Impact of the reemergence of North Pacific subtropical mode water on the multi-year modulation of marine heatwaves in the North Pacific Ocean during winter and early spring

Marine heatwaves (MHWs) are events during which the sea surface temperature (SST) is extremely high compared to typical climatological values. MHWs have recently attracted significant interest owing to their broad impacts on numerous processes, including global warming, marine ecosystems, and aquaculture industries [1–4]. MHWs have varied drivers such as the horizontal and vertical advection of heat, reduced vertical mixing, anomalous surface heat uptake, and shifts in the SST front [4]. In the North Pacific, MHWs were found to be related to low-frequency climate variabilities such as the Pacific decadal oscillation (PDO) [5] and North Pacific gyre oscillation [6], highlighting the effects of large-scale oceanic phenomena on MHWs.

One of the important factors regulating SST in the North Pacific, especially during winter and early spring, could be North Pacific subtropical mode water (STMW). STMW is typically characterized by a vertically homogeneous layer with a temperature range of 16 ^∘C–19 ^∘C in the North Pacific subtropical gyre [7–10]. STMW is normally formed by vertical mixing near the Kuroshio Extension from December to March. It then becomes isolated from the atmosphere when surface heating enhances the stratification [11–14]. Observations and numerical experiments have revealed the mechanisms and processes controlling the formation of STMW and its variability. Local air-sea buoyancy flux can directly control the mixed layer depth (MLD) and the formation of STMW [11, 15–18]. Large-scale climate variability and the instability of ocean currents also influence the annual variation of STMW formation, with a time lag of several years [9, 14, 19–24]. Finally, stratification intensity during the previous summer can modulate the wintertime MLD and STMW formation [19].

The area east of the STMW formation zone is occupied by the Transition Zone Chlorophyll Front (TZCF), across which the chlorophyll concentration changes rapidly [25, 26]. Its latitudinal position migrates with the seasons. It is found near 45^∘ N in summer but shifts further south in winter, near 30^∘ N, due to the southward Ekman transport caused by strong westerlies [26]. The TZCF shows elevated net community production [27, 28], and its winter position affects the survival rate of Hawaiian monk seal pups [29]. Consequently, MHWs occurring near 30^∘ N in the North Pacific need to be monitored closely as they can potentially damage the ecosystem near the TZCF.

STMW has a broad influence on both the physical and biogeochemical properties of the North Pacific in the timescale longer than a year. If capped by the seasonal thermocline, STMW could be entrained into the mixed layer in the following winter and affect the marine ecosystem or even the basin-scale climate [24, 30, 31]. STMW can be transported to the east where the TZCF locates while its properties are altered, and can surface in the following winter through a process referred to as reemergence [30, 32–37]. The volume of newly formed STMW can indicate the ability of the ocean to take up carbon and nutrients from the atmosphere, not only in the formation region but also near the TZCF and over a broad region of the North Pacific Ocean. This is because the transport of STMW by the subtropical gyre can transport STMW downstream to the interior of the ocean, followed by repeating reemergence [24, 38].

MHWs and STMW can regulate each other. Warming of the ocean surface during MHWs can increase stability and suppress the formation of STMW. A smaller volume of STMW is likely to accompany positive SST anomalies in the following winter because of the lower amount of heat during reemergence, which in turn increases the probability of future MHW activities. Therefore, the annual variation in the volume of STMW may be closely related to MHW activities in the North Pacific during winter and early spring. Although a close relationship has been suggested between MHWs and subsurface conditions in the Northeast Pacific [39], this has not yet been explored in the North Pacific, particularly in the context of connectivity between these two phenomena. In this study, we investigated the impact of STMW reemergence on MHWs during winter and early spring, at both the STMW formation site and in remote areas near TZCF in the North Pacific, using satellite observations and ocean state estimation.

2.1. North Pacific STMW

STMW is identified as a water mass that satisfies the two following conditions: nearly uniform temperature (16 ^∘C–19 ^∘C) and a small vertical temperature gradient ($\lt$1.5 ^∘C per 100 m) [7–10, 20]. As previous studies [10, 36] have shown, STMW forms near 30^∘ N over a 45^∘ wide area, spanning from 130^∘ E to 175^∘ E, while reemergence occurs to the east of 180^∘ over the same latitudes as the formation region. Hence we defined the formation region of STMW and the remote region from 130^∘ E to 175^∘ E and 175^∘ E to 160^∘ W, respectively, extending meridionally from 24^∘ N to 35^∘ N (blue solid and dashed boxes in figures 1(a) and (b)). This zonal band encompasses the winter TZCF. We applied these criteria to EN.4.2.1 data, which is a monthly objective analysis dataset [40], and computed the volume of STMW in the formation region from 1982 to 2017. Additionally, we used Estimating the Circulation and Climate of the Ocean, version 4, release 4 (ECCOv4r4) [41, 42], which is a three-dimensional global ocean state estimate, with a 1^∘ horizontal resolution, covering 1992 to 2017, to trace the STMW formed in March through a passive tracer experiment.

Zoom In Zoom Out Reset image size

The volume of STMW exhibits distinct seasonal variations. It increases from December to March, indicating that STMW is mainly formed during that time period with active vertical convection (see figure 4 in [13]). In March, the core of the formation site is situated near 30^∘ N, from 145^∘ E to 165^∘ E, and south of the Kuroshio Extension, where the thickness of the STMW is greater than 250 m (figures 1(a) and (c)). A zonal section of water temperature in the formation region shows that the formation of STMW occurs at a depth of approximately 100 m to the east of 175^∘ E (figure 1(c)). After its formation in March, the STMW is isolated from the surface by seasonal stratification and from April to November and moves to the remote region via the subtropical gyre in the subsurface layer [36]. The volume of STMW decreases over time, because it dissipates while being advected near a depth of 200–300 m (figure 1(d)). The thickness of STMW gradually decreases toward the downstream region of the Kuroshio Current due to the downstream cooling of the Kuroshio Extension [12, 16]. Nevertheless, STMW can affect the surface conditions of the ocean when reemerging in winter and early spring at the remote region near TZCF [30].

2.2. MHWs

STMW captures the thermal condition of the ocean when it forms in winter and early spring and potentially alters MHW activity when it reemerges in the following winter. This relationship was explored using the time series of normalized STMW volume and cumulative MHW intensity in the formation and remote regions in March (figure 2(a)). As anticipated, the STMW volume was found to be inversely related to the activity of MHWs. In general, the cumulative intensity of MHWs in the formation region is concurrently affected by variations in the STMW volume, while in the remote region, intensity shows a delayed response with a time lag of one or two years.

Zoom In Zoom Out Reset image size

Details of the relationship between MHWs and STMW were determined through correlation analysis. We calculated the correlation and time-lagged correlation between the monthly cumulative intensity of MHWs and the volume of STMW, as well as the MLD for the period 1982–2017. The MLD was estimated as the depth at which the potential density increases by 0.125 kg m⁻³, relative to the density at 10 m, using EN4.2.1 data, which was interpolated over 1 m depth intervals [46, 47]. Correlation with the PDO index was also analyzed because previous studies have suggested that the occurrence of MHWs in the North Pacific is related to the PDO, and that oceanic Rossby waves driven by the PDO can affect the formation of STMW with a time lag of several years. [4, 14, 48, 49].

The monthly cumulative intensity of MHWs is negatively correlated with STMW volume, and this negative correlation becomes statistically significant from October to May (red line in figure 2(b)). A particularly high negative coefficient from February to April indicates that a smaller volume of STMW can lead to higher MHW activity during winter and early spring. In summer, MHW activity and STMW volume are less correlated as expected, based on the separation of STMW from the surface ocean. MHWs are also negatively correlated with MLD. This correlation is statistically significant throughout the year (blue line in figure 2(b)), indicating weaker MHW activity with a thicker mixed layer. This is consistent with our understanding that deeper vertical mixing reduces the SST and decreases the probability and intensity of MHWs in all seasons. Spatial distributions of STMW thickness and the cumulative intensity of MHWs in March distinctly present this inverse relationship (figures 1(a) and (b)). The PDO index does not exhibit a significant correlation with MHW activity in the formation region except in summer (black line in figure 2(b)).

The long 'memory' of the ocean may allow STMW created in the formation region in winter and early spring to impact MHWs arising in the remote region in the following winter owing to reemergence. This hypothesis was investigated by performing time-lagged correlation analysis between MHW activities in the formation, remote regions and the STMW volume, MLD in the formation region, and PDO index in March (figures 2(c) and (d)). MHWs activity in the formation region had the most significant relation to all parameters with zero time lag and the correlation coefficients in the previous isolation stage (June (1) to November (1)) were not statistically significant (figure 2(c)). It suggests that newly formed STMW instantly modulates the probability of MHWs in winter and early spring and the remnant STMW in the previous isolation stage plays a secondary role in the MHWs' occurrence in the formation region. The activity of MHWs in the remote region did not show a significant correlation with the MLD in the formation region (the blue line in figure 2(d)). It was only negatively correlated with the PDO index with zero or one month of lag, indicating that there is a high probability of finding stronger MHWs in the remote region during the cold phase of PDO. Interestingly, the correlation between the MHW activity in the remote region and the volume of STMW in the formation region is greater with a one or two years of lag than with zero lag. This indicates that MHW activity in the remote region is more closely related to the volume of STMW in the previous winters than in the current winter. In fact, the correlation coefficient was the lowest (most significant negative correlation) with a time lag of one year (r = −0.65). The coefficients were relatively low over the summer and passed the significance threshold in the fall and winter, suggesting that there is a particular process responsible for the stronger correlation in this season.

The spatial patterns of the time-lagged correlation coefficient can shed light on the mechanisms behind the strong connection between the volume of STMW and MHW activity in the following years (figure 3). With zero lag, statistically significant negative correlation is found at the formation site of STMW, which is consistent with our expectations (figure 3(a)). With one year of lag, a patch of negative correlation appears to the east of the STMW formation region (figure 3(b)). The statistically significant negative correlation occupies a broader area with one year of lag than with zero lag, and extends to near 150^∘ W. This indicates that a lower volume of newly formed STMW within the formation region can lead to greater MHW activity in the remote region during the following year. Although it shrinks toward the east, the 'blue patch' can still be observed in the remote region for up to three years (figures 3(c) and (d)). This result supports the idea that STMW in the formation area is advected eastward and affects the MHW activity in the remote region through reemergence. This was examined further using passive tracer numerical experiments.

Zoom In Zoom Out Reset image size

The spatial pattern of the lag correlation in figure 3 indicates that a portion of STMW can move eastward before reemerging to the surface and affecting MHWs in the following winter. This indicates that it takes nearly one year for STMW to travel over approximately 30^∘ to the east near 30^∘ N. In other words, the speed of this water mass is approximately 10 cm s⁻¹, which is consistent with observations [50, 51]. The eastward shift in the water mass decelerates after the first year, which has also been verified by drifter observations [51]. A previous study proposed that the STMW formed in winter is isolated by a seasonal thermocline and moves eastward in the subsurface layer. During this process, it undergoes the property change and affects SST in the remote region through entrainment to the surface [36].

To investigate the mechanism behind the time-lagged relationship between reemergence and MHWs, we conducted passive tracer experiments using the ocean states from ECCOv4r4. Two simulations were performed; one starting in March 1999 (Exp-99) and the other starting in March 2015 (Exp-15), when the annual volumes of STMW were low and high (figure A1(b)), respectively. We integrated the Massachusetts Institute of Technology general circulation model (MITgcm) [52] with the passive tracer, initially injected in the STMW in the formation region uniformly with a value of 100, for two years, driven by the surface forcing, initial condition, and model configurations from ECCOv4r4.

The results from Exp-99 revealed a patch of passive tracers moving eastward, which is consistent with eastward movement of the STMW under the influence of the subtropical gyre. In summer, the core of the patch was isolated at a depth of 100 m by the seasonal thermocline (figure 4(a)). In March 2000, the eastward-moving patch reached approximately 180^∘, where the passive tracers appeared in the surface layer. This indicates that the STMW at the subsurface was entrained to the surface layer by vertical mixing (figure 4(b)), although the longitudinal position of the tracer in the surface layer was shifted less toward the east than the negative correlation patch in the time-lagged correlation map (figure 3(b)). The reason for this difference in the reemergence location between the model experiment and correlation map is likely related to model discrepancies; ECCOv4r4 simulates a lower STMW volume and weaker cumulative MHW intensity than those in observation datasets (figure A1). After the first year, the passive tracer stagnated near 180^∘ in the subsurface layer (figure 4(c)). The passive tracer again appeared in the surface layer during the second winter, although its concentration was relatively low compared to that in the previous year because of dissipation (figure 4(d)). This passive tracer experiment clearly highlights the eastward shift of the tracers and its reemergence in the remote region.

Zoom In Zoom Out Reset image size

In Exp-15, the motion of the STMW exhibited different spatial distribution to that in Exp-99 (figures 4(e)–(h)). The eastward shift and reemergence of the tracers still appeared in the remote region, which was comparable to the result from Exp-99. However, in the following winter (figure 4(f)), the mean concentration of the tracer in the surface layer was much higher in Exp-15 than in Exp-99. This was because a larger portion of the STMW was located in the subsurface layer, which was evidenced by the 19 ^∘C isotherm occurring at a shallower depth than in Exp-99 (figures 4(a) and (e)). As a result, STMW was easily entrained to the surface layer during the following winter (figure 4(f)), lowering the MHW activity. The passive tracer experiments clearly demonstrate that the volume of STMW can affect surface conditions in the remote region in the following winters by moving along the ocean circulation. This supports the lagged negative correlation between the STMW volume in the formation region and MHW activity in the remote region. This is an important finding because it implies that MHWs in the TZCF can be predicted at least one year prior using the volume of STMW in the North Pacific formation region.

MHWs have recently attracted significant attention owing to their wide range of influence. These extreme events occur under various conditions and are controlled by air-sea heat exchange and oceanic heat transport. We demonstrated that MHWs and STMW are closely related to each other and have mutual influences. When a seasonal thermocline develops near the surface, STMW is kept beneath it and transported eastward by the North Pacific subtropical gyre while experiencing property changes. In the following winter, when the seasonal thermocline collapses, the water mass reemerges at the surface and affects surface conditions in the remote region. While remaining in the subsurface layer in the remote region, the water mass continues to be entrained to the surface two and three years after its formation, though the properties of the water mass continue to change due to mixing, subduction, and entrainment. Hence, a larger volume of STMW formed during the previous winter causes a greater volume of STMW to be preserved in the subsurface layer and regulate the conditions of the ocean surface during the winter when reemerged. In this case, the activity of MHWs in the remote region is reduced, as indicated by the negative time-lagged correlation and the passive tracer model experiments.

Interestingly, a statistically significant correlation was repeatedly observed in the remote region during every winter, for up to three years, suggesting that STMW can serve as a predictor of MHWs near the TZCF which is a boundary between low and high surface chlorophyll. The prediction of MHWs on an inter-annual timescale is possible owing to the long memory of the ocean. Behrens et al [53] also reported a prolonged predictive ability for MHWs in the Tasmanian Sea, using upper ocean heat content on inter-annual and multi-decadal timescales. Because this multi-year predictive ability is the result of interactions between the atmosphere and subsurface ocean during the winter, wintertime MHWs may be intrinsically different from those occurring in summer when the ocean surface is separated from the subsurface by strong stratification. The time-lagged correlation analysis proposed that the annual variation in STMW volume would be a crucial factor for forecasting the multi-year variation of MHWs near the TZCF in the North Pacific Ocean during the winter and early spring.

Oceanic teleconnection plays an important role in explaining the lagged correlation between MHWs and STMW. The subtropical gyre connects the STMW in the formation region with the MHWs in the remote region to the east, while reemergence connects the surface and subsurface ocean. The proposed mechanism is simple and applicable for perdicting MHWs in other STMW formation areas, such as the North and South Atlantic, and South Pacific Ocean. Future studies will investigate whether MHWs in these regions are indeed correlated with the volume of STMW, as in the North Pacific.

MHWs in the north of the Kuroshio Extension (along 36^∘–42^∘ N) is also important in terms of societal significance because of their intensity being the greatest in the North Pacific (figure 1(b)). This large intensity may be associated with a poleward displacement of the Kuroshio Current and mesoscale eddy activities [4]. In addition, the atmospheric factors, such as the surface net flux, may be also related to the MHWs because the SST change was largely contributed by the surface heat flux [54–56]. Thus, quantitative contributions of oceanic and atmospheric factors to MHWs in the north of the Kuroshio Extension will be further investigated.

This work was funded by the Korea Meteorological Administration Research and Development Program under Grant KMI (2020–01210). The authors are grateful to three anonymous reviewers for their constructive comments, which have helped improve the manuscript.

The data generated and/or analysed during the current study are not publicly available for legal/ethical reasons but are available from the corresponding author on reasonable request. The data that support the findings of this study are available upon request from the authors. AVHRR data was provided by GHRSST and the NOAA National Centers for Environmental Information. Dataset accessed (2020-08-23) at https://doi.org/10.5067/PATHF-MOD50. ECCOv4r4 is available at http://ecco-group.org/products-ECCO-V4r4.htm. EN4.2.1 data was obtained from the Met Office Hadley Centre and downloaded at http://metoffice.gov.uk/hadobs/en4/download-en4-2-1.html. A source code of MITgcm is available at http://mitgcm.org/public/source_code.html.

The seasonal variation in STMW volume was estimated to evaluate the validity of the ECCOv4r4 dataset (figure A1(a)). The volume increased from December to March, indicating that STMW mainly forms during this time period. Although ECCOv4r4 (blue line) shows a smaller volume of STMW than that shown by EN.4.2.1 (red line), the seasonal variation is consistent between both data products and comparable to the results reported in previous studies [13, 57]. The annual variation in STMW volume in March simulated from ECCOv4r4 data is also comparable to that from EN4.2.1 data with a significant coefficient of r = 0.76 (figure A1(b)). In addition to variations in STMW, the reproduction of MHWs in March is essential for evaluating the impact of STMW reemergence on MHWs. ECCOv4r4 simulates a similar annual variation of MHWs in the remote region (175^∘ E–160^∘ W, 24^∘ N–35^∘ N) in March to the variation shown by AVHRR satellite data with a significant coefficient of r = 0.95, although the mean cumulative intensity of 1.0 ^∘C days from ECCOv4r4 is smaller than the intensity of 1.4 ^∘C days from AVHRR (figure A1(c)).

Zoom In Zoom Out Reset image size