Potential impact of wintertime Arctic forcing on the subsequent sea surface temperature anomalies in the tropical eastern Pacific

Despite extratropical forcing being recognized as an important factor that can modulate El Niño-Southern oscillation (ENSO) properties on the interannual time scale, little is known about whether and how Arctic forcing changes the tropical sea surface temperature (SST). This current study reveals a significant link between the net surface sensible heat flux (SHF) in the Arctic and the SST anomalies in the tropical eastern Pacific (TEP). Specifically, anomalous upward SHF into the Arctic atmosphere in February leads to a warmer TEP in the subsequent spring and summer. A northeast-southwest-tilted North Pacific Oscillation-like atmospheric pattern associated with the upward Arctic SHF anomaly induces SST cooling in the subtropical North Pacific via positive Wind-Evaporation-SST feedback, which further promotes TEP SST warming via meridional heat advection, thermocline feedback, and nonlinear processes. The spring-to-summer TEP SST anomalies driven by the preceding anomalous Arctic SHF can potentially modulate the seasonal evolution of ENSO. Our findings imply that we should take into account the Arctic-tropics linkages when comprehensively understanding the ENSO variability and improving ENSO projection skills.


Introduction
El Niño-Southern oscillation (ENSO), the dominant mode of interannual tropical climate variability, has gathered significant attention due to the global reach of its influence.Typically, ENSO grows in boreal summer to autumn, matures in winter, and decays in the following spring and summer.Although ENSO peaks during the cold season, its influences during the warm season (both the onset and decaying phases) also has important implications for regional climates [1][2][3][4][5].For example, the fast decay of El Niño leads to a colder tropical eastern Pacific (TEP) in the subsequent summer, resulting in stronger rainfall over most of eastern China [4] via an intensified anomalous Western North Pacific anticyclone [6].In contrast, the slow decay of El Niño causes suppressed rainfall over northern and southern China but enhanced rainfall over the Yangtze River valley owing to a northeastward-shifted Western North Pacific anticyclone anomaly in the post-El Niño summer [4].Thus, understanding the onset and decaying rates of ENSO is of great scientific importance.
Extratropical forcings are important sources of diverse ENSO properties, especially the impacts of the North Pacific variability on the tropical sea surface temperature (SST) [7][8][9][10][11][12][13]. The North Pacific oscillation (NPO), characterized by a meridional dipole of sea level pressure (SLP) over the North Pacific [14], can induce SST anomalies that resemble the North Pacific meridional mode (NPMM) [15][16][17], significantly altering ENSO's timing, flavor, etc. [9,18,19], and therefore acting as a bridge between the mid-latitude forcings and ENSO [19,20].Different mechanisms were previously proposed for explaining how extratropical forcings influence ENSO, including the seasonal foot-printing mechanism [10,21], the trade wind charging [22], and ocean Kelvin waves [23,24].As is well recognized, the Arctic (further north compared to the aforementioned extratropical regions) undergoes dramatic changes and exhibits increasing interannual variability [25]; however, whether and how the Arctic can affect ENSO, in particular its life cycle, has received little attention.
Recently, several studies tried to build a link between Arctic variability and ENSO emergence.Nakamura et al [26] and Chen et al [27] suggested that the spring variability of Arctic oscillation (AO), the leading mode of Arctic atmospheric variability [28], is statistically correlated with the tropical Pacific SST in the following winter.Chen et al [29] proposed a physical process for this AO-ENSO connection: the positive AO in spring causes westerly anomalies over the tropical western Pacific by eddy-mean flow interaction, which triggers eastward-propagating Kelvin waves and warms the central Pacific and TEP in the following autumn, exciting the succeeding El Niñolike warming in the subsequent winter.Kim et al [30] suggested that spring sea-ice loss near the Bering Sea can contribute to the onset of the central Pacific El Niño in the following winter by stimulating a positive NPO-like atmospheric circulation pattern.Chen et al [31] proposed that the Greenland-Barents Seas sea-ice expansion in the preceding winter will lead to an El Niño-like warming in the tropical Pacific via the atmospheric Rossby waves and equatorial Kelvin waves.The conclusions of both Kim et al [30] and Chen et al [31] are consistent with an earlier study by Yeo et al [32], who linked the climate variability in the Bering Sea and large-scale circulations in the tropical Pacific.
In this study, we will present another pathway potentially connecting the Arctic and the tropical Pacific SST and discuss its influence on the ENSO cycle.Our results suggest that anomalous net surface sensible heat flux (SHF) on the Arctic surface into the atmosphere in February tends to result in a warmer TEP in the following spring and summer.Arctic surface heat flux anomalies are associated with the thinning sea ice and/or open ocean, and can change the air state above and further affect the remote atmospheric circulation [33].Climatological SHF field is upward across the Arctic in the cold season (figure 1(a)); thus, the stronger upward SHF represents that the atmosphere obtains more heat from the Arctic Ocean than usual.Given the strongest interannual variability of SHF among various surface fluxes including the latent heat flux and radiation fluxes (figure S1), we considered the SHF as a proxy indicating the forcing on the Arctic atmosphere from the Arctic Ocean.Note that Arctic sea-ice loss or atmospheric circulation patterns (e.g.AO), which are used as proxies for Arctic anomalies in previous studies on the Arctic-tropics link, are less indicative of causal responses [34].

Data
ERA5 [35], the most recent reanalysis from the European Centre for Medium-Range Weather Forecasts, assimilates advanced observations of SST, sea-ice concentration (SIC), surface air temperature, and other surface fields in the Arctic.Previous studies have evaluated its promising performance in the Arctic using independent observational records that are not assimilated in ERA5, such as those from radiosondes [36], research vessels and aircraft [37], though certain biases exist in the marginal ice zone.Thus, in this study, monthly net surface radiation and heat fluxes in the Arctic, together with SLP, 10 m horizontal wind and 10-m wind speed, were derived from ERA5.All the surface fluxes were multiplied by −1; thus positive (negative) anomalies denote upwards (downwards) anomalies across the air-sea/ice interface, representing the atmosphere gains (loses) heat from the bottom boundary.We also used monthly SST and SIC from the Met Office Hadley Centre's SST datasets [38].The wind stress, sea-ice thickness (SIT), depth of 17 • C isotherm (estimating the thermocline depth), and the ocean potential temperature at 0.5 m (PT_0.5m)are obtained from the Ocean Reanalysis System 5 (ORAS5) [39].The ocean current and sea surface height are provided by the National Centers for Environmental Prediction (NCEP) Global Ocean Data Assimilation System (GODAS) [40].
Our studying period is 1979-2020 with relatively plentiful and homogeneous observations.All anomalies are relative to the 1979-2020 mean.Linear trends are removed at each grid point for all variables to focus on the connections internally generated by the climate system.We referred to the area north of 60 • N as the Arctic, while using the area north of 65 • N or 67 • N will not strongly alter the correlations.

Oceanic mixed-layer heat balance
To diagnose the dominant physical processes changing the ocean temperature, we conducted a mixedlayer heat budget presented as follows: and where T is the ocean temperature, u, v, and w are the zonal, meridional, and vertical currents, respectively.Q net represents the net surface radiation and heat fluxes, that is, the sum of net longwave (LW), shortwave (SW), latent heat (LH), and sensible heat (SH) fluxes.ρ is the ocean water density, and C p is the specific heat of the water.H is the constant of the mean mixed layer depth (H = 50 m).Results are not sensitive to the value of H we choose, e.g.H = 70 m.
R is the residual term.The overbar and prime are the climatological mean and perturbations.

Eliassen-Palm (E-P) flux
The horizontal component of the E-P flux [41] was calculated by the following formula: In equation ( 3), u, v, and ϕ are horizontal geostrophic wind and geopotential height anomalies, respectively.R is the Earth radius, Ω is the rotation rate of Earth, g is the acceleration of gravity, and p is the ratio of pressure to 1000 hPa.φ and λ are latitude and longitude.The E-P flux was used as a diagnostic tool of the Rossby wave propagation associated with the Arctic forcing in this study.

Links between Arctic SHF and tropical SST variations
We defined the area-averaged SHF anomalies (positive upwards) north of 60 • N as the Arctic SHF index, which is displayed by the blue line in figure 1(d).
Its lead-lag correlations with the SST anomalies averaged over the TEP are shown in figure 1(c).There are two sticking relationships emerging from figure 1(c): the Arctic SHF index in February leads to warm TEP in the following spring and summer and the Arctic SHF index in March leads to cold TEP in the next autumn and winter.The latter is in line with the relationship proposed by Chen et al [27,29] and Chen and Chen [42] that the positive AO in March leads to an El Niño-like SST pattern over the tropical Pacific, as the small SHF index (representing less heat into the Arctic atmosphere from the bottom boundary) is generally coherent with cold air and low pressure over the Arctic and the positive phase of AO.The former has not been previously revealed and is the focus of the current study.Note that the upward SHF is largely limited in February due to a wide sea-ice cover in the Arctic Ocean.However, the February SHF exhibits more prominent variance on the interannual time scale compared with that in the warm season (figure S2).It is reasonable to examine the interannual relationship between the Arctic SHF in February and tropical SST.
According to figure 1(c), the Arctic SHF index in February has significant correlations with TEP SST anomalies from the following March to August above the 95% confidence level, with the highest value of 0.53 in April.Further analysis shows that coherent changes in the Arctic SHF index and TEP SST anomalies are widely seen from 1979-2020 rather than resulting from a few extreme cases (figure 1(d)).For example, in the years with particularly warm TEP SST in April (1983,1987,1992,1993,1998,2010,2015), positive anomalies of the Arctic SHF index are found in the prior February; more than half of the years with particularly cold TEP SST in April (1985,1988,1994,1999,2003,2007,2018) are preceded by negative anomalies of the Arctic SHF index in February.The robustness of this Arctic-tropic relationship is evidenced since the correlation is insensitive to whether the linear trend is excluded (figure S3) and applying different observational and reanalysis SST datasets (figure S4).Additionally, this SHF-SST correlation is weakened after the late 1990s, which will be briefly discussed in section 4.
Associated with the large Arctic SHF index in February, first the SST cooling anomaly appears in the subtropical North Pacific (figures 2(c)-(e)), and subsequently the SST warming anomaly develops off the coast of South America and then spreads across the TEP (figures 2(d)-(h)).The significant SST warming even extends to the tropical Atlantic in summer (figures 2(g) and (h)).In contrast, associated with the large Arctic SHF index in February, the tropical SST shows faint anomalies in the preceding winter months before February (figures 2(a) and (b)).This suggests that the Arctic SHF changes are unlikely a result of the preceding tropical SST anomalies, which is further confirmed by partial-correlation maps of the Arctic SHF index and the following tropical SST after removing effects of the preceding-winter Niño-3 index (figure S5).
Previous studies have built the linkage between the ENSO and the Arctic air-sea-ice system with a focus on Arctic atmosphere patterns, regional sea ice in the Arctic and so on [29][30][31][32].Here, we examine the correlation of the tropical SST anomalies with various variables which are usually used to measure the Arctic air-sea-ice state, e.g. the near-surface atmospheric temperature, sea-ice extent and subsurface oceanic temperature (figure S6).By contrast, the Arctic SHF has the strongest and most continuous correlation with the TEP SST in the following seasons, suggesting that the atmosphere will respond to heat fluxes directly, instead of to the atmospheric or ice state.Thus, we argue that the Arctic SHF could nicely symbolize the surface forcing into the Arctic atmosphere, which helps deduce the remote effects of Arctic forcing on the tropical SST anomalies.Next, we will attempt to establish a physical link for this Arctictropic relationship.

Physical processes linking the Arctic and TEP
The Arctic SHF anomaly in February is determined by the concurrent and preceding Arctic air-sea-ice interaction.Associated with large Arctic SHF index in February, negative anomalies of SIC and SIT are found across most of the Arctic Ocean (figures 3(b) and (c)), except some positive signals in the marginal seas (e.g. the Greenland Sea and Bering Sea).In fact, reduced sea ice over the Arctic Ocean is seen from the preceding October (figure S7).By the large thermal contrast between the Arctic air and its lower boundary, the continuous sea-ice loss over the Arctic Ocean and the large-scale low air temperature (figure 3(e)) jointly result in positive SHF anomalies (figure 3(a)) that represent more upward heat transferring from the underlying surface to the atmosphere [43].The anomalous heating in the Arctic atmosphere can excite eastwardpropagating Rossby waves along northern Canada and Eurasia (vectors in figure 3(f)), which possibly leads to high-pressure anomalies in northeastern Canada and low-pressure anomalies along northern Eurasia and in northwestern Canada (shadings in figures 3(f) and 4(a)), consistent with the subpolar atmospheric responses to the Arctic forcing in observational evidence and numerical simulations [44].One month after February, the lower-middle-level warming and positive geopotential height anomalies appear above the Arctic Ocean in March (figure S8), suggesting an active role is played by the SHF anomalies in forcing the thermodynamical and dynamical atmospheric anomalies in the Arctic.
Driven by the anomalous upward SHF, atmospheric circulation over the Pacific sector shows a subpolar low-pressure center lying over northwestern Canada, and a high-pressure center over the subtropical North Pacific (figure 4(a)).This North Pacific dipole pattern, to some extent, shares a resemblance to the spring NPO [45,46], but with a more northeast-southwest orientation.The southern lobe of this dipole pattern induces significant northeasterly anomalies between 180 • -150 • W, south of 20 • N (vectors in figure 4(a)), which can intensify the climatological trade wind, leading to an increase in the near-surface wind speed (contours in figure 4(a)).Significant SST cooling anomalies are found in the subtropical North Pacific near the dateline (shadings in figure 5(a)).The most pronounced SST cooling appears from February (shadings in figure 5(a)) and it becomes stronger in March (figure 2).
Consistently, the SST tendency shows a manifest negative anomaly east of the dateline from February to March (contours in figure 5(a)).Our diagnoses using the mixed-layer ocean heat budget show that the cooling tendency in the subtropical North Pacific is governed by the surface heat and radiation fluxes (T10; figure 5(c)), which largely represents an upward anomaly of LH anomalies (the inner panel in figure 5(c)) coupled with the increased 10 m wind speed (contours in figure 4(a)).These results indicate that the enhanced trade wind associated with the southern lobe of the NPO-like pattern triggers SST cooling in the subtropical North Pacific via the WES feedback mechanism.
Regarding causes of the TEP SST growth (figure 5(b)), our diagnosis underlies the primary impact of the meridional heat advection process (T5), particularly in April (figure 5(d)).In March, the SST cooling generating in the subtropical North Pacific (figures 2(d) and 5(a)) produces a negative meridional SST anomaly gradient north of the equator ( ∂T ′ ∂y < 0).Owing to the northward component of the climatological surface current between 20 • N and the equator (v > 0) (figure S9), the meridional heat advection process (T5 = −v ∂T ′ ∂y ) will heat up the ocean temperature in the TEP.The associated mixed-layer warming, in turn, further reinforces negative meridional SST anomaly gradients north of the equator.Therefore, a positive feedback is initiated, and the effect of the meridional heat advection process (T5) becomes stronger from March to April (figure 5(d)), which mainly contributes to the TEP SST warming in spring and summer.
We find that the equatorial westerly anomaly east of the dateline significantly reinforces after March (figures 4(b)-(d)), which may be excited by the equatorward Rossby waves propagation [47,48] and the combination of the WES feedback and tropical ocean dynamics [9,48,49].The reinforced anomalous westerlies produce eastward-propagating equatorial downwelling Kelvin waves given the apparent eastward-propagation of the positive sea surface height anomalies (figure 6(a)).The downwelling Kelvin waves can lead to a warm TEP by the warmwater advection [50] and thermocline feedback process (T8) associated with a deepening thermocline (figure 6(b)) [51,52].
We note that the effects of nonlinear advection processes (i.e.T3 and T9) should not be overlooked in TEP warming (figure 5(d)), although most previous studies focused on the linear feedback processes (e.g.T1, T7 and T8) in ENSO development [51,53].Furthermore, the effect of surface heat and radiation fluxes (T10) hinders the warming tendency in TEP (figure 5(d)), likely due to locally less downward SW which is closely related to more cloud cover (not shown).
So far, as summarized in figure 7, we found that the upward Arctic SHF anomaly in February results in SST cooling in the subtropical North Pacific by stimulating a northeast-southwest-shifted NPO-like atmospheric pattern.The coherent northeasterly anomalies over the subtropical North Pacific initiate the positive WES feedback, intensifying and maintaining the subtropical SST cooling from February to March.The subtropical SST cooling further induces westerly anomalies along the equatorial Pacific, triggering meridional heat advection feedback, thermocline feedback, and nonlinear heat advections, which warms the mixed-layer ocean in the TEP in April and results in a warm TEP in the late spring and subsequent summer.
It is noteworthy that the physical process linking the Arctic SHF and TEP SST changes is somewhat different from previous studies that examined how the spring subpolar signals triggers ENSO in the following winter [9,29,30,47].They claimed that the negative phase of NPMM or NPO in spring [54,55] can induce a La Niña-like pattern in the subsequent winter.In our study, the upward Arctic SHF anomaly in February results in the subsequent TEP SST warming via the North Pacific dipole circulation, similar to the negative phase of NPO but with a northeast-southwest orientation.We highlight the major contribution of meridional heat advection feedback (T5) due to the large subtropicaltropical SST gradient, which is different from the previously-proposed mechanisms that the equatorial ocean waves mainly give rise to the generation of ENSO once the subtropical SST anomalies emerge [9][10][11][12]29].

Potential impacts on the ENSO cycle
The above results suggest that Arctic SHF anomalies can affect the TEP SST in spring and summer, thereby potentially affecting the development, maintenance or transition of ENSO.Regardless of whether the Arctic forcing has indeed affected the ENSO events in recent decades, we identified 17 El Niño and 14 La Niña years for 1979-2020 based on the definitions provided by the NOAA Climate Prediction Center (CPC).Each type of event was classified into three groups based on the TEP SST state in the early spring as well as the seasonal phase: on the onset phase during the early spring, on the prolonged phase from the last winter, already on the decay phase in the early spring (see table S1).Details can be found at https://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php.Events that have relatively short duration less than or equal to 5 consecutive seasons (e.g.1979/80, 2005/06, etc.) were excluded.Years with the Arctic SHF index in February greater (smaller) than 1 (−1) standard deviation are marked by "+" ("−") in table S1.
In the years with El Niño events onset, the cases with large Arctic SHF index in February (colorful lines in figure S10(a)) exhibit similar TEP SST in winter but warmer TEP SSTs in the late spring and summer than the cases without a large Arctic SHF index in February (compare the solid black line with the dash black line in figure S10(a)), suggesting that the Arctic forcing tends to warm the TEP in advance and favors the emergence of El Niño.In the years with El Niño events heading toward a neutral or an opposite phase from winter to spring, the TEP warming in the cases without a large Arctic SHF index in February decays rapidly from January and shifts to the cooling phase after June (dash black line in figure S10(b)), whereas the TEP decaying rate in the cases with large Arctic SHF indexes in February seemingly becomes slower after the early spring (solid black line in figure S10(b)).This implies a slowdown decay and a delayed transition of El Niño events under the impact of a large February Arctic SHF.Moreover, it is interesting to note that the upward Arctic SHF anomaly may somehow facilitate the multi-year 1986/87/88 El Niño event (the blue line in figure S10(b)), since the positive Arctic SHF anomalies in 1987 can result in TEP SST warming in the subsequent spring and summer.By contrast, the influence of Arctic SHF anomaly on La Niña events is less clear, probably due to the fact that the La Niña years that coincide with significant Arctic SHF anomalies in February are rare in observations (table S1).It may also result from only anomalous upward SHF affecting the remote SST in the TEP since a negative Arctic SHF index represents less upward SHF into the atmosphere or downward SHF per se, which could hardly drive the changes in the Arctic atmosphere.Our results imply an asymmetry effect of the Arctic SHF anomaly on the TEP SST.Further investigations using large ensemble simulations or ocean-atmosphere coupled model simulations may help confirm and quantify the effect of Arctic SHF on the seasonal evolution of ENSO.

Summary and discussion
The Arctic has undergone dramatic changes in recent decades, but whether and how the Arctic variability is influencing the tropical climate system remains unclear.Previous studies have reported a link between the spring AO/Arctic sea-ice loss and the central Pacific ENSO generation in the following winter.In this study, we introduced a new pathway connecting the late winter Arctic and the spring and summer TEP.We found that the anomalous upward Arctic SHF in February, which indicates direct forcing from the Arctic surface to the local atmosphere, has a potential impact on the TEP SST via the NPO-like atmospheric circulation anomaly, Wind-Evaporation-SST feedback over the subtropical North Pacific, and feedback in the tropical Pacific mixed layer.
Our findings show that the effect Arctic SHF forcing on individual events is rather as ENSO are largely determined by the coupled oceanatmosphere state in the tropical Pacific.Nevertheless, in the presence of anomalous upward Arctic SHF in February, the budding El Niño events tend to grow faster in spring/summer, and the decaying El Niño events are inclined to persist into summer.These results have profound implications regarding the influof high-latitude systems on the seasonal of ENSO, which entails more in-depth and relevant research in the future.
According to figure 2, significant tripolar patterns in SST anomalies are clearly seen in the Atlantic Ocean, hinting that the Atlantic may be another pathway linking the Arctic SHF and TEP SST.We note that the El Niño-like pattern is preceded SST warming over the tropical northern Atlantic (figures 2(c)-(f)), which has been detected as a potential trigger for a La Niña event via the equatorial ocean dynamics [56,57].This suggests, to the contrary, that this Atlantic-related mechanism is unlikely to explain the Atlantic-Pacific connection observed in figure 2. We hypothesize that the Atlantic SST anomalies may have  modulated the TEP SST via the atmospheric bridge [58], e.g.changing the regional Hadley and Walker circulations (figure S11).
The link between Arctic SHF and TEP SST from winter to spring/summer is more promising before the late 1990s and seems weakened afterwards (figure 1(d)), as we mentioned in section 3.1.Their correlation reaches 0.72 during 1979-1998 but drops to 0.33 during 1999-2020.Further examination shows that the linkage between the Arctic SHF and NPO-like pattern is significantly weakened during 1999-2020 (not shown).Some studies have reported considerable interdecadal changes in the NPO properties (e.g. its periodicity, location, spatial distribution) around 1990s [44,59,60], which provides some clues to understanding the interdecadal reduction in the SHF-SST correlation.Overall, the exact physical mechanism that links the wintertime Arctic anomalies and the spring/summer tropical Pacific climate under different background states, especially how the Arctic SHF stimulates an NPO-like pattern or anchors the phase of the NPO, is not fully understood and calls for further investigation.

Figure 1 .
Figure 1.Characteristics of the Arctic sensible heat flux (SHF) and its connection with the sea surface temperature (SST) over the tropical eastern Pacific (TEP).(a) Climatological Arctic SHF (shading; interval: 15 W m −2 , positive upwards) and (b) its standard deviation (shading; interval: 5 W m −2 ) in the cold season (November-March) during 1979-2020.(c) Correlations between monthly-mean SHF timeseries averaged over the Arctic (SHF index) and SST timeseries averaged over the TEP (5 • S-5 • N, 120 • -60 • W) in the same year and the subsequent year, denoted by the numerals 0 and 1 in the subscript.Hollow circles, crosses and asterisks denote the correlations significant at the 90%, 95% and 99% confidence levels, respectively.(d) Normalized timeseries of the Arctic SHF index in February and the TEP SST in April.Their correlation coefficient and the p-value are shown in the top-right box.

Figure 2 .
Figure 2. SST anomalies related to large Arctic SHF index in February.Correlation maps between the Arctic SHF index inFebruary and SST anomalies from the preceding December (abbreviated as Dec-1) to the following July (Jul0) during the period 1979-2020.The stippling denotes the regions significant at the 95% confidence level.

Figure 3 .
Figure 3. Arctic anomalies related to large Arctic SHF index in February.Correlation maps between the Arctic SHF index in February and the simultaneous (a) SHF, (b) sea-ice concentration (SIC), (c) sea-ice thickness (SIT), (d) oceanic potential temperature at 0.5 m (PT_0.5m),(e) 2 m air temperature (T2m), and (f) sea level pressure (SLP, shading) over and around the Arctic.The vectors in (f) represent regressions of the horizontal component of Eliassen-Palm (E-P) flux at 850 hPa onto the Arctic SHF index in February.Only the vectors exceeding the 0.1 m 2 s −2 are shown.The stippling denotes the significance at the 95% confidence level.

Figure 4 .
Figure 4. Atmospheric circulation anomalies related to large Arctic SHF index in February.Regression maps of SLP anomalies (shading; interval: 0.3 hPa), wind stress anomalies (vectors; unit: N m -2 ), and 10 m wind speed anomalies (contours; interval: 0.2 m s -1 , starting from ±0.1 m s -1 ) in (a) February, (b) March, (c) April, and (d) May against the Arctic SHF index in February.The stippling estimates the significance in SLP at the 90% confidence level.The thick contour means the significance in 10 m wind speed at the 90% confidence level.Purple (orange) contours denote positive (negative) values of 10 m wind speed.Only the wind stress exceeding 90% confidence level are shown as vectors.

Figure 5 .
Figure 5. Mixed-layer heat balance in the subtropical North Pacific (NP) and the TEP related to large Arctic SHF index in February.(a)-(b) Regression maps of NP SST anomalies (shading; interval: 0.05 • C) and the tendencies (contours; interval: 0.04 • C month -1 , starting from ±0.02 • C month -1 ) averaged along (a) 10 • -20 • N, and (b) 5 • S-5 • N from January to June against the Arctic SHF index in February.The stippling denotes the significance at the 90% confidence level.(c)-(d) Individual terms in the mixed-layer heat budget from equation (1) (bar; unit: • C month -1 ) averaged over (c) the subtropical North Pacific (10 • -20 • N, 170 • E-120 • W) in February and (d) the tropical Pacific (5 • S-5 • N, 150 • -60 • W) in March and April against the Arctic SHF index in February.According to equation (2), the term 10 in (c) is further decomposed into net surface fluxes of longwave radiation (LW), shortwave radiation (SW), latent heat (LH), and sensible heat (SH), as shown in the inner panel.

Figure 6 .
Figure 6.Tropical sea surface height and thermocline depth anomalies associated with the large Arctic SHF index in February.Regressed (a) sea surface height anomalies (interval: 0.5 cm) and (b) thermocline depth anomaly (interval: 1 m) averaged along 5 • S-5 • N from January to June against the Arctic SHF index in February.The zeros contour is highlighted by the dashed line.The stippling denotes the significance at the 90% confidence level.

Figure 7 .
Figure 7. Schematic diagram of the physical processes responsible for the linkage between anomalous upward SHF over the Arctic in February and the El Niño-like pattern in following spring and summer.Black phrases show the four steps linking Arctic SHF and tropical SST anomalies, while green phrases highlight the physical processes controlling the SST evolution in the third and fourth steps.