Modulation of the cold tongue mode (CTM) in tropical cyclone genesis over the western North Pacific during 1975–2021

The cold tongue mode (CTM) in the tropical Pacific enhanced with the global warming in the recent decades. Here, we show that the change of oceanic thermal condition and atmospheric circulation plays a critical role in the tropical cyclones (TCs) genesis in the western North Pacific (WNP). The frequency of TCs in the WNP decreases with its northwest movement, when the CTM transitions from its negative phase in 1975–1997(P1) to positive phase in 1998–2021 (P2). Intercomparison of the TC activities between P1 and P2, we found that the main reason responsible for decreasing of TC genesis is from reducing of meridional and zonal heat transportation. And the northwest movement of TC position is driven by the heat increasing at the western and northern boundary areas. In addition, the northwest movement of TC is dominated by the atmospheric circulation associated with local Walker circulation and Hadley cell.


Introduction
Tropical cyclones (TCs) are non-frontal and form over the tropical ocean with a warm-core structure (Frank and Roundy 2006).The tropical western North Pacific (WNP), one of the warming pools, experiences the most frequent TC activity globally.As one of the most destructive weather conditions, TCs cause property damage and human casualties in coastal areas (Peduzzi et al 2012, Lin et al 2014).Therefore, it is important to study the climate change of TC activity in WNP in the context of global warming (Goldenberg et al 2001, Mei et al 2015, Sun et al 2017).Gray (1968) noted the thermal conditions for TC generation: midlevel moisture coupled with deep conditional instability and warm oceanic mixed layer.Due to global warming, TC activity in the WNP has undergone considerable changes (Emanuel 2005).Since the late 1990s, the frequency of TC generation in WNP has decreased, shifting northwestward (Choi et al 2015, He et al 2015, Cao et al 2020).After 1998, the Interdecadal Pacific Oscillation shifted into a negative phase and a La Niña-like distribution of sea surface temperature anomalies (SSTAs) occurred in the Pacific, leading to a decrease in the frequency of TCs in the WNP (Zhao et al 2018).Huangfu et al (2018) found that the weakening of convective activity over the Central Pacific in the late 1990s also led to an interdecadal decrease in the frequency of TCs in the WNP.
Zhang et al (2010) first decomposed SSTAs in the tropical Pacific using an empirical orthogonal function (EOF) and found that the second mode of SSTAs exhibited a seesaw pattern with warming in the western Pacific and cooling over the tropical eastern Pacific.This mode is known as the cold tongue mode (CTM) and exhibits an increasing trend with global warming (Solomon and Newman 2012).The long-term cold trend of the CTM is mainly caused by the response of the ocean dynamic feedback to global warming.Therefore, the CTM is strongly related to global warming (Li et al 2015).When a strong positive CTM is superposed on El Niño, the eastern equatorial cooling SSTA in the CTM weakens the eastern equatorial SSTA of El Niño, further causing a westward movement of the warm center of El Niño along the equator, contributing to a more frequent occurrence of Central-Pacific-El Niño (Li et al 2017).The increase in extreme cold and warm events (ENSO diversity) in the central equatorial Pacific has been attributed to a positive cold tongue pattern in the context of recent global warming (Jiang and Zhu 2018).
The decadal variability of SSTAs in the Pacific Ocean and changes in weather systems are likely closely related to lower-frequency TCs in recent decades (Huangfu et al 2017, An 2018, Kim et al 2020).The CTM is likely a reflection of global warming.However, very few studies have investigated the possible impact of CTM on TCs.In the present study, we explored the potential impact of the CTM on TCs over the WNP from the perspective of anomalous oceanic thermal conditions and atmospheric circulation, and examined the responses of TC activity to SSTA under recent global warming.

Data
The TC Best Track data set was obtained from the International Best Track Archive for Climate Management (IBTrACS v04), with a temporal resolution of 6 h (Knapp et al 2010).Monthly atmospheric reanalysis data were obtained from ERA5, with a horizontal resolution of 0.25 • × 0.25

Method
The location of TC genesis is defined when the central wind speed of a tropical storm is greater than 17.2 m s −1 .Since routine satellite observations of TCs began in 1975 (Kossin et al 2007), and the active season of TC spans from May to November (Wang et al 2010), the present study mainly focused on TC activity over the WNP in the May-November period during 1975-2021.
The structure of the Hadley cell can be measured using the meridional mass stream function, which is obtained by integrating the latitudinally averaged meridional wind from a given isobaric region to the top of the atmosphere (Oort and Yienger 1996).Similarly, the Walker circulation can be described by the zonal mass stream function (Yu and Zwiers 2010): where u d and v d are the divergence components of the zonal and meridional winds, respectively.x and y are the longitude and latitude, respectively.p is the pressure; g is the acceleration due to gravity; and a is the radius of the Earth.We calculated oceanic zonal heat transport at a given longitude φ i and oceanic meridional heat transport at a given latitude Ø i , respectively (Liu et al 2020): with ρ 0 the density of seawater (1027 kg•m −3 ), c po the specific heat capacity of seawater (3987 ,and z 0 and z b stand for sea surface and the depth to the bottom (270 m here), respectively.

Linear trend of CTM during 1975-2021
El Niño Southern Oscillation (ENSO) was the leading EOF mode in the tropical Pacific SSTA, accounting for 46.3% of the total variance.When ENSO variability was removed, the tropical Pacific SST exhibited a La Niña-like CTM mode.The CTM mode accounts for 19.31% of the total variance (figure 1

Statistical analysis of TCs between P1 and P2
Since 1998, the CTM has changed from a negative to a positive phase.This may result in changes in atmospheric and oceanic conditions as well as TCs over the WNP.The decadal variability of TC frequency exhibited a seesaw pattern around 1998, and the mean frequency of TCs decreased significantly from 24.1 in P1 to 21.8 in P2 (p < 0.05) (figure 2  Therefore, the influence of the CTM on TC formation in the WNP may be mediated by anomalies in the oceanic thermal conditions or atmospheric circulation.

Modulation of CTM on TC over the WNP
In terms of ocean thermal conditions, the coverage of the 29.5 • C isotherm in P2 extends eastward compared to that in P1.The mean SST over the western Pacific warm pool in P2, that is, approximately 29.5 • C, is higher than that in P1 at approximately 28.5 • C. The SST zonal gradient of the warm pool (dSST = SST(125 Consequently, the frequency of TCs decreases.This suggests that the main reason for the decrease in TC frequency is the adjustment of the atmospheric circulation caused by the change in SST, whereas the impact of local SST variations is quite minimal.This conclusion is also consistent with the findings of Chen and Huang ( 2006) and Fedorov et al (2015).
In the WNP region, the western and southern boundaries are ocean heat input boundaries, and the eastern and northern boundaries are ocean heat output boundaries.In P1, the zonal heat transportation is 1.75 × 10 8 W (1.8 minus 0.05), the meridional heat output in the region is 2.49 × 10 8 W (2.34 minus −0.15), and the total heat transportation is 4.24 × 10 8 W. In P2, the zonal heat transportation is 1.55 × 10 8 W (1.71 minus 0.16).The meridional heat transportation is 2.32 × 10 8 W (2.46 minus 0.14), and the total heat transportation is 3.87 × 10 8 W (figure 3(c)).The reduced meridional and zonal heat exports in P2 led to a reduced total heat export and higher sea surface and subsurface temperatures.The heat transportation at the eastern boundary of 165 • E decreases from 0.2 × 10 8 W in P1 to 0.06 × 10 8 W in P2.There was heat accumulation in the warm pool, resulting in higher mean temperatures in the western Pacific warm pool in P2 compared to P1.The SST zonal gradient of the warm pool increased in P2 compared to P1 (figures 3(a) and (b)).
The regression equation between TC frequency and zonal and meridional heat transportation can be expressed as follows: TC frequency = 0.66 * zonal heat transportation + 2.49 * meridional heat transportation + 15.2.
The regression coefficients passed the 95% hypothesis test and the proportion of variance in the dependent variable that can be explained by the independent variable is 15%.From the formula, we can see that TC frequency over the WNP was positively correlated with zonal and meridional heat transportations.Given the decrease in the meridional and zonal heat transportations in P2, the TC generation frequency decreased.Ocean heat transfer to the atmosphere is favorable for TC generation and development (Wada and Usui 2007, Neerja and Ali 2014).The reduction in TC frequency in P2 may decrease ocean heat loss, leading to surface and subsurface warming in the western Pacific, and warming of the warm pool.Compared with P1, the increased western boundary heat input and the increased northern boundary heat transport in P2 lead to increased ocean heat in the western and northern WNP, which favors more TC generation in the western and northern WNP.
Owing to global warming, a large amount of heat is stored in the oceans, resulting in a change in the thermal structure.This, in turn, affects atmospheric circulation (Wang et al 2020).The Walker circulation was enhanced and shifted westward in P2 (figure 1

Interannual variability of TC frequency
Compared with P1, the interannual fluctuation in TC frequency was larger in P2 (figure 3(a)).The variance increased significantly from 0.17 in P1 to 0.18 in P2 (p < 0.05).Under different phases of the CTM, ENSO variability may have different effects on the interannual variability of TC frequency (figure 4).Before the 1990s, the tropical Pacific was dominated by Eastern-Pacific El Niño, after 1990s, the frequency of Central-Pacific El Niño events has increased (Huang et al 2020).In P1, the frequency of TCs significantly correlated with the Nino3 index and Nino1 + 2 index, with a correlation coefficient above 0.8 (table S1).In P2, the correlation coefficient of the TC frequency with the Nino3 index decreased to 0.65 and had no correlation with the Nino1 + 2 index.The Nino4 index and Nino3.4 index had no significant correlation with the TC frequency in either P1 or P2 (table S1).This suggests that the commonly used ENSO index became less indicative of TC frequency in P2.We used factor variance analysis to analyze the effect of zonal heat export (110 • E-180 • ) on TC frequency and found that zonal heat export had a significant effect on TC frequency in P2 (p < 0.1).This suggests that the inter-annual variability in TC frequency in P2 may be related to WNP zonal heat transportation.

Summary and discussion
We explored the potential impact of the CTM on TCs in the WNP from the perspective of oceanic thermal conditions and atmospheric circulation during 1975-2021.The positive phase of the CTM was characterized by an enhanced warming pool in the WNP, resembling a La Niña pattern.In contrast to the negative phase of the CTM during 1975-1997 (P1), the observed frequency of TCs in the WNP was significantly lower.The generation position of TCs shifted westward and northward during the positive phase of the CTM during 1998-2021 (P2).Driven by a La Niña-like SST trend distribution in the tropical Pacific, Walker circulation was enhanced, and the WPSH expanded westward.The Hadley cell was anomalously weakened in the upper layers and strengthened in the lower layers, with a concomitant poleward extension of the Hadley circulation.The westward movement of the Walker circulation and the northward expansion of the regional Hadley cell were the dominant factors responsible for the westward and northward movement in TC generation.The frequency of TCs was positively correlated with zonal and meridional heat transportation.The reduced meridional and zonal heat export in P2 suppressed TC generation frequency.Compared with P1, the western boundary heat input increased, and the northern boundary heat transportation increased in P2.This resulted in a westward and northward movement of the TC generation.Meanwhile, the reduction in TC frequency for P2 may have decreased ocean heat loss, leading to surface and subsurface warming in the western Pacific and warming of the warm pool (figure 5).The interannual variability of the TC frequency in P2 increased.In P1, the TC frequency was closely related to ENSO, whereas that in P2 was influenced by the zonal heat transportation.
Liu and Chen (2018) showed that, since the early 1990s, the duration of central Pacific El Niño has increased, and the range of SSTA has increased.Therefore, it is more capable of influencing the frequency of summer TCs in the northwestern Pacific.Bengtsson et al (2007) (a)).The seesaw pattern of the SSTAs is consistent with that of a previous study (L'Heureux et al 2013).PC2 shows the long-term trend of the CTM with global warming.The transition phase occurred around 1998 (figure1(b)).With reference to Jiang's findings, the period 1950-2021 is divided into three time periods, with the periods 1950-1974 as before CTM formation, 1975-1997(P1) as the CTM negative phase, and 1998-2021(P2) as the CTM positive phase, respectively (Jiang and Zhu 2020).The thermodynamic structure in the upper Pacific was significantly different between P1 and P2 (figures 1(c) and (d)).The SST in the WNP warms significantly (p < 0.1), with a warming trend in subsurface sea temperatures above 270 m.The maximum of the warming amplitude (+0.74 • C) occurs around 120 m.SST cooling occurred in the east-central equatorial Pacific, where significant cooling was located in the 120 • W-90 • W region (p < 0.1).The subsurface cold tongue in the eastern Pacific shrinks, and the 50-150 m subsurface layer becomes cooler, with

Figure 1 .
Figure 1.(a) Cold tongue mode (CTM) of the tropical Pacific seas surface temperature anomalies during 1950-2021; (b) corresponding PC of the CTM; (c) Spatial distribution of SST differences between P2 and P1; (d) longitude-depth profiles of sea temperature differences over 30 • S-30 • N between P2 and P1; (e) difference in the zonal mass stream function (shaded; 10 10 kg•s −1 ) and the averaged divergent zonal wind (shown in vector) between P1 and P2 over 5 • S-5 • N. The mean state of the zonal mass stream function (blue line; 10 10 kg•s −1 ); (f) differences in meridional mass stream function (shaded; 10 10 kg•s −1 , zero line in black) and averaged divergent meridional wind (expressed as vectors) between P1 and P2 along the Western Pacific Ocean (100 • E-180 • ).The shaded areas show values that are statistically significant at the 90% confidence level.
(a)).The location of TC generation showed an increase in the proportion of TCs generated west of 130 • E and north of 20 • N in P2.The number of TCs increased in the South China Sea, whereas the number of TCs decreased in the western Pacific warm pool (125• E-165 • E, 0 • -16 • N) (figure 2(f)).Statistically,the frequency of TCs over 100 • E-130 • E increased from 360 (65%) in P1 to 379 (73%) in P2.Meanwhile, the frequency of TCs over 130 • E-180 • decreased from 194 (35%) in P1 to 143 (27%) in P2.The frequency of TCs south of 20 • N decreased from 429 (77%) in P1 to 366 (70%) in P2, whereas the frequency of TCs north of 20 • N increased from 125 (23%) in P1 to 156 (30%) in P2.The average longitude of TC generation decreased significantly from 136.8 • E in P1 to 134.3 • E in P2 (p < 0.05, figure 2(b)).The mean latitude of TC generation increased significantly from 16.2 • N in P1 to 17.3 • N in P2 (p < 0.05, figure2(c)).Therefore, the average frequency of TCs in the WNP showed a significant decrease in P2 with the northwestward movement of the TC location.The CTM

Figure 2 .
Figure 2. (a) Frequency of TCs from May to November in 1975-2021.The blue line indicating the average frequency in P1 and P2, respectively.Changes in longitude (b) and latitude (c) during 1975-2021.The blue and black lines indicate mean longitude and latitude before and after 1998, respectively.Changes in the spatial distribution of TCs before (d) and after 1998 (e).(f) Difference in the spatial distribution of TC frequencies for P1 and P2.The shaded areas only show values that are statistically significant at the 90% confidence level.
38 • C in P1 to 0.48 • C in P2 (figures 3(a) and (b)).The SST zonal gradient of the warm pool is significantly negatively correlated (p < 0.05) with TC frequency, with a correlation coefficient of −0.61.During the P2 phase, accompanied by the warming of the western Pacific warm pool, the increase in the SST zonal gradient of the warm pool, leading to the westward shift of the Walker circulation (figure 1(e)) and the northward expansion of the Hadley circulation (figure 1(f)).

Figure 3 .
Figure 3. Spatial distribution of SST (shaded, in • C) and difference fields for P1 (a) and P2 (b) from May to November.The black dashed line representing the 28.5 • C and 29.5 • C isotherms.(c) Ocean heat transportations (10 8 W) at each boundary at 0-270 m below the sea surface of the WNP for P1 (in red) and P2 (in blue), with black boxes indicating the study area.Distribution of geopotential height at 500 hPa and differential field (f) for P1 (d) and P2 (e) during May-November.Contours (dashed lines, unit: gpm) in the WNP.
(e)).During P2, there was an anomalous upward flow between 100 • E and 150 • E and an anomalous downward flow east of 150 • E. In P1 (figure 3(d)),the West Pacific Subtropical High (WPSH) was located in the region of 19 • N-28 • N, 141 • E-173 • E, whereas in P2 (figure 3(e)), the WPSH extended northwestward and covered a larger area (14 • N-30 • N, 131 • E-184 • E).The geopotential height at 500 hPa showed a positive anomaly in the western Pacific (figure 3(f)).The westward movement of the Walker circulation and the westward extension of the WPSH led to a westward movement in TCs, enhanced TC frequency in the South China Sea, and suppressed TCs in the WNP (figure 2(f)).The Hadley cells weakened in the upper layers (above 300 hPa) and strengthened in the lower layers with a concomitant poleward extension (figure 1(f)).The weakening of the upper-level anomalies in the ascending branch of the Hadley cell may suppress the occurrence of low-latitude TCs in the WNP (Zhou and Cui 2008, Studholme and Gulev 2018).The northward expansion of the regional Hadley cells may have favored the northward movement of the TCs in P2 (figure 2(f)).

Figure 4 .
Figure 4. Scatter plot of the effect of tropical Pacific sea temperature patterns on the frequency of TCs.The gray shading masks non-ENSO years, the red shading represents positive CTM years, and the color and size of the scatter represent the frequency of TCs.
suggested that the decrease in TC frequency associated with global warming is related to the weakening of the mass flux associated with tropical deep convection.Hong et al (2016) suggest that the abrupt change in the Pacific climate pattern in the mid-to-late 1990s led to a reduction in the frequency of TCs in the northwestern Pacific in summer and autumn and a northwesterly movement in the mean location of TC generation.However, we only considered the decrease in the frequency of TCs in the WNP after 1998 from the perspective of the decrease in meridional and zonal heat export.In this context, the detailed mechanisms require further confirmation via dynamic model simulations in the future.

Figure 5 .
Figure 5. Schematic map of oceanic and atmospheric contrast in the Pacific Ocean before and after 1998.Orange arrows indicate the Walker circulation.Gray arrows indicate the Hadley cell.Black symbols indicate TCs.Blue line represents the confluence of warm and cold water.Pink areas represent warm pools, and yellow areas indicate WPSH.

sea temperature and zonal velocity (Zuo et al 2018). The monthly reanalysis data for the present study are available from 1940 to 2021.
• , including 500 hPa potential height and pressure (Hersbach et al 2020).