1 Introduction

The ocean plays a crucial role in the evolution of tropical cyclones (TCs) not only because the ocean supplies water vapor to TCs (e.g. Riehl 1950; Emanuel 2003), but also because TC-induced sea surface cooling (e.g. Leipper 1967) inhibits development of TCs (e.g. Bender et al. 1993; Schade and Emanuel 1999; Ginis 2002; Cione and Uhlhorn 2003; Shen and Ginis 2003). Since the sea surface cooling is expected to be strongly influenced by the oceanic stratification (e.g. Lloyd and Vecchi 2011; Vincent et al. 2012a, b), it is important to investigate the upper ocean structure prior to the passage of TCs.

In this regard, Leipper and Volgenau (1972) introduced the tropical cyclone heat potential (TCHP), which is a measure for the amount of heat stored in the upper ocean above the 26 °C isotherm. Since TCs develop when sea surface temperature (SST) is above 26 °C (e.g. Palmén 1948; Gray 1968; Emanuel 2003), which is an average temperature in the subtropical atmospheric boundary layer, SST-26 °C related to the thermal disequilibrium between SST and the air temperature (Price 2009) is used as the threshold. Although the TCHP is useful in considering potential impacts of the upper ocean (e.g. Wada and Usui 2007; Lin et al. 2008; Mainelli et al. 2008), several issues have been raised (Price 2009). First, although this measure is based on “vertically integrated temperature”, vertical mixing that plays the dominant role in the sea surface cooling (e.g. Price 1981; D’Asaro et al. 2007) is better represented by “vertically averaged temperature”. In addition, the TCHP cannot be defined where SST is below 26 °C due to the use of 26 °C as the threshold temperature, although the sea surface cooling in such oceans is of interest. Furthermore, this measure does not account for salinity stratification, which often makes important contribution to stability of the upper ocean in the tropics.

More recently, Vincent et al. (2012b) proposed a unique measure that is based on the change in upper ocean potential energy caused by TCs (see Sect. 2 for more details about the definition of this measure). This measure is superior in the sense that it can incorporate effects of haline stratification and uses “vertically averaged temperature” that is more appropriate for the sea surface cooling associated with vertical mixing. Taking advantage of this new measure, Neetu et al. (2012) showed that the haline effect accounts for as much as about 50% of seasonal variations in the inhibition of the sea surface cooling during the post-monsoon season in the Bay of Bengal. Regarding the northwestern Pacific, where about one third of tropical storms develop (Gray 1968), many studies had examined influences of the El Niño/Southern Oscillation (ENSO) mainly from the atmospheric viewpoint (e.g. Chan 1985; Camargo and Sobel 2005; Iizuka and Matsuura 2008), but approaches from the oceanic viewpoint (e.g. Zheng et al. 2015; Gao et al. 2022) are relatively few. In this regard, Vincent et al. (2014) were the first to study the influences of the ENSO on upper ocean stratification using that measure. Based on a dynamical downscaling approach, they suggested that interannual variations in oceanic stratification associated with the ENSO may modulate the strength of strongest TCs by 10 m/s. However, they did not quantify the relative role of thermal and haline stratification even though salinity stratification may strongly influence the sea surface cooling (Neetu et al. 2012) and TC intensification under global warming (Balaguru et al. 2016) in the northwestern Pacific, and tropical SST is known to be significantly influenced by anomalies of upper ocean salinity through anomalous freshwater fluxes (e.g., Lukas and Lindstrom 1991) associated with ENSO (e.g. Vialard et al. 2002; Maes et al. 2006; Zheng and Zhang 2012). Moreover, Vincent et al. (2014) did not distinguish between the development and decay phases of El Niño and La Niña, despite the known differences in oceanic conditions (e.g. Bjerknes 1969; Delcroix 1998).

Motivated by the above, using an ocean reanalysis product, this study investigates interannual variations in potential impacts of the upper ocean stratification on TC-induced sea surface cooling associated with the evolution of El Niño and La Niña events, with a special focus on the role of haline stratification. This paper is organized as follows. A brief description of ocean reanalysis products and observational datasets, and methods is provided in Sect. 2. Section 3 describes main results with some supporting analyses. The final section provides a summary and discusses potential implications of this study. Since the main focus of this paper is to present a new measure and discuss potential impacts of the upper ocean salinity stratification, we briefly discuss how the proposed measure may be further validated with observed TCs and/or numerical simulations.

2 Data and methods

2.1 Ocean reanalysis products and observational data

This study mainly uses the Ocean Reanalysis System 5 (ORAS5) (Zuo et al. 2019) for monthly temperature and salinity during 1989–2018. The horizontal resolution is 1°x1° and the number of vertical levels is 75. Since salinity observations were quite limited before the Argo era and there may be large uncertainty in the salinity field, we have conducted the same analyses by using the Simple Ocean Data Assimilation version 3.15.2 (SODA 3.15.2) (Carton et al. 2018), and the objective analyses of the EN 4.2.2 c14 (Good et al. 2013; Cheng et al. 2014; Gouretski and Cheng 2020) to test the robustness of our results obtained from ORAS5. The horizontal resolution and the number of vertical levels are respectively 0.5°x0.5° and 50 for SODA 3.15.2, and 1°x1° and 42 for EN 4.2.2 c14. Since results are qualitatively the same in general, we mainly present results from ORAS5, and results from the other two are presented in the Supplementary Information (Figs. S1, S2).

To investigate possible mechanisms of interannual variations in the upper ocean salinity, we also use the CPC Merged Analysis of Precipitation (CMAP) (Xie and Arkin 1997) during 1989–2018 for precipitation. This monthly data is based on gauge and satellite estimates. Its horizontal resolution is 2.5°x2.5°.

2.2 Change in the upper ocean potential energy ΔEp

To investigate potential impacts of the upper ocean stratification on sea surface cooling induced by TC passages, Vincent et al. (2012b) proposed \(\:\varDelta\:{E}_{p}\) defined by

$$\:\varDelta{E}_{p}\left(\varDelta{T}\right)={\int}_{0}^{{h}_{m}(\varDelta{T})}\left({\rho}_{i}\left({T}_{i},\:{S}_{i}\right)-{\rho}_{f}\left({T}_{f},\:{S}_{f}\right)\right)gzdz.$$
(1)

Here, the density of the seawater \(\:\rho\:\) is derived from potential temperature \(\:T\) and salinity \(\:S\), the subscripts \(\:i\) and \(\:f\) represent before and after a TC passage, respectively, \(\:g\) is the acceleration due to gravity, and \(\:{h}_{m}(\varDelta\:T)\) is the mixing depth to which the upper ocean is homogeneously mixed by a TC. The vertically averaged temperature from the surface to \(\:{h}_{m}\) is \(\:\text{S}\text{S}\text{T}-\varDelta\:T\) (Fig. 1). This measure can incorporate effects of haline stratification, and uses “vertically averaged temperature” that is more appropriate for the SST cooling associated with vertical mixing. However, \(\:\varDelta\:{E}_{p}\) does not directly relate the strength of a TC and induced sea surface cooling. Thus, we here propose a new measure that is based on an assumption that \(\:\gamma\:\)(= 2%) of the total energy input from a TC is used for the vertical mixing (Oey 2021):

Fig. 1
figure 1

Schematic temperature profiles before (blue line) and after (green line) a tropical cyclone (TC) passage. Here, \(\:{h}_{i}\) is the initial mixed layer depth (MLD), and \(\:{h}_{m}\) is the mixing depth to which the upper ocean is homogeneously mixed

Full size image
$$\ {\int}_{0}^{{h}_{m}\left(\varDelta{T}\right)}\left({\rho}_{i}\left({T}_{i},{S}_{i}\right)-{\rho}_{f}\left({T}_{f},\:{S}_{f}\right)\right)gzdz=\gamma{\rho}_{a}{C}_{d}{V}^{3}\varDelta{t}.$$
(2)

Here, \(\:{\rho\:}_{a}\) (= 1.26 kg/m3) is the density of the air, \(\:{C}_{d}\) is the drag coefficient, \(\:V\) is surface wind speed, and \(\:\varDelta\:t\) is the time during which a given location is influenced by a TC. Although these parameters vary between TCs, here we have simply set \(\:{C}_{d}\) = 2 × 10−3 and \(\:\varDelta\:t\) = 105 s, assuming that winds at a constant speed blow about one day. To estimate the amplitude of sea surface cooling against the strength of a TC, we have first calculated the right-hand side of Eq. (2), and derived \(\:{h}_{m}\) as a function of \(\:V\). Then, we have obtained \(\:\varDelta\:T\) by using the vertically averaged temperature from the surface to \(\:{h}_{m}\).

2.3 Salinity effect

To quantify the thermal effect without salinity, this study proposes \(\:\varDelta\:{T}^{temp}\), which is inversely derived from \(\:{h}_{m}^{temp}\):

$$\ {\int}_{0}^{{h}_{m}^{temp}\left({\varDelta\:T}^{temp}\right)}\left({\rho}_{i}\left({T}_{i},{S}_{f}\right)-{\rho}_{f}\left({T}_{f},{S}_{f}\right)\right)gzdz=\gamma{\rho}_{a}{C}_{d}{V}^{3}\varDelta{t}.$$
(3)

Here, \(\:{h}_{m}^{temp}\) is the mixing depth to which the upper ocean is homogeneously cooled to \(\:\text{S}\text{S}\text{T}-{\varDelta\:T}^{temp}\) by a TC. The right-hand side of Eq. (3) is the same as that of Eq. (2). Then, the salinity effect on the sea surface cooling is estimated by \(\:{s}_{T}\) as follows:

$$\ {s}_{T}=\frac{\varDelta{T}-{\varDelta{T}}^{temp}}{\varDelta{T}}.$$
(4)

2.4 Definition of El Niño and La Niña

In this study, the Niño-3 index, which is defined as the five-month running mean of SST anomalies averaged over the eastern equatorial Pacific (5°S-5°N, 150°W-90°W), is used to define El Niño and La Niña. If the Niño-3 index were higher (lower) than + 0.5℃ (-0.5℃) for six consecutive months or longer, the event is defined as El Niño (La Niña). For each El Niño event, the developing (decaying) year is determined by the increase (decrease) in the Niño-3 index across + 0.5℃ at the beginning (end) of the consecutive months. For a La Niña event, the developing (decaying) year corresponds to a year with a decrease (an increase) in the Niño-3 index across − 0.5℃ at the beginning (end) of the consecutive months. Table 1 summarizes the developing and decaying years of El Niño and La Niña.

Table 1 Developing and decaying years of El Niño and La Niña
Full size table

3 Results

3.1 Climatological distributions

To examine seasonality of \(\:\varDelta\:T\), Fig. 2 shows the mean \(\:\varDelta\:T\) for \(\:V\)= 22 m/s in each month from March to November. Areas of large \(\:\varDelta\:T\) greater than 2 °C are found especially to the north of 20°N from May to October, where the strong seasonal thermocline develops. Also, \(\:\varDelta\:T\) is relatively large over the South China Sea. Moreover, to examine seasonality of interannual variations of \(\:\varDelta\:T\), Fig. 3 shows the ratios of the standard deviation in \(\:\varDelta\:T\) to the mean \(\:\varDelta\:T\) for each month. Organized large values are seen to the south of 20°N throughout a year. Although most TCs do not develop equatorward of 10°N (Goni et al. 2009; Knapp et al. 2010), this suggests that \(\:\varDelta\:T\) has stronger interannual variability to the south of 20°N. Since the peak typhoon season is July-October (JASO) (Gao et al. 2022), we focus on this season. We note that results are qualitatively the same even when \(\:V\) is varied (\(\:V\) = 18, 22, 26, and 30 m/s) (Fig. 4), and we thus present results for \(\:V\)= 22 m/s.

To examine the climatological distribution of \(\:\varDelta\:T\) and the relative contributions of thermal and haline stratification, mean \(\:\varDelta\:T\), \(\:\varDelta\:{T}^{temp}\), their difference \(\:\varDelta\:T-\varDelta\:{T}^{temp}\), and \(\:{s}_{T}\) in JASO are shown in Fig. 5. The distributions of \(\:\varDelta\:T\) and \(\:\varDelta\:{T}^{temp}\) are qualitatively similar, but equatorward of 20°N, their difference becomes relatively large (\(\:\varDelta\:T-\varDelta\:{T}^{temp}\:\)~ -0.3℃) with non-negligible values of about − 0.2 for \(\:{s}_{T}\). This suggests that the salinity stratification is also important for the potential impacts of the upper ocean on the sea surface cooling.

To investigate the effects of stratification on \(\:\varDelta\:T\) in more detail, we have prepared mean meridional cross sections of density, temperature, and salinity at 135°E (Fig. 6). Given that the vertical structure is qualitatively similar zonally from 135°E to 180°E equatorward of 20°N, we present profiles at 135°E. The density stratification below the mixed layer depth, which is defined as the depth at which the potential density is 0.125 kg/m3 larger than at the surface, is weaker in 10°N-20°N than that to the north of about 20°N, and this is related to deeper \(\:{h}_{m}\) (dashed line in Fig. 6) and smaller \(\:\varDelta\:T\) (Fig. 5a). The change in density stratification is largely determined by that in thermal stratification, and this relationship between \(\:\varDelta\:T\) and strength of thermal stratification below the mixed layer is qualitatively consistent with Vincent et al. (2012a, b). However, equatorward of about 20°N, distinct salinity stratification below the mixed layer is found (Fig. 6c) with large amplitudes of negative \(\:{s}_{T}\) (Fig. 5d), suggesting that haline stratification could also be important to suppress the sea surface cooling.

Fig. 2
figure 2

Mean \(\:\varDelta\:T\) (in ℃) for \(\:V\) = 22 m/s in each month from March to November

Full size image
Fig. 3
figure 3

Ratios of the standard deviation of \(\:\varDelta\:T\) to the mean \(\:\varDelta\:T\) for \(\:V\) = 22 m/s in each month from March to November

Full size image
Fig. 4
figure 4

Mean \(\:\varDelta\:T\) (in ℃) for \(\:V\) = (a) 18 m/s, (b) 22 m/s, (c) 26 m/s and (d) 30 m/s in July-October (JASO). The gray shading indicates regions where the values are not defined

Full size image
Fig. 5
figure 5

Mean (a) \(\:\varDelta\:T\), (b) \(\:\varDelta\:{T}^{temp}\), (c) \(\:\varDelta\:T-\varDelta\:{T}^{temp}\) (in ℃), and (d) \(\:{s}_{T}\) for \(\:V\) = 22 m/s in JASO

Full size image
Fig. 6
figure 6

Mean meridional cross sections of (a) potential density (in kg/m3), (b) potential temperature (in ℃), and (c) salinity (in psu) along 135°E in JASO. The white solid and dashed lines indicate the mean MLD and the mean mixing depth (\(\:{h}_{m}\)) for \(\:V\) = 22 m/s in JASO, respectively

Full size image

3.2 Interannual variations

Although the peak typhoon season is JASO (Gao et al. 2022), the decaying and developing signals associated with the ENSO are relatively stronger in the first half (July-August or JA) and the second half (September-October or SO) of the season, respectively. Indeed, significant and large amplitudes of \(\:{s}_{T}\) anomalies during the decaying years are more distinct in JA than in SO (Fig. S3, S4, S5, S6). Since our main focus is to examine how the evolution of the ENSO affects interannual variations of \(\:\varDelta\:T\), we have prepared composites in JA for the decaying years and SO for the developing years of El Niño and La Niña (Fig. 7).

Except for the opposite sign, the developing years of El Niño and La Niña show similar anomaly spatial patterns. More specifically, significant and organized positive (negative) anomalies are found to the east of about 120°E (135°E) during the developing years of El Niño (La Niña). On the other hand, positive (negative) anomalies are relatively weak and are limited to the west of about 150°E and the eastern South China Sea during the decaying years of El Niño (La Niña). Also, there are strong negative anomalies off Vietnam during the decaying El Niño and developing La Niña years.

Fig. 7
figure 7

Composites of \(\:\varDelta\:T\) anomalies (in ℃) in September-October (SO) of developing (a) El Niño and (c) La Niña years, and in July-August (JA) of decaying (b) El Niño and (d) La Niña years. The green dots indicate anomalies significant at the 90% confidence interval by a two-tailed t-test

Full size image

To examine contributions from interannual variations of salinity to \(\:\varDelta\:T\), we have constructed composites of \(\:\varDelta\:{T}^{temp}\) (Fig. 8) and \(\:{s}_{T}\) (Fig. 9) anomalies. Composites of \(\:\varDelta\:{T}^{temp}\) anomalies south of 20°N resemble those of \(\:\varDelta\:T\) anomalies, suggesting that the thermal stratification makes a dominant contribution. However, composites of \(\:{s}_{T}\) anomalies show sizeable contributions of haline stratification to \(\:\varDelta\:T\) anomalies in some regions. More specifically, significantly positive anomalies of \(\:{s}_{T}\) are found to the west of about 160°E during the developing and decaying years of El Niño (Fig. 9a, b). On the other hand, statistically significant organized anomalies are limited in the South China Sea and near the equator to the east of about 150°E during the developing years of La Niña (Fig. 9c). The decaying years of La Niña are close to a mirror image of the decaying years of El Niño. Although the peak amplitude of ~ 0.2 is also similar, negative anomalies during the decaying years of La Niña are less significant (Fig. 9d). Thus, due to anomalous haline effects, the upper ocean to the west of about 160°E is more susceptible to the sea surface cooling during the developing and decaying years of El Niño (Fig. 9a, b), while the sea surface cooling in the same region could be suppressed during the decaying years of La Niña (Fig. 9d).

Fig. 8
figure 8

As in Fig. 7, but for \(\:\varDelta\:{T}^{temp}\) (in ℃)

Full size image
Fig. 9
figure 9

As in Fig. 7, but for \(\:{s}_{T}\)

Full size image

Since the haline effect on \(\:\varDelta\:T\) stems from salinity stratification in the upper ocean, we have prepared vertical cross sections of salinity anomalies along 135°E (Fig. 10), because the peak \(\:{s}_{T}\) anomalies during the developing El Niño years and the decaying El Niño and La Niña years are found around this longitude. Significantly positive (negative) anomalies of salinity are found in the upper 50 dbar to the south of 20°N during the developing and decaying years of El Niño (decaying years of La Niña) (Fig. 10).

Fig. 10
figure 10

As in Fig. 7, but for meridional cross sections of composited salinity anomalies (in psu) along 135°E. The white solid (dashed) line indicates composites of MLD (\(\:{h}_{m}\))

Full size image

On the other hand, associated with anomalous shoaling (deepening) of the thermocline, strong negative (positive) temperature anomalies with their peak in the subsurface are found in the equatorial region during the developing El Niño (La Niña) years (Fig. 11a, c). Also, weak positive (negative) temperature anomalies are found around 20°N during the decaying El Niño (La Niña) years with positive temperature anomalies during the decaying El Niño extending deeper (Fig. 11b, d).

Fig. 11
figure 11

As in Fig. 10, but for potential temperature anomalies (in ℃)

Full size image

To examine the temperature and salinity effects on density, Fig. 12 shows composites of potential density anomalies along 135°E. In the upper 50 dbar, density anomalies are significantly positive during the decaying years of El Niño south of about 15°N, suggesting that positive salinity anomalies (Fig. 10b) dominate positive temperature anomalies (Fig. 11b). Since positive density anomalies in the upper ocean result in weaker density stratification and lead to positive \(\:\varDelta\:T\) anomalies, these results suggest that during the decaying years of El Niño (Fig. 12b), \(\:\varDelta\:T\) in this region is influenced by haline stratification more strongly than other regions. During the decaying La Niña years, weak negative salinity anomalies (Fig. 10d) and weak negative temperature anomalies (Fig. 11d) seem to compensate each other and result in no significant density anomalies (Fig. 12d). During the developing years of El Niño and La Niña, \(\:\varDelta\:T\) to the south of about 15°N is dominantly influenced by thermal stratification, resulting in small amplitudes of \(\:{s}_{T}\) especially during the developing La Niña years (Fig. 9c).

We note that interannual variations in the contribution of salinity anomalies (Zhang and Busalacchi 2009; Zhi et al. 2019) to the sea surface cooling may depend on events. To the south of 20°N, salinity anomalies that weaken the upper ocean stratification during the decaying El Niño year of 1998 may lead to about 25% increase in the salinity effect on the sea surface cooling (Fig. 13a). On the other hand, such as during the decaying La Niña year of 2011, the salinity effect on the sea surface cooling in 120°E-150°E may be suppressed by as much as about 40% due to enhanced haline stratification (Fig. 13b).

Fig. 12
figure 12

As in Fig. 10, but for potential density anomalies (in kg/m3)

Full size image
Fig. 13
figure 13

Anomalies of \(\:{s}_{T}\) in JA for (a) the decaying El Niño of 1998 and (b) the decaying La Niña of 2011

Full size image

3.3 Potential mechanism for upper ocean salinity variations

Since salinity near the surface makes an important contribution to \(\:{s}_{T}\) (Fig. 9), we have prepared composites of sea surface salinity (SSS) anomalies associated with the ENSO (Fig. 14). The horizontal distributions of anomalous SSS are qualitatively consistent with \(\:{s}_{T}\); positive SSS anomalies cover almost all western tropical Pacific during the decaying years of El Niño, while negative SSS anomalies are found mostly to the west of 135°E during the decaying years of La Niña. Significant SSS anomalies are confined to the equatorial region during the developing years of El Niño and La Niña.

Since precipitation primarily explains the evolution of SSS in the tropical oceans (Delcroix et al. 1996), composites of precipitation anomalies are prepared to examine potential mechanisms of SSS anomalies. As is well known (e.g. Delcroix 1998), the ENSO causes significant precipitation anomalies in the northwestern Pacific associated with the anomalous Walker circulation, and the patterns of the anomalies in June-July for the decaying years and in August-September for the developing years (Fig. 15) are mostly consistent with that of SSS anomalies. In particular, negative precipitation anomalies are found over positive SSS anomalies during the decaying years of El Niño and positive precipitation anomalies over negative SSS anomalies during the decaying years of La Niña. This suggests that SSS and its impact on \(\:{s}_{T}\) are largely influenced by precipitation anomalies associated with the ENSO.

Fig. 14
figure 14

As in Fig. 7, but for sea surface salinity (in psu)

Full size image
Fig. 15
figure 15

As in Fig. 7, but for precipitation (in mm/day) in June-July for the decaying years and in August-September for the developing years

Full size image

4 Summary and discussions

Using an ocean reanalysis product during 1989–2018, this study has investigated interannual variations in potential impacts of the upper ocean stratification on TC-induced sea surface cooling associated with the evolution of El Niño and La Niña events, with a special focus on the role of haline stratification. For this purpose, based on an assumption that 2% of the total energy input from a TC is used for the vertical mixing (Oey 2021), we extend the measure of Vincent et al. (2012b) so that the amplitude of TC-induced sea surface cooling (\(\:\varDelta\:T\)) is crudely but directly estimated as a function of the TC intensity represented by the surface wind speed.

First, seasonality of interannual variations in \(\:\varDelta\:T\) is investigated. From March to October, the ratios of the standard deviation in \(\:\varDelta\:T\) to the mean \(\:\varDelta\:T\) for each month are large to the south of 20°N, suggesting that \(\:\varDelta\:T\) has stronger interannual variability in this region. Although thermal stratification contributes more dominantly to \(\:\varDelta\:T\), haline stratification makes a sizeable contribution (~ 20%) in the South China Sea and equatorward of 20°N in the western tropical Pacific during the peak typhoon season (JASO) (Gao et al. 2022).

To examine interannual variations of \(\:\varDelta\:T\) associated with the evolution of the ENSO, we have prepared composites in JA for the decaying years and SO for the developing years of El Niño and La Niña considering the seasonality of ENSO. More specifically, ENSO events generally start to develop from boreal spring to summer and thus ENSO signals become more conspicuous in the latter half of the peak season during the ENSO developing years. On the other hand, ENSO events tend to decay in the following spring to summer and thus stronger anomalies are remaining in the first half of the peak season during the ENSO decaying years. It is found that the developing (decaying) years of El Niño and the developing (decaying) years of La Niña show similar anomaly spatial patterns in \(\:\varDelta\:T\) except for the opposite sign. Positive (negative) anomalies in \(\:\varDelta\:T\) are relatively weak and are limited to the west of about 150°E and the eastern South China Sea during the decaying years of El Niño (La Niña). Furthermore, significantly positive anomalies in haline effects \(\:{s}_{T}\) are found to the west of about 160°E during the developing and decaying years of El Niño. On the other hand, significant organized anomalies are limited in the South China Sea and near the equator to the east of about 150°E during the developing years of La Niña. Negative anomalies during the decaying years of La Niña are close to a mirror image of the decaying years of El Niño and with a similar peak amplitude, but less statistically significant. Thus, due to anomalous haline effects, the upper ocean to the west of about 160°E is more susceptible to the sea surface cooling during the developing and decaying years of El Niño, while the sea surface cooling in the same region could be suppressed during the decaying years of La Niña. Further analyses have revealed that negative precipitation anomalies in the western tropical Pacific lead to near-surface positive salinity anomalies and destabilize the upper ocean especially during the decaying years of El Niño. However, the contribution of anomalous haline stratification to the amplitude changes in TC-induced sea surface cooling associated with the ENSO is about 40% at most. Although the effect of haline stratification has been found less important than that of thermal stratification, potential impacts of upper ocean salinity on TC-induced sea surface cooling associated with the ENSO have been quantitatively estimated for the first time.

We note that there are some caveats to this study. We have discussed the generation mechanism of salinity anomalies only in a qualitative manner by checking precipitation anomalies, but an online salinity budget analysis using an ocean model (e.g. Kido et al. 2019) that can realistically simulate interannual variations of salinity may allow us to more quantitatively examine the mechanism. Furthermore, we have only investigated potential impacts of the upper ocean stratification on TC-induced sea surface cooling. In realistic situations, however, the sea surface cooling occurs not only through vertical mixing, but also through air-sea heat fluxes, upwelling, and horizontal advection (e.g. Vincent et al. 2012a). Therefore, it is necessary to conduct oceanic mixed layer heat budget analyses through numerical simulations forced by realistic or idealized TCs to verify the potential impacts. For example, sensitivity experiments in which an ocean model is initialized with anomalous oceanic stratifications associated with different phases of the ENSO and forced by the same TC forcing may be conducted. Also, we currently do not have a sufficient number of TCs that have the same characteristics (e.g. maximum wind speed, translation speed, and size) and pass through regions with statistically significant anomalies in \(\:\varDelta\:T\) during developing and decaying years of El Niño and La Niña. However, if we could accumulate more observations, it may become possible to directly verify the potential impacts with observational data.