the global ocean–sea-ice state estimate ECCO2 (Estimating the Circulation and Climate of the Ocean, Phase II; [28]). ECCO2 is built on the MITgcm and assimilates multi-source observations under dynamical constraints in a least-squares sense, thereby reducing systematic biases while maintaining dynamical consistency among variables. We use daily mean fields with 0.25° × 0.25° horizontal resolution and 50 vertical levels for 1992–2018 over 25.5–50° N, 130.5–180° E. Variables include net surface heat flux (positive upward), wind stress, freshwater flux (positive downward, defined as precipitation minus evaporation), temperature, and salinity. ECCO2 has been justified for studying eddy dynamics in various regions, including the Kuroshio Extension [29,30,31]. Its resolution and accuracy are sufficient for our investigation.

2.2. GOTM

The General Ocean Turbulence Model (GOTM) is a 1-D vertical water-column model [32]; we use version 6.0.0. It solves 1-D vertical transport–diffusion equations for momentum, temperature, and salinity coupled with turbulence-closure parameterizations to represent key processes such as wind-driven mixing, convective mixing, and shear instability, and is widely used to study the rapid evolution of the mixed layer and thermocline. We employ a second-order k–ε closure to compute vertical eddy viscosity and diffusivity [33] and include the Craig and Banner wave-breaking parameterization, which injects a turbulent kinetic energy (TKE) flux determined by friction velocity into the surface boundary condition of the TKE equation to represent wave-induced mixing [34]. GOTM has been extensively applied and evaluated for mixed-layer studies [35,36,37]; therefore, we do not perform an additional independent validation here.

2.3. Experimental Design

GOTM does not explicitly include horizontal advection or eddy transport, which allows the upper-ocean response to local atmospheric forcing to be largely separated from 3-D dynamical effects. We partitioned the study region into 578 independent water columns on a 1.25° × 1.25° grid and ran GOTM for each column without horizontal exchange between columns. Vertically, the integration depth for each column is set to twice the local winter-mean mixed-layer depth and discretized into 100 uniform layers; initial temperature–salinity profiles and mixed-layer depth are prescribed from ECCO2. This ensemble single-column framework follows Wu et al. [17].

Diagnostically, the tendency of SST (or mixed-layer mean temperature) can be written as a combination of a surface heat-flux term, an entrainment/mixing term, and an advection term (equations omitted). The net heat-flux term depends on the net air–sea heat flux, seawater density, specific heat capacity, and mixed-layer depth. The mixing term represents stratification adjustment and entrainment/vertical mixing driven by wind stress and wave breaking. The advection term includes wind-driven Ekman temperature advection, background advection, and vertical temperature advection associated with Ekman pumping. In addition to changing mixed-layer heat content, surface heat flux also modulates mixing and entrainment by altering stratification stability. Freshwater flux affects mixing by modifying salinity and density stratification and can indirectly affect the heat-flux term through changes in density and mixed-layer depth; we do not separately account for the direct thermal input of rainfall to temperature. Wave-induced mixing is represented using the Craig and Banner scheme [34], in which the upper boundary condition for the TKE equation is represented by a TKE flux estimated from friction velocity:

$$F_k=\eta u_{\ast }^3$$

The parameter $\eta$ is set to 100.

Ekman temperature advection is not produced internally by the single-column model; instead, it is estimated for each column by ${\mathit{u}}_E\cdot \nabla T$, i.e., combining the Ekman velocity ${\mathit{u}}_E$ computed from wind stress with the background horizontal temperature gradient $\nabla T$ from the ECCO2. Here

$${\mathit{u}}_E\left(x,y,z,t\right)={\displaystyle \frac{1}{\rho \sqrt{2fA}}}\exp \left({\displaystyle \frac{z}{D}}\right)\mathit{R}\left(-{\displaystyle \frac{z}{D}}\right)\mathit{\tau }\left(x,y,t\right)$$

where $\mathit{\tau }$ is the wind stress and

$$\mathit{R}\left(\theta \right)=\left[\begin{array}{cc}\cos \theta & -\sin \theta \\ \sin \theta & \cos \theta \end{array}\right]$$

is a rotation matrix. $D=\sqrt{2A/f}$ is the frictional depth, and $A$ = 0.05 m²/s is the viscosity. $\rho$ = 1025 kg/m³ is a reference density, and $f$ is the Coriolis parameter. SST gradient is calculated from the SST field simulated at each time step.

Background advection cannot be represented by the single-column experiments either, but it can be qualitatively inferred from the difference between the SST tendency in ECCO2 and the purely forced variability simulated by GOTM; this residual also contains uncertainties arising from reanalysis and model errors (numerical and parameterization) and the lack of coupled feedbacks and is therefore used only for qualitative discussion. Vertical temperature advection induced by Ekman pumping is estimated from wind-stress curl and the background vertical temperature gradient; tests show that its magnitude is much smaller than other terms and it is neglected in the analyses below.

Numerical integration is forced by the ECCO2 daily mean atmospheric forcing. For each year during 1992–2018, the model is integrated from 1 October to 28 February of the following year with a 1 h time step. Because October is mainly for spin-up, we only use outputs after 1 November for statistical analyses. On this basis, for each extreme event we conduct an independent short-term simulation from the event start date to one day after the event ends to capture the lagged SST response; a 1-day lag is adopted because the correlation between event intensity and SST peaks at a 1-day lag.

To isolate contributions from different forcings, we designed a set of stepwise removal experiments (

Table 1. Experimental design and forcing treatment.

Experiment	Ekman Advection	Wave-Breaking Mixing	Net Heat-Flux	Wind Stress	Freshwater Flux
FORCED	On	On	Original	Original	Original
FE	Off	On	Original	Original	Original
FEW	Off	Off	Original	Original	Original
FEWH	Off	Off	Linear	Original	Original
FEWHW	Off	Off	Linear	Linear	Original
FEWHWF	Off	Off	Linear	Linear	Linear

). Note that in this framework, it is the forcing terms that are separated, not the oceanic processes that they drive. The FORCED experiment is a full experiment retaining all forcings and considered processes. Based on FORCED, the FE experiment removes Ekman advection to represent a purely local response. The FEW experiment further removes wave-induced mixing, so that the response is driven only by local flux forcing. For net heat flux, wind stress, and freshwater flux, removing synoptic anomalies means replacing the original forcing within the event window by a linearly interpolated sequence connecting the pre- and post-event forcings, thereby suppressing flux extremes while retaining the slowly varying background during the events. This yields FEWH, FEWHW, and FEWHWF experiments to isolate, respectively, the extreme contributions from net heat flux, wind stress, and freshwater flux. Because the linearly interpolated forcing still induces some ocean adjustment, we take FEWHWF as the baseline state, and all other responses are expressed as anomalies relative to this baseline.

3. Results

3.1. Atmospheric Forcing Characteristics of Cold- and Warm-Sector Extreme Events

We use synoptic net surface heat-flux anomalies to characterize extreme atmospheric forcing. Net heat flux (NHF; positive upward) consists of turbulent heat fluxes and radiative heat fluxes; the former includes sensible and latent heat fluxes, and the latter includes shortwave and longwave radiation. Over the Kuroshio Extension and adjacent waters in winter, anomalous air–sea exchange is often associated with extratropical cyclones and accompanying cold-air activity, which can trigger strong winds, enhanced evaporation, or heavy precipitation on day-to-few-day time scales and substantially alter the upper-ocean heat budget [19,22,25]. Meanwhile, cyclone cloud systems can modulate both net radiation and turbulent fluxes to influence surface cooling, and longwave radiation is an important damping mechanism for ocean temperature, making radiative terms non-negligible at the event scale [27]. Therefore, compared with the commonly used turbulent fluxes alone, NHF is more suitable for identifying and characterizing extreme forcing in this study.

Specifically, we apply a 10-day high-pass filter to NHF to obtain NHF′, highlighting transient signals on 2–10-day time scales. For each grid point, the 80th percentile of winter (November–February, NDJF) NHF′ is used as a threshold: days when NHF′ exceeds this threshold are defined as extreme cold-flux days. Consecutive days are merged into a single event, and the event center is defined as the day when NHF′ reaches its maximum, corresponding to the most intense ocean heat loss. Similarly, NHF′ in the lower 20% of the winter distribution (strong negative anomalies) defines extreme warm heat-flux events. Note that under the winter background, the ocean is still, on average, losing heat; thus, an extreme warm event corresponds to a pronounced weakening of heat loss rather than net ocean heat gain.

In order to examine how the two extreme heat-flux event types correspond to cyclone structure, we composite the high-pass-filtered surface wind stress and freshwater flux (P−E; positive downward) with respect to the NHF′ event-center day (Figure 1). On cold-event days, the target point is located in the southwest quadrant of the cyclone, with strong northerly winds and markedly enhanced evaporation (negative P−E anomalies), indicating rapid transitions in momentum and buoyancy forcing. Following classical extratropical cyclone anatomy [7], this is in the cold sector of the cyclone, characterized by southwesterly winds. Warm events correspond to the passage of the cyclone warm sector and typically feature strong winds with enhanced precipitation, while the upward heat flux is substantially reduced. These composite features are consistent with observational and reanalysis results linking anomalous flux events and cyclonic activity over the KOE region [22,23,38].

Based on the above physical correspondence, we further screen extreme heat-flux events to identify cold- and warm-sector extreme events associated with extratropical cyclones. Cold-sector events require that extreme upward heat flux, strong wind stress, and strong evaporation occur simultaneously. Warm-sector events require that markedly weakened upward heat flux coincides with strong wind stress and heavy precipitation. Overall, during the 27 years (1992–2018), each grid point experiences on average about 3–6 cold-sector and 3–6 warm-sector events per winter (Figure 2a,b), most lasting 1–3 days (Figure 2c). Cold-sector events occur most frequently over 30–45° N in the Sea of Japan and the western side of the western North Pacific, whereas warm-sector events are more common over 35–50° N and are biased toward the downstream region. This difference is consistent with the typical meridional configuration of cyclone cold and warm sectors and with flux asymmetry caused by SST being substantially higher than air temperature near the western boundary current.

To quantify event intensity, we compute event-scale anomaly amplitudes of NHF, wind stress, and freshwater flux and take the extrema within each event window as the event intensity, yielding the multi-winter mean spatial distributions for the three flux intensities (Figure 3a–f). Using the climatological winter-mean SST gradient (Supplementary Figure S1), we extract the KEF and SAF location (black curves in Figure 2) and define the region between them as the KOE. The fronts move on seasonal and interannual scales but are treated as fixed during winter. The regions north of the SAF and south of the KEF are referred to as the subpolar and subtropics, respectively. Results show that for both cold- and warm-sector events, NHF intensity peaks along the KEF; wind-stress intensity exhibits a northeast–southwest gradient; and evaporation and precipitation intensities are strongest south of the SAF and near the KEF (Figure 3e,f). Further decomposing heat-flux anomalies into shortwave heating and non-solar terms (sensible, latent, and longwave) indicates that although cold-sector events are often less cloudy, non-solar terms dominate in amplitude (Supplementary Figure S2), implying that synoptic NHF anomalies are mainly contributed by turbulent and longwave processes.

We normalize the winter-mean intensities of the three fluxes and average them to define a hybrid event-intensity index (Figure 3g,h). Although the occurrence-frequency patterns differ between cold- and warm-sector events (Figure 2a,b), their hybrid intensity spatial structures are highly similar: the KOE region between the KEF and SAF is much stronger than regions to the south and north, consistent with the notion that extreme flux events are more readily organized and enhanced along frontal zones [19,20]. The spatial pattern of the hybrid intensity resembles that of the North Pacific storm track (contours in Figure 3g,h). Averaging flux intensities, the hybrid intensity, and the storm-track intensity over the strong-frontal KOE region yields 27 wintertime time series (Supplementary Figure S3); the hybrid intensity correlates with the storm-track intensity at over 0.7. This indicates that the identified extremes are primarily modulated by extratropical cyclone activity and are intensified along the fronts, consistent with the physical basis that strong SST gradients can maintain and enhance near-surface baroclinicity, promoting transient baroclinic eddy development and influencing storm-track intensity and position [3,5,6].

3.2. Cold-Sector Events: Roles of Ekman Advection, Wave Breaking, and Flux Forcing

In this section, we investigate oceanic response to cold-sector-event forcing based on GOTM experiments. The forced response is separated into non-local and local parts, the latter of which is further divided into wave effects and flux effects. Section 3.3 splits flux effects into the heat, freshwater, and wind stress components, and Section 3.4 compares oceanic responses under warm- and cold-sector event forcings. The response is examined in terms of wintertime synoptic temperature variability (STV) and mixed-layer depth changing rate. STV is defined as the variance of 10-day high-pass filtered SST, and mixed-layer depth is represented by the depth where the vertical density gradient exceeds 0.01 kg/m³/m [39,40].

Figure 4a shows the relative contribution of cold-sector-event STV to the winter-mean STV (regardless of in or out of the event windows) in the FORCED experiment. Even without background advection induced by eddy or other dynamic processes, extreme events can generate STV comparable to or greater than the winter-mean level over the KOE and subpolar regions, with a banded maximum along the KEF. The mixed-layer-depth response (Figure 4b) shows pronounced deepening over the KOE and south of the KEF, and slight shoaling over much of the subpolar and subtropical regions. Relative to the climatological mixed-layer depth (Supplementary Figure S4), deepening is concentrated around the KEF, where the climatological mixed layer is shallow, forming a circumfrontal pattern. Note that although FORCED excludes background advection, it still includes the nonlocal wind-driven Ekman advection and local processes driven by NHF, freshwater flux, and vertical mixing. In a pure local perspective, the co-occurrence of enhanced STV and mixed-layer deepening during cold-sector events implies that STV changes cannot be explained solely by deeper mixed layers, which would otherwise dilute surface anomalies and reduce STV. Possible explanations include: (i) STV and mixed-layer depth are driven by different atmospheric processes and may vary concurrently without direct causality; (ii) enhanced STV reflects stronger surface cooling, and cooling-induced convection deepens the mixed layer; or (iii) Ekman advection simultaneously modulates SST and stratification, altering the relationship between STV and mixed-layer depth. We use sensitivity experiments to separate nonlocal and local contributions to disentangle these mechanisms.

Comparing FORCED and FE isolates the Ekman-advection effect (Figure 4c,d), whereas FE represents the purely local response after removing Ekman advection (Figure 4e,f). In cold-sector events, near-surface winds are predominantly northwesterly (Figure 1), and the associated Ekman transport crosses the frontal temperature gradient, tending to produce cold advection within the mixed layer and weaken stratification stability. Results show that Ekman advection enhances STV mainly within a narrow band along the KEF (Figure 4c); elsewhere, it contributes negatively, indicating that STV amplification outside the frontal zone in FORCED is primarily driven by local processes (Figure 4e), while along the KEF, both contributions matter. In contrast to STV, Ekman advection exerts a much stronger influence on mixed-layer depth: Figure 4d indicates that the deepening around the KEF is dominated by Ekman advection and explains most of the deepening in FORCED (Figure 4b), whereas the deepening effect from local processes is relatively limited (Figure 4f). This implies that at the cold-event scale, wind-driven nonlocal processes more readily reshape mixed-layer structure near the front, while local processes dominate SST-variability responses outside the frontal zone.

Within the local response, comparing FE and FEW further isolates the contribution from wave-breaking mixing (Figure 5a,b). Wave-induced mixing suppresses STV across the KOE and subpolar regions (Figure 5a) and further shoals the shallow mixed layer around the KEF (Figure 5b). In the Craig and Banner parameterization, injection of surface TKE flux forms a strongly turbulent near-surface layer [34]; downward transport and dissipation of TKE lead to a quasi-equilibrium that can be viewed as a wave-enhanced transport layer. This process strengthens near-surface vertical diffusion of momentum and heat, while potentially reducing shear production and entrainment at the mixed-layer base, leading to mixed-layer shoaling; similar effects have been reported in both observations and modeling studies [41,42,43]. Enhanced near-surface diffusion also accelerates mixing and dissipation of temperature anomalies, consistent with reduced STV. Note that the Craig and Banner scheme provides a simplified representation of wave state and its quantitative impacts on mixed-layer structure remain uncertain [34]; parameterization differences and applicability have been discussed [43,44]. Future work could incorporate more detailed wave information, ideally using a wave model, to improve the representation of wave-induced mixing [45].

After removing Ekman advection and wave-induced mixing, the FEW experiment depicts the response driven by local buoyancy and momentum fluxes (Figure 5c,d). Local flux forcing significantly enhances event-period STV over most of the KOE and subpolar regions and deepens the mixed layer, with the most pronounced relative amplification in subpolar waters. Local flux forcing and wave-induced mixing act in opposite directions on STV and mixed-layer depth; they partially compensate within the local response, but overall local flux forcing remains dominant (Figure 4e). In Section 3.3, we further decompose the local flux forcing into contributions from heat flux, wind stress, and freshwater flux.

3.3. Cold-Sector Events: Contributions from Heat, Momentum, and Freshwater Fluxes

Next, we compute the differences FEW−FEWH, FEWH−FEWHW, and FEWHW−FEWHWF to obtain the effects of net heat flux, wind stress, and freshwater flux, respectively (Figure 6). Here, the wind-stress effect reflects only mechanical mixing because Ekman advection has already been removed in Section 3.3. Results show that net heat-flux anomalies dominate the local enhancement of STV and the main spatial structure of mixed-layer deepening over most of the KOE and subpolar regions; mechanical mixing and freshwater-flux contributions are generally weaker. The sum of the three contributions agrees well with the total response in FEW, indicating that at the event scale, the combined effects of the three fluxes are approximately additive and strongly nonlinear coupling is relatively limited. Net heat flux simultaneously enhances STV and deepens the mixed layer, implying that the two are not independent parallel responses. In the climatological shallow mixed-layer band surrounding the KEF, strong heat loss likely triggers convective adjustment and entrainment, deepening the mixed layer and amplifying surface temperature perturbations; over the KEF and in subpolar deep mixed-layer regions, further deepening is constrained by the background depth, yielding weaker signals. Combined with Section 3.3, the full adjustment of mixed-layer structure in the frontal zone still requires nonlocal advection, and net heat flux and Ekman advection jointly shape event-period mixed-layer responses in space.

The direct effect of wind stress through mechanical mixing on STV is generally weak: it barely changes STV over the KOE, and its effect on mixed-layer depth is scattered (Figure 6c,d). By contrast, freshwater flux has a very limited influence on STV over most regions but shows a significant positive contribution southeast of the SAF (155–165° E, 42–48° N; Figure 6e). In this region, contributions from net heat flux and wind stress to STV are both weak, forming a pronounced low center; freshwater flux compensates this minimum and is associated with mixed-layer deepening (Figure 6f). Freshwater flux modulates mixing by altering surface salinity and density stratification. During cold-sector events, enhanced evaporation increases surface salinity and density, potentially reducing static stability and strengthening convective mixing; entrainment can bring colder subsurface water into the mixed layer, enhancing surface cooling and amplifying STV. This sensitivity region lies southeast of the SAF and is the formation area of Transition Region Mode Water, where mixed-layer processes are more sensitive to salinity anomalies [46,47,48]. Elsewhere, freshwater-flux effects tend to manifest as slight mixed-layer shoaling with weak temperature responses, possibly because evaporation anomalies are confined to a very shallow surface layer and form a thin saline layer; under the winter background controlled by a strong thermocline, salinity-induced density perturbations are often insufficient to trigger deep convection or substantially enhance entrainment, leading to weak modulation of SST variability.

Overall, the local flux response to cold-sector extreme events is dominated by anomalous net heat loss, with wind-stress-driven mechanical mixing playing a secondary role. The freshwater-flux influence is strongly regional, producing appreciable modulation of STV only southeast of the SAF.

3.4. Warm-Sector Events and Total Forced Response

During warm-sector extreme events, the composite forced response in the FORCED experiment is shown in Figure 7; quantitative contributions from individual forcing factors to STV and mixed-layer depth are given in Supplementary Figure S6, and Figure 8 summarizes and contrasts the contribution structures for cold- and warm-sector events. Compared with Figure 4, warm-sector events feature overall reductions in STV and mixed-layer shoaling, with spatial patterns broadly similar to those of cold-sector events but with opposite signs, reflecting the opposite directions of flux anomalies. Among the factors, heat flux remains the primary contributor (Supplementary Figure S6e,f) and shares the sign of the FORCED response, because warm-sector events correspond to a minimum in upward NHF, slowing surface cooling and suppressing mixed-layer development.

Wave-induced mixing depends only on wind speed; thus, under the strong winds of warm-sector events, it still produces a thin, strongly mixed surface layer while reducing the amplitude of surface temperature perturbations (Supplementary Figure S6c,d). Freshwater flux during warm-sector events manifests as net precipitation, freshening the surface and inhibiting convective mixing, thereby shoaling the mixed layer and enhancing STV; the region southeast of the SAF again emerges as a response center (Supplementary Figure S6i,j). This indicates a clear asymmetry in the upper-ocean response to freshwater forcing over the KOE (except southeast of the SAF): both evaporation and precipitation tend to shoal the mixed layer, but only precipitation effectively modulates STV. Ekman advection generally enhances STV in warm-sector events (Supplementary Figure S6a), corresponding to northeastward warm advection that induces surface warming on 1–3-day time scales, opposite to the cold-sector effect (except along the KEF). However, its impact on mixed-layer depth still tends to be deepening (Supplementary Figure S6b), suggesting that the shallow mixed-layer band around the KEF is surrounded by deeper mixed-layer waters and that Ekman transport, regardless of direction, can import deep-mixed-layer waters into the local area. Wind-driven mechanical mixing is still incapable of penetrating the main thermocline and substantially altering mixed-layer depth, but it can moderately enhance STV over the KOE (Supplementary Figure S6g,h). Given comparable wind-stress magnitudes but opposite directions between cold and warm events (Figure 3c,d), this difference may arise from different near-inertial energy input associated with wind-stress turning in different cyclone quadrants, which would affect shear production of turbulence and its dissipation of STV [26,49]. However, given our experimental configuration, these processes cannot be accurately simulated and quantified here and warrant further investigation. As noted in Section 2.3, vertical temperature advection associated with Ekman pumping is negligible in our framework.

Combining cold- and warm-sector events, the overall impacts on STV and mixed-layer depth are shown in Figure 7c,d. Forcing associated with different cyclone sectors does not cancel out; instead, it produces a distinctly asymmetric net effect: STV increases upstream along the KEF, decreases between the KEF and SAF, and the shallow mixed-layer band surrounding the KEF deepens overall. These results demonstrate that extratropical-cyclone-related extremes can substantially modulate upper-ocean temperature variability and mixed-layer structure, and the controlling factors and pathways have been systematically identified.

3.5. Background Winter Ocean State and Responses During Cold- and Warm-Sector Events

Given the above results on pure-forced oceanic responses, this section discusses the full STV, including oceanic intrinsic processes, and its relationship with the forced SST variability. Based on ECCO2 data, the wintertime-mean STV (regardless of in or out of the event windows) shown in Figure 9a reveals that high STV values are concentrated near the KEF, with a peak variance of about 0.6 K² (standard deviation ~0.77 K); a secondary band also appears along the SAF. We further estimate the mesoscale-eddy contribution to STV using the method of Zhou et al. [50] (Figure 9b). In the KOE region, the contribution commonly reaches 40–100%, whereas outside the frontal zone it is mostly below 10%. This contrast is consistent with Wu et al. [17], indicating that in the frontal zone, intrinsic ocean variability provides the primary background for winter STV, and thus extreme weather events are more of a modulation on top of this existing variability.

The ratio of event-window and winter-mean STV in ECCO2 is shown in Figure 9c,d. Contrasting Figure 7c,d, here the event-window STV includes both forced and intrinsic variability. In most regions (including subtropical, subpolar, and downstream KOE waters), cold-sector events yield ratios significantly greater than 1, implying enhanced STV. However, near the KEF and along the SAF—where event frequency and intensity are high and background STV is strongest—the ratio is close to 1 or below 1, indicating weak suppression or little change. Warm-sector events show a similar pattern but with generally weaker amplification/suppression (Figure 9d); combining both event types gives consistent anomalies (Supplementary Figure S5).

Comparing the purely forced STV response during events (Figure 7c) with the total STV represented by ECCO2 (Figure 9c,d and Supplementary Figure S5) reveals marked differences, further highlighting the key role of intrinsic ocean dynamics in synoptic air–sea interaction in this region. These contrasts suggest that in regions such as the KEF, where background STV is strong, cyclone-related extreme forcing does not simply add to intrinsic ocean variability. Instead, by modifying mixed-layer structure and upper-ocean flows, it may adjust eddy–front-related advection and heat redistribution, statistically reducing or offsetting part of the intrinsic variability, leading to inconsistency between forced and total variability. In subtropical and subpolar waters where background STV is weaker, the qualitative agreement between forced STV amplification and total STV enhancement suggests that extreme events behave more like externally forced responses and thus more readily manifest as relative STV amplification.

Previous studies have shown that eddy and frontal processes can strongly reshape the mixed-layer heat budget and redistribute temperature perturbations induced by surface heat flux in both horizontal and vertical directions [51,52]. Wind–current feedback can also transfer kinetic energy from ocean mesoscale circulation to the atmosphere, causing dissipative effects that weaken eddies [53,54]. These insights provide a physical backdrop for the spatial contrast in Figure 9c,d, with suppression within the frontal zone and amplification outside. Although both model and reanalysis inevitably include parameterization and numerical errors and a strict quantitative comparison is difficult, the above physical inference is mechanistically self-consistent and plausible.

4. Conclusions

Focusing on the Kuroshio Extension and adjacent frontal zones, we identify two types of extreme events associated with extratropical cyclone cold and warm sectors using ECCO2 reanalysis and a GOTM single-column ensemble framework. Starting from heat, momentum, and freshwater fluxes and the associated local mixing processes, and further distinguishing wind-driven Ekman advection and wave-breaking mixing, we characterize the main pathways through which synoptic extreme forcing affects the upper ocean and its spatial differences.

Our results indicate that the direct upper-ocean forcing during both cold- and warm-sector events is dominated by heat-flux anomalies, with effects strongly structured by the frontal background. In cold-sector events, intense net heat loss strengthens buoyancy loss and becomes the core source of STV amplification and mixed-layer evolution; in warm-sector events, the pronounced weakening of upward NHF suppresses surface cooling, leading to overall STV reduction and mixed-layer shoaling. By comparison, the direct contribution of wind stress through mechanical mixing is generally weak, acting more as a secondary modulation of temperature perturbations in specific regions rather than as a dominant mechanism. Wave-breaking mixing damps temperature perturbations and accompanies mixed-layer shoaling in both event types, implying that within our framework it behaves more like a diffusive–dissipative process that stabilizes event-scale STV. The influence of freshwater flux is highly selective: southeast of the SAF, both enhanced evaporation in cold events and enhanced precipitation in warm events can substantially modulate STV and mixed-layer structure, suggesting higher sensitivity of the upper-ocean structure to freshwater perturbations in the frontal transition zone.

Contrasting our heat-flux-dominant conclusion with Kobashi et al. [27] for a subtropical North Pacific frontal region highlights regional differences in which components of the heat flux matter. They suggested that cyclone-induced surface cooling in the subtropical frontal zone is mainly driven by cloud-reduced shortwave radiation together with enhanced latent heat loss and ocean mixing, with impacts extending to ~100 m and persisting longer in the subsurface. In our winter KOE events, synoptic NHF anomalies are mainly contributed by non-solar terms rather than shortwave, implying that turbulent and longwave processes are stronger in the storm-track entrance region, where air–sea temperature contrasts and wind anomalies more readily amplify latent and sensible heat exchange. In the subtropical frontal zone, in contrast, cloud shading can more readily become the key trigger for cooling.

A more frontal-zone-specific finding is that wind-driven Ekman advection exerts a robust and non-negligible shaping effect on event responses near the KEF. Its impact on STV changes sign between cold and warm events with the direction of cross-front advection, reflecting how sector-dependent cyclone differences project onto SST perturbations through cross-front transport. In contrast, its impact on the climatological shallow mixed layer surrounding the KEF tends to be deepening in both event types, underscoring the importance of nonlocal wind-driven transport for mixed-layer structure under strong fronts. Thus, the impacts of extremes on the frontal-zone upper ocean are not determined by local fluxes alone, but jointly by heat-flux-dominated buoyancy adjustment and Ekman-advection-dominated nonlocal transport, with their relative importance varying with frontal position and background stratification.

The superposition of cold- and warm-sector events yields a structural net effect, indicating that sector-dependent extremes do not cancel statistically; instead, under the joint constraints of frontal gradients and wind-driven transport, they leave a recognizable spatial fingerprint. This fingerprint provides a clearer process-based explanation of how synoptic forcing at the storm-track entrance region leaves a cumulative imprint on ocean frontal zones. In regions like the KEF, where background variability is intense and intrinsic ocean advection and eddy processes are active, the total forced response to extreme events does not necessarily translate into synchronous amplification or suppression of total STV; instead, it is more likely to be redistributed or partly offset by intrinsic ocean dynamics. In subtropical and subpolar waters where background STV is weaker, extreme forcing behaves more like an external driver and can project onto total statistics. This contrast echoes conclusions that submonthly SST variability over the KOE is mainly controlled by intrinsic ocean processes [17] and is consistent with studies of scale-dependent heat-flux feedback in frontal regions [18]. Therefore, in strong western boundary current fronts, synoptic atmospheric extremes do not unidirectionally drive ocean variability; their net effects depend on their event-scale coupling with intrinsic ocean dynamics.

5. Discussions

This study focuses on thermodynamic and turbulent adjustments within the mixed layer under typical extratropical cyclones and does not address explosive cyclones or deeper-ocean responses. Modeling studies [6] showed that winter explosive cyclones over the North Pacific can induce strong surface divergence and upwelling southeast of the SAF, with signals extending downward to ~2000 m and generating day-scale temperature and vertical-velocity perturbations in the deep ocean via vertical advection and near-inertial wave processes. This suggests that the impact depth of strong cyclones may exceed the mixed-layer scale, and that deep-ocean responses may be dynamically connected to upper-ocean heat-content changes through vertical processes. However, as the deep response is mostly confined to the southeast of the SAF (which, in our study, is where freshwater flux plays its role), this finding may not be applicable to the entire North Pacific frontal zone.

In our framework, the impact of Ekman vertical advection is proven to be weak, and the daily mean forcing cannot fully represent near-inertial energy input and high-frequency wind-stress turning effects on turbulence generation, nor can it capture energy and heat transfer to the deep ocean. The effect of waves can also be improved in the future using wave models to better account for wave-induced turbulence [45] and heat flux anomalies [55] during extreme weather. Accordingly, our discussion of offsetting and redistribution in the frontal zone is more suitable for interpreting event responses of upper-ocean heat budgets and mixed-layer structure, while deep processes should be viewed as a possible extension for future work. Given quantitative uncertainties from numerical errors and parameterizations, we emphasize robust identification of process chains and relative contributions rather than strict quantitative attribution. The use of a 1-D model has advantages in explicitly separating the forced response from the intrinsic ocean dynamics but inevitably underrepresents the interaction between forced and intrinsic variability. Future studies combining higher-temporal-resolution forcing, regional 3-D ocean models with explicit mesoscale dynamics, and targeted air–sea-wave coupling experiments could further quantify the coupling between extreme events and intrinsic variability and assess their potential impacts on longer-time-scale frontal SST anomalies and storm-track changes.