Vertical structures and drivers of marine heatwaves and cold-spells in the Kuroshio Extension region

Marine heatwaves (MHWs) and marine cold-spells (MCSs) are prolonged oceanic extreme temperature events that can severely impact large-scale ecosystems, fisheries, and human activities with consequent socioeconomic impacts. Although some studies have contributed valuable insights into the vertical structure and related mechanisms of MHWs, equivalent research on MCSs remains unclear. Thus, comprehensive and systematic analysis of the vertical structures and related mechanisms of MHWs and MCSs remains area of an active research. In this study, we classified MHWs/MCSs into two types in the Kuroshio Extension region: extended MHWs/MCSs that can extend through more than 70% of the water column and shallow MHWs/MCSs that are restricted from the surface layer to less than 70% of the water column. Analysis revealed that shallow events are characterized by stronger intensity and shorter duration compared with extended events. All shallow events are driven by surface heat flux anomalies, with shortwave radiation (latent heat flux) mostly inducing those in MHWs (MCSs). However, extended MHWs/MCSs are primarily driven by ocean anticyclonic/cyclonic eddies. These findings provide deeper understanding of the statistical characteristics, vertical structures, and physical drivers of MHWs and MCSs.


Introduction
Marine heatwaves (MHWs), characterized by prolonged extreme high temperature in the ocean, are escalating in severity and have resulted in notable ecosystem disasters and substantial socio-economic losses (Meehl and Tebaldi 2004, Frölicher et al 2018, Oliver et al 2018, Smith et al 2021).An MHW event can persist for several days to months, spanning spatial scales from tens to thousands of kilometers (Waliser 1996, Qin et al 2007, Pearce and Feng 2013, Chen et al 2014, Di Lorenzo and Mantua 2016, Jackson et al 2018, Holbrook et al 2019).While MHWs have garnered significant focus, their inverse counterpart, marine cold-spells (MCSs), involving extreme ocean cooling events, have similarly devastating impact on marine ecosystems (Wang et al 2020, Schlegel et al 2021, Wang et al 2022, Yao and Wang 2022) but have been relatively less well investigated.Understanding MHWs and MCSs is essential for comprehending broader climate dynamics and predicting climatic variations.
MHWs are driven by atmospheric and/or oceanic processes, and they are also closely linked to largescale climate modes (Holbrook et al 2019, Yao et al 2020, Yao and Wang 2021, Zhang et al 2021, Liu et al 2022, Han et al 2023).The development of MCSs is steered by multiple factors related to atmospheric dynamics, such as atypical winds and air-sea heat exchange (Economidis and Vogiatzis 1992, Gómez and Souissi 2008, Pirhalla et al 2015).Additionally, variations in ocean currents and notably vigorous upwelling play pivotal roles in their formation (Yuan 2006, Schlegel et al 2017, Wijffels et al 2018).
Studies of MHWs and MCSs have typically been conducted using only surface data, facilitated by the availability of a 30 year (or longer) satellite-derived record of sea surface temperature (Reynolds et al 2007, Banzon et al 2016).However, following research advances, studies of MHWs are expanding in scope from surface examination to consideration of the subsurface (Schaeffer and Roughan 2017, Elzahaby and Schaeffer 2019, Scannell et al 2020, Hu et al 2021, Sun et al 2023b) and even the bottom of the ocean (Amaya et al 2023).Recently, several studies reported that MHWs exhibit intricate vertical structures shaped by various influences (Elzahaby et al 2021, Ryan et al 2021, Schaeffer et al 2023, Zhang et al 2023).For example, in the coastal waters of eastern Australia, the vertical coherence of MHWs is regulated by seasonal stratification, large-scale circulation, and local downwelling processes (Schaeffer et al 2023).Regional MHWs in the East Australian Current System, which are driven by anomalous advection, typically extend three times deeper than those influenced by surface air-sea heat fluxes (Elzahaby et al 2021).These findings underscore the intricate nature of MHW formation and stress the importance of considering subsurface MHWs.Although there is preliminary understanding of the vertical structures of MHWs, those of MCSs are less well comprehended, with notable lack of systematic comparison between the characteristics and mechanisms of the two types of events.
MHW hotspots are predominantly located in regions characterized by substantial temperature fluctuations, such as western boundary currents and their extensions (Oliver et al 2018, Holbrook et al 2019).Within the Northwest Pacific, the Kuroshio Extension (KE) region exhibits the greatest temperature variability at various depths (figure 1(a)).This pronounced temperature variation, coupled with intensified ocean warming (Wu et al 2012, Yang et al 2016), can easily lead to intense MHWs.Previous investigations conducted in the KE region predominantly focused on surface MHWs during summer, attributing the driving factors to anomalies in air-sea heat flux and the ocean's memory of winter warming (Du et al 2022).Nevertheless, knowledge remains limited regarding the typical characteristics and related mechanisms of the vertical structures of both MHWs and MCSs in the KE region.
In this study, we focused on a subset of extreme temperature events in the KE region that occurred during 1993-2020.Recognizing the complexity of the vertical structures in those events, we simplified our categorization into shallow and extended MHWs/MCSs based on their depth extent, thereby allowing comparison of their vertical structures and mechanisms.This work provides a relatively comprehensive understanding of oceanic extreme temperature events in the KE region.

Data
The Global Ocean Reanalysis and Simulations (GLORYS) 12v1 product (Jean-Michel et al 2021) from the Copernicus Marine Environment Monitoring Service (CESMS), which includes daily ocean data such as potential temperature and velocities with high horizontal resolution of (1/12) • × (1/12) • and 50 vertical levels, has been validated against in-situ observations and alternative models.It has been proven to accurately capture ocean processes at both regional (Artana et al 2019, Poli et al 2020, Verezemskaya et al 2021) and global scales (Jean-Michel et al 2021).We further assessed its robustness by comparing the monthly sea temperature anomalies at different depths with Global Ocean Data Assimilation System (GODAS) datasets (Behringer 2007) from the National Centers for Environmental Prediction (NCEP).The results showed reasonable consistency in terms of temporal variation and trend (figure S1).
Daily atmospheric reanalysis data on a 0.25 • × 0.25 • grid, including surface net shortwave radiation (SWR) and longwave radiation (LWR), surface latent heat flux (LHF) and sensible heat flux (SHF), 10 m u-component and v-component of wind fields, total cloud cover (TCC), and 1000 hPa specific humidity, were extracted from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) (Hersbach et al 2020).The net surface heat flux (Qnet) taken as the sum of the SWR, LWR, SHF, and LHF, and a positive value indicates surface heat flux into the ocean.
We employed the Mesoscale Eddy Trajectory Atlas version 3.2 Delayed-Time allsat version (META3.2DT allsat) dataset obtained from AVSIO+, to determine the location, edge contour, and polarity of each identified eddy.
The analysis period adopted for all the above datasets is 1993-2020.All daily anomalies were calculated by removing day-of-the-year averages over the entire duration of the datasets.

Depth categorization of MHWs and MCSs
Using the method for identification of MHWs and MCSs specified in section 2.2, we obtained a matrix of MHWs and MCSs over time and depth.Following Schaeffer et al (2023), we defined shallow MHWs/MCSs based on events identified at the 0.49 m depth.With respect to the duration of each MHW/MCS event identified at 0.49 m depth, we investigated the cooccurrence of MHWs/MCSs across different depth levels.If the MHW/MCS extends from the surface to no more than 70% of the entire water column (the MHW/MCS conditions could be met only at depths of <130.29 m) for more than half of the event period, we classified this as a shallow MHW/MCS.However, we defined extended MHWs/MCSs based on events identified at the 92.33 m depth, and we explored the cooccurrence of MHWs/MCSs across different depth levels during each MHW/MCS event determined at the 92.33 m depth.If 70% or more of the entire water column exhibited MHW/MCS conditions for more than half of the event period, we classified this as an extended MHW/MCS.We used the conditions (with or without MHW/MCS) at each depth level to represent the water column conditions above it, which extended up to the next depth level.For example, the conditions at the 0.49 m depth reflected the water column conditions in the depth range of 0-0.49m, and those at the 1.54 m depth level represented the water column conditions in the depth range of 0.49-1.54m.In this analysis, we estimated MHWs/MCSs from the surface to the depth of 186.13 m; therefore '70% or more of the entire water column' corresponded to depths of ⩾130.29 m.The duration and mean intensity of shallow events are then extracted from the events at 0.49 m depth, while those of extended events are extracted from the events at 92.33 m depth.

Analysis of the structures and metrics of MHWs and MCSs in the KE region
During 1993-2020, 26 out of the 47 MHWs identified at the depth of 0.49 m in the KE region were classified as shallow MHWs (table S1).At the depth of 92.33 m, 12 of the 23 identified MHWs were classified as extended MHWs (table S2).Similarly, at 0.49 m depth, 24 out of the 42 identified MCSs were classified as shallow MCSs (table S3).At 92.33 m depth, 15 of the 25 identified MCSs were classified as extended MCSs (table S4).Shallow MHWs/MCSs were limited to surface mixed layers and peaked at the surface.Extended MHWs/MCSs, which peaked below the surface, could span the entire (or most of) water column.In the four typical examples presented in figure S3, the maximum intensities were 1.30 • C at 0.49 m, 0.82 • C at 9.57 m, −1.16 • C at 0.49 m, and −0.90 • C at 18.50 m.The KE region is in the mid-latitudes, where the mixed layer depth exhibits significant seasonal variability (figure S1).Shallow MHWs/MCSs occur most frequently in summer (12 times/8 times) due to strong stratification, and extended MCSs are most common in winter (6 times) due to weak stratification (Schaeffer et al 2023).However, extended MHWs are more evenly distributed across seasons.This suggests that stratification is not the only factor affecting the seasonal distribution characteristics of shallow and extended events.
Figure 1 shows the mean vertical structures and statistical characteristics of MHWs and MCSs identified in the KE region.The two types of MHWs and MCSs exhibited strong symmetry in terms of vertical temperature anomalies.In this regard, the maximum temperature anomaly of shallow MHWs/MCSs exceeding +1.0 • C/−1.0 • C was confined to the 20 m surface layer and decreased most rapidly within 40 m.Additionally, the warm/cold temperature anomalies decreased to +0.5 • C/−0.5 • C below depths of 40-50 m.For extended MHWs/MCSs, the temperature anomalies were more homogeneous across depth, i.e. the difference in temperature anomalies from the surface to the subsurface was less than 0.2 • C. In addition to the strong consistency between the two types of MHWs and MCSs, there were some minor differences such as the decreasing rate of shallow MCSs was slightly faster than that of shallow MHWs.The temperature anomaly difference in extended MHWs across depth was greater than that in extended MCSs.The metrics of both types of MHWs and MCSs such as duration (figure 1(c)) and mean intensity (figure 1(d)) had strong similarities.Extended MHWs and MCSs persisted for longer (median duration of 27 and 25 d, respectively) but were less intense (median mean intensity of 0.74 • C and −0.65 • C, respectively) than shallow MHWs and MCSs (median duration of 11 and 10 d, respectively; median mean intensity of 1.07 • C and −0.95 • C, respectively).Among the two types of MHWs and MCSs, the duration of extended MHWs (ranging between 8 and 371 d) and the mean intensity of shallow MCSs (ranging between −0.50 • C and −1.73 • C) had the greatest variability.
The horizontal temperature and current anomalies of the two types of MHWs and MCSs at depths of 0.49 and 92.33 m are given in figure 2. It is suggested that shallow MHWs were characterized by uniform warming at 0.49 m in the KE region (figure 2(a)).The KE region had small and weak warming centers at 92.33 m during shallow MHW periods, such that the amplitude of overall warming could not meet the threshold of a MHW (figure 2(b)).The horizontal distributions of shallow MCSs at depths of 0.49 and 92.33 m (figures 2(e) and (f)) exhibited distinct patterns compared with those of shallow MHWs.Specifically, the shallow MCSs had a colder center in the western KE region at 0.49 m depth, which was maintained at the depth of 92.33 m.However, the shallow MCSs had a colder center in the western KE region at 0.49 m depth, which was maintained at the depth of 92.33 m.For the extended MHWs/MCSs, the warmest/coldest anomalies were concentrated around the KE axis at 0.49 m depth (figure 2(c)/(g)) and extended to the depth of 92.33 m (figure 2(d)/(h)), accompanied by strong anomalous anticyclonic/cyclonic eddies.However, the maximum warming centers of MHWs were oriented in an east−west direction, while the maximum cooling centers of MCSs were oriented in a northwest−southeast direction.

Drivers of MHWs and MCSs in the KE region
To further identify the contributions of dynamic and thermodynamic processes to the two types of MHWs/MCSs in the KE region, we conducted composite analysis of the influencing factors associated with the different types of MHWs/MCSs (figures 3 and 4).During shallow MHWs, an anomalous lower-level anticyclone centered at the southeastern KE markedly reduced the TCC over the southeastern KE by suppressing local convection (figure 3(a)), leading to increase in net downward SWR in the southeastern KE (figure 3(e)).This anticyclonic anomaly also reduced the wind speed and associated surface evaporation over the centraleastern KE (figure 3   accompanied by anticyclonic eddies reaching 1.97 • C and that accompanied by cyclonic eddies reaching −0.52 • C. At both depths, the warm anomalies were much greater than the cold anomalies, which enabled them to meet the conditions for MHWs. Figure 4(a) also shows that in shallow MHWs, the effects of cyclonic and anticyclonic eddies at depth of 92.33 m were notable.However, the warm and cold anomalies under both effects were broadly equal, resulting in a total temperature anomaly that was too low to meet the threshold for MHWs.Therefore, mesoscale eddies are crucial in shaping the vertical structure of MHWs.
The factors that influenced shallow MCSs were different from those that influenced shallow MHWs, although the Qnet anomalies related to an anomalous lower-level cyclone was critical.Specifically, during shallow MCSs, an anomalous lower-level cyclone was observed over the northeastern KE, which led to marked increase in the TCC (figure 3(c)) but small reduction in the net downward SWR (figure 3(g)).
Additionally, the heightened wind speed (figure 3(c)) resulted in enhanced release of latent heat of evaporation from the ocean to the atmosphere (figure 3(k)), resulting in the Qnet anomalies being mainly contributed by LHF.While Qnet anomalies (figure 3(o)) explained most of the KE surface cooling, they did not account for strong western KE cold anomalies.
strong cold anomalies might have been related to a local anomalous cyclonic eddy (figures 2(e) and (f)).These conditions collectively favored surface cooling.
In extended MCSs, the anomalous lower-level cyclone dominated over the KE region, but its center was shifted to the southeast of the KE.The southward movement of the cyclone caused the Qnet anomalies within the entire KE region to be negligible (figure 3(p)).However, the cold anomalies accompanied by cyclonic eddies were extremely strong, both at the surface and subsurface (figure 4(b)).In contrast to the extended MHWs, cyclonic eddies assumed a more prominent role.Although the effect of the airsea heat flux was not notable, the cooling temperature anomalies induced by eddies were sufficient to meet the conditions of MCSs.The role of mesoscale eddies in shaping the vertical structure of MCSs was similar to that related to MHWs.The difference is in the dominant force: in MHWs, anticyclonic eddies are superior, whereas cyclonic eddies are prevalent in MCSs.

Conclusion and discussion
We classified extreme temperature events identified in the KE region from 1993 to 2020 into shallow and extended MHWs/MCSs based on their position in the water column.Despite their differences, the MHWs and MCSs shared many characteristics.Strong symmetry was found between the vertical structure of the MHW and MCS composites within the KE region (figure 1(b)), suggesting similar driving processes for both positive and negative extremes.Shallow MHWs and MCSs, driven primarily by atmospheric processes, were characterized by shallow depth, short duration, and high intensity.Shallow MHWs in the KE region, similar to many iconic extratropical MHWs (Lee et al 2010, Bond et al 2015, Holbrook et al 2019, Rodrigues et al 2019), were found linked to persistent anticyclonic systems over the ocean and their resulting air-sea interactions, which collectively led to reduced ocean heat loss and increased solar radiative heating that result in sea surface warming.Conversely, shallow MCSs were found associated with persistent cyclonic systems over the ocean and their resulting air-sea interactions, which collectively led to reduced solar radiative heating and increased ocean heat loss that result in sea surface cooling.Shallow events are driven by Qnet anomalies.However, in shallow MHWs, these anomalies are primarily induced by SWR, whereas they are mainly caused by LHF in shallow MCSs.Extended MHWs and MCSs, driven primarily by oceanic processes, were characterized by deep depth, long duration, and low intensity.Elzahaby et al (2021) suggested that surface flux-driven MHWs are shallower and shorter in duration than advection-driven MHWs.This is because the air-sea heat flux is mainly absorbed by the upper ocean, and there is rapid feedback with the atmosphere.In shallow events, the warm anomalies associated with anticyclonic eddies and the cold anomalies associated with cyclonic eddies have similar amplitude.This similarity in amplitude results in reduction of the overall temperature anomalies because the warm and cold anomalies broadly counterbalance each other.Consequently, this counterbalancing effect prevents the formation of MHWs or MCSs in the subsurface.In extended events, although Qnet is weak, the role of anticyclonic or cyclonic eddies is crucial.Mesoscale eddies are important in driving ocean subsurface temperature extremes (He et al 2023).The effect of mesoscale eddies is spatially heterogeneous, becoming more dominant in the western boundary currents and their extensions (Bian et al 2023).
MHWs and MCSs represent major challenges for marine ecosystems and the sustainability of marine resources due to their negative impacts on many marine organisms and ecosystems (Lirman et al 2011, Frölicher and Laufkötter 2018, Tuckett and Wernberg 2018, Smale et al 2019, Smith et al 2021).Shallow MHWs, extended MHWs, shallow MCSs, and extended MCSs have different characteristics and drivers and they can affect ecosystems at different depths.Comprehensive understanding of these extreme oceanic warm/cold water events could provide critical information and guidance for marine conservation, fisheries and aquaculture management.Our method effectively delineates and characterizes the vertical structures of MHWs and MCSs.It is designed with specific focus on the vertical structures, which allows in-depth insight into such phenomena.However, the complete spatial structure and the evolution of such events, which unfold in a four-dimensional spatiotemporal framework, present additional layers of complexity.This study focused on the vertical structures, but future studies could potentially incorporate more comprehensive assessment of these events by considering their full spatiotemporal dynamics (Sun et al 2023a(Sun et al , 2023b)), which would further enhance our understanding of MHWs and MCSs.Additionally, the dynamics of subsurface MHWs and MCSs should be quantitatively investigated in future research.

Figure 1 .
Figure 1.(a) Three-dimensional distribution of the standard deviation (STD) of de-seasoned temperature (units: • C) in the western North Pacific.The black box (32 • -38 • N, 140 • -165 • E) represents the KE region.(b) The mean vertical temperature anomalies (units: • C) for shallow MHWs (magenta curve), extended MHWs (red curve), shallow MCSs (cyan curve) and extended MCSs (blue curve); the corresponding shading indicates the ±1 standard deviation uncertainty range of the vertical temperature anomalies for each type of event.Boxplots of (c) the duration (units: days) and (d) the mean intensity (units: • C) for different types of events after removing extremes (marked by the dots in (c) and (d)) with values that are more than 1.5 times the interquartile range from the top or bottom of the box.

Figure 4 .
Figure 4. (a) Average temperature anomalies overlain on anticyclonic eddies (TaAE; orange bar; • C) and cyclonic eddies (TaCE; green bar; • C) during the two types of MHWs.Red shading represents the results at 0.49 m depth, and the blue shading represents the results at 92.33 m depth.(b) As in (a) but for the two types of MCSs.