Upper Ocean Response to Two Sequential Tropical Cyclones over the Northwestern Pacific Ocean

The upper ocean thermodynamic and biological responses to two sequential tropical cyclones (TCs) over the Northwestern Pacific Ocean were investigated using multi-satellite datasets, in situ observations and numerical model outputs. During Kalmaegi and Fung-Wong, three distinct cold patches were observed at sea surface. The locations of these cold patches are highly correlated with relatively shallower depth of the 26 ◦C isotherm and mixed layer depth (MLD) and lower upper ocean heat content. The enhancement of surface chlorophyll a (chl-a) concentration was detected in these three regions as well, mainly due to the TC-induced mixing and upwelling as well as the terrestrial runoff. Moreover, the pre-existing ocean cyclonic eddy (CE) has been found to significantly modulate the magnitude of surface cooling and chl-a increase. With the deepening of the MLD on the right side of TCs, the temperature of the mixed layer decreased and the salinity increased. The sequential TCs had superimposed effects on the upper ocean response. The possible causes of sudden track change in sequential TCs scenario were also explored. Both atmospheric and oceanic conditions play noticeable roles in abrupt northward turning of the subsequent TC Fung-Wong.


Introduction
Tropical cyclones (TCs), which are among the most devastating natural disasters, pose significant threat on the lives and property of global coastal regions [1]. During the passage of a TC, the strong wind stress could typically produce intense oceanic entrainment and upwelling [2][3][4], which leads to physical and biological responses of the upper ocean. Sea surface temperature (SST) decrease, the most striking phenomenon, is usually observed in the range of 1-9 • C with stronger cooling to the right side of a TC's track [5][6][7][8][9]. The sea surface cooling has a negative feedback to a TC by suppressing the TC development in turn [10]. Furthermore, the SST decrease would have an impact on the subsequent TCs if they pass by these cold wakes coincidentally [11]. The distribution of phytoplankton biomass, as is measured by chlorophyll a (chl-a) concentration, is also influenced by TCs. Previous studies have found that the surface chl-a concentration would be largely increased after TCs, with estimated maximum contribution of 20-30% for the annual new production in the South China Sea (SCS) [12,13]. It is well recognized that the magnitude and spatial distribution of SST decrease and chl-a enhancement mainly depend on the intensity and translation speed of TCs and the background ocean environments, such as the mixed layer depth (MLD) and mesoscale eddies, as well as topography [2,[13][14][15][16]. Owing to the Fung-Wong (right panel). The blue and red dashed lines indicate tracks of Kalmaegi and Fung-Wong, respectively. The symbol "*" in (a) denotes the center location of two TCs at 00:00 UTC on the corresponding date. The black triangles with symbols from "B1" to "B5" indicate five array stations. The magenta pentagrams with symbols "A1" and "A2" indicate the locations of the two Argo floats.

Satellite Observations
The daily SST product used in this study is Microwave and Infrared OI SST (MW-IR OISST) from the Remote Sensing Systems (http://www.remss.com/) with a spatial resolution of 9 km × 9 km.
It combines the through-cloud capability of the microwave OI SST product [35] with the high spatial resolution of the infrared SST by using Optimum Interpolation method [36].
The daily SSHA dataset, provided by Archiving, Validation and Interpretation of Satellite Data in Oceanography (AVISO) has a spatial resolution of 0.25 • × 0.25 • . The altimeter products are produced and distributed by the Copernicus Marine and Environment Monitoring Service (CMEMS, http://marine.Copernicus.eu/). Closed contours of negative SSHA are normally corresponding to mesoscale cyclonic eddies (CEs) with cold water in the core region [8].
Because the daily chl-a concentration dataset includes plenty of data gaps caused by severe cloud contamination, we use the 8-day averaged data to study the biological response of the upper ocean to the sequential TCs. Previous studies pointed out that phytoplankton growth, which attributes to chl-a enhancement under the influence of a TC, needs a delay of several days (∼5 days) to reach its peak [21]. The accurate days for chl-a concentration to rise to its maximum remains unclear [22]. The merged 8-day chl-a concentration data, with a spatial resolution of 4 km × 4 km, produced and distributed by the GlobColour Project (http://hermes.acri.fr/) are satisfied with the basic research requirements [22]. The GlobColour data were developed, validated, and distributed by ACRI-ST, France.
The daily ocean surface winds on a global 0.25 • grid derived from the Blended Sea Winds dataset are acquired from the National Oceanic and Atmospheric Administration's (NOAA's) National Climatic Data Center, via their website (http://http://www.ncdc.noaa.gov/oa/rsad/air-sea/seawinds. html). The gridded wind speeds were created from the blending observations of multiple satellites [37].

In Situ Measurements
Temperature and salinity profiles used to investigate the thermal responses are obtained from Argo observations and a cross-shaped observational array in the SCS deployed by the Second Institute of Oceanography, Ministry of Natural Resources of China in 2014. The Argo profiles are provided by the China Argo Real-time Data Center (http://www.argo.org.cn/) and the quality control has been performed [38]. The locations of Argo floats and observational array are shown in Figure 1a. There are two Argo floats located in the area influenced by both Kalmaegi and Fung-Wong. The observational array consists of five moored buoys and four subsurface moorings organized into five stations. Considering the validity of the data and the distance between the stations and two TCs, the data at Stations 4 and 5 are analyzed in this study.

Ocean Numerical Model Outputs
The daily ocean analysis data with a spatial resolution of 0.08 • × 0.08 • are obtained from the global HYbrid Coordinate Ocean Model (HYCOM) + Navy Coupled Ocean Data Assimilation (NCODA) based ocean prediction system (hereafter called HYCOM, https://www.hycom.org/) [39]. As shown in Figure 2, HYCOM simulations of SST, SST difference (SSTD) during and before Kalmaegi passed by and SSHA are in good agreement with satellite observations. It is clear that the SSHA from HYCOM and AVISO both reveal the same CE region. Therefore, the data can be used for further analysis.

Atmospheric Reanalysis Data
The atmospheric fields are obtained from the National Centers for Environmental Prediction-Department of Energy (NCEP-DOE) AMIP-II reanalysis 2 data, which is from the National Oceanic and Atmospheric Administration (NOAA) Earth System Research Laboratory's Physical Sciences Division (https://www.esrl.noaa.gov/psd/). The reanalysis data have a spatial resolution of 2.5 • × 2.5 • on a global grid and 17 pressure levels from 10 to 1000 hPa [40]. on a global grid and 17 pressure levels from 10 to 1000 hPa [40].

149
In this study, SST on September 9 is considered to be the pre-TCs condition. The SSTD, which 150 denotes the SST variation during and before TCs pass by, can be calculated from satellite 151 observations or HYCOM data by subtracting SST during the TCs and that on September 9.

152
The MLD, defined as the depth where the temperature decreases 0.5 °C compared to that at 153 surface [41], or that was based on density variation according to Equation (1) of Kara et al. (2000) 154 [41], is calculated from temperature sections of Argo profiles, observational array and HYCOM data.

155
The temporal variations of MLDs are similar when we use temperature or density criterion,

159
The UOHC is calculated using HYCOM output of ocean temperature profiles T(z):

Methods
In this study, SST on September 9 is considered to be the pre-TCs condition. The SSTD, which denotes the SST variation during and before TCs pass by, can be calculated from satellite observations or HYCOM data by subtracting SST during the TCs and that on September 9.
The MLD, defined as the depth where the temperature decreases 0.5 • C compared to that at surface [41], or that was based on density variation according to Equation (1) of Kara et al. (2000) [41], is calculated from temperature sections of Argo profiles, observational array and HYCOM data. The temporal variations of MLDs are similar when we use temperature or density criterion, although the absolute values of the later are a little smaller. Here, the results of MLD calculated by temperature criterion are adopted. Mixed layer temperature (MLT) and mixed layer salinity (MLS) are taken as the average values in the mixed layer.
The UOHC is calculated using HYCOM output of ocean temperature profiles T(z): where ρ is the seawater density (1024 kg/m 3 ); C p is the heat capacity (4186 J/kg • C). D26 is the depth of the 26 • C isotherm, which can characterize the relative warm or cold of the upper ocean. The potential upwelling speed w due to TC winds is calculated using the classical Ekman pumping velocity (EPV) formula [4]: where f is the Coriolis parameter. The wind stress → τ can be obtained by the bulk formula [42]: where ρ a is the air density; U 10 and → u 10 are the wind speed and wind vector 10 meters above sea surface, respectively; C d is the drag coefficient and is calculated using the formula proposed by Oey et al. (2006) [43], which fits the formula at low-to-moderate wind speeds [44] and the results at high wind speeds [45], C d × 1000 = 1.2 U 10 < 11 ms −1 = 0.49 + 0.065U 10 11 < U 10 < 19 ms −1 = 1.364 + 0.0234U 10 − 0.00023158U 2 10 19 < U 10 < 100 ms −1 (4)

Sea Surface Cooling
As shown in Figures 3a and 4a,c,e, the background ocean environments on September 9 provided favorable conditions for TC generation, with high SST, relatively deep D26, MLD and high UOHC. The SST in most part of study area was higher than 30 • C. During the passages of both Kalmaegi and Fung-Wong, there was a cool trail along their tracks with rightward bias (Figure 3b-f). In particular, three distinct cold patches were detected, which were marked with black, gray and red box and correspondingly labeled as R1, R2, R3 in Figure 3, respectively.
In region R1, the maximum decrease of SST located to the right side of Kalmaegi's track was 3.18 • C on September 14 ( Figure 3b). The ocean surface cooled again 5 days later when Fung-Wong passed through the same area (Figure 3d), with a temperature decrease of 2.36 • C. Overall, during the passages of the sequential TCs, SST dropped by 3.28 • C. Kalmaegi and Fung-Wong were both strong with MSW ranging between 28.3~36.0 m s −1 and 15.4~18.0 m s −1 , respectively. Therefore, the induced strong mixing effects were able to bring colder water from deep to surface. Furthermore, this region was characterized by shallower MLD, D26 and lower UOHC before the arrival of TCs (Figure 4), which is beneficial to the cooling of sea surface. Interestingly, as Figure 5 shows, whether during the passage of Kalmaegi or Fung-Wong, this cold patch was highly consistent with the location of a pre-existing CE ( Figure 5). Negative SSHA inside the CE indicates that the MLD was relatively shallower than the background ocean. It means the thermodynamic structure inside the CE was relatively unstable, which makes it possible to uplift the cold water below easily. Therefore, the pre-existing CE is an essential factor contributing to the stronger surface cooling, which has also been noticed in previous studies [7,8].
As can be seen in Figure 3c, the maximum SST drop in region R2 reached 3.73 • C after Kalmaegi passed. The ocean surface then gradually recovered until September 19 when Fung-Wong affected this region. The surface cooled down again by 1.93 • C, but much weaker than the first cooling. The magnitude of temperature decrease in this region after sequential TCs was larger than that in the other two regions. The possible reasons are: (1) Shallower MLD, D26 and lower UOHC were all favorable for the surface cooling. (2) The intensity of Kalmaegi increased continuously to its peak MSW of 41 m s −1 when passing by Region R2. While during Fung-Wong, its slow translation speed (from 3.1 to 4.8 m s −1 ) could induce strong upwelling. However, because this region was located to the left of Fung-Wong, the intensity of which was much smaller than Kalmaegi, the weaker mixing effect resulted in a weaker second cooling.
For Region R3, the strongest surface cooling caused by Kalmaegi was 2.54 • C on September 15. During Fung-Wong on September 20, SST was reduced again by 2.38 • C after its recovery from the influence of Kalmaegi. The magnitude of SST decrease induced by the sequential TCs reached up to 2.92 • C in this region compared to pre-TC conditions, which was the smallest among the three regions owing to the relatively unfavorable ocean conditions of MLD, D26 and UOHC. Meanwhile, the translation speed of Kalmaegi was the fastest when passing through this region. However, the MSW of Fung-Wong over the same region ranged from 18.0 to 25.7 m s −1 and the translation speed was

Subsurface Response
In order to explore subsurface thermodynamic changes induced by Kalmaegi and Fung-Wong, temperature and salinity data observed by two Argo floats (A1 and A2 in Figure 1a) and two stations (B4 and B5 in Figure 1a) of the observational array were analyzed here. Float A1 was close to the center of Kalmaegi while A2 was on the left side of its track. They were both located on the left of Fung-Wong. Neither float A1 nor A2 has moved a long distance from September 11 to 23. The variations of temperature and salinity profiles from two Argo floats are shown in Figure 6, from which the trends in MLD, MLT and MLS were calculated and listed in Table 1.
One can see from Table 1, the MLD measured by float A1 remained basically unchanged during the transit of Kalmaegi, probably because A1 was near the TC center. After Fung-Wong, the mixed layer deepened and the MLD increased to 47.5 m on September 23. The MLT dropped by 0.96 • C during Kalmaegi and the cooling was strengthened with a further temperature decrease of 0.40 • C after Fung-Wong. Although the MLS increased by~0.15 as Kalmaegi passed by, it is notable that from September 15 to 19, it was reduced by 0.22 due to the influence of Fung-Wong which was on the right of float A1.    -c show the changes of temperature and salinity in the upper ocean as well as that of the mixed layer before and after the sequential TCs at Station B4, which was located to the right of Kalmaegi but left of Fung-Wong. It can be seen that the increase in MLD and decrease in MLT influenced by Kalmaegi were much larger than that induced by Fung-Wong, whereas the MLS increased during Kalmaegi but then decreased during Fung-Wong. The differences in the variations of MLD, MLT and MLS caused by Kalmaegi and Fung-Wong were mainly due to the large difference in the MSW of the two TCs. The MSW of Kalmaegi ranged from 33 to 36 m s −1 when it passed over this region, while that of Fung-Wong was only 18 to 21 m s −1 . In addition, the turning of wind stress to the right side of TC's track may resonate with the near-inertial oscillations generated by a TC, which resulted in stronger vertical mixing [3,19]. Figure 7a,b shows that, the subsurface temperature at depth of~50 to 100 m increased obviously, while the salinity above the depth of~50 m increased first and then decreased after the two TCs. Compared with Figure 3c, the observations at Stations B4 and B5 reproduced the process of SST change measured from satellite, captured the two cooling events and the difference in cooling intensity as well.
Station B5 was located to the left side of both TCs' tracks. As can be seen from Figure 7c,d and Figure 8d-f, compared with Station B4, the trends of MLD and MLT were similar but with weaker amplitudes, especially during the passage of Kalmaegi. There was no distinct change in MLS, probably because the salinity variation is constrained by many factors, such as precipitation and evaporation. It is worth mentioning that there was an obvious signal around September 16 that all isotherms within 200 m of the upper layer sank sharply (Figure 7c,d). This is possibly caused by the resonation between Kalmaegi-generated near-inertial oscillation and the internal wave, which could produce more energy and lead to stronger vertical mixing. the resonation between Kalmaegi-generated near-inertial oscillation and the internal wave, which the entrainment and upwelling induced by the TC. Meanwhile, the MLS increased on the right side 282 of their tracks but decreased on the left side. The difference is mainly due to the asymmetry of   region (region R1 in Figure 9) was a little higher than that in other areas of the WNP but fell within 296 the predominant value of ≤ 0.1 mg m -3 in typical summer in SCS [12]. Not surprisingly, three 297 typical regions with obvious surface cooling are also noted with pronounced growth of chl-a after 298 sequential TCs. As mentioned in Section 3.1.1, the translation speed and intensity of the TCs were 299 both conducive to chl-a rise when they passed over these three regions. However, the extent of 300 enhancement varied from region to region due to the specific mechanisms involved. Consequently, the MLD deepened and MLT dropped after Kalmaegi or Fung-Wong, owing to the entrainment and upwelling induced by the TC. Meanwhile, the MLS increased on the right side of their tracks but decreased on the left side. The difference is mainly due to the asymmetry of precipitation (the figure is not shown) brought by the TC [46]. Figure 9 depicts the temporal evolution of surface chl-a concentration before, during and after the passages of Kalmaegi and Fung-Wong. Before TCs, the chl-a concentration in the SCS and CE region (region R1 in Figure 9) was a little higher than that in other areas of the WNP but fell within the predominant value of ≤ 0.1 mg m −3 in typical summer in SCS [12]. Not surprisingly, three typical regions with obvious surface cooling are also noted with pronounced growth of chl-a after sequential TCs. As mentioned in Section 3.1.1, the translation speed and intensity of the TCs were both conducive to chl-a rise when they passed over these three regions. However, the extent of enhancement varied from region to region due to the specific mechanisms involved.   Figure 9 depicts the temporal evolution of surface chl-a concentration before, during and after 294 the passages of Kalmaegi and Fung-Wong. Before TCs, the chl-a concentration in the SCS and CE 295 region (region R1 in Figure 9) was a little higher than that in other areas of the WNP but fell within 296 the predominant value of ≤ 0.1 mg m -3 in typical summer in SCS [12]. Not surprisingly, three 297 typical regions with obvious surface cooling are also noted with pronounced growth of chl-a after 298 sequential TCs. As mentioned in Section 3.1.1, the translation speed and intensity of the TCs were 299 both conducive to chl-a rise when they passed over these three regions. However, the extent of 300 enhancement varied from region to region due to the specific mechanisms involved. Inside the CE (region R1), the maximum increase (0.21 mg m −3 ) in chl-a concentration occurred near the center of the eddy. Owing to the special thermodynamic structure of the CE, the subsurface water with rich nutrients inside the eddy would be more likely to be brought to the euphotic zone by TC-enhanced vertical mixing. Therefore, the pre-existing CE could not only modulate the thermodynamic response of the upper ocean, but also cause changes in biological processes.

Biological Response
As for the near-shore area (region R2), the average surface chl-a concentration increased from 0.38 to 0.53 mg m −3 . At certain locations (e.g., 118.4 • E, 20 • N) of the coastal area, the chl-a concentration reached as high as 20 mg m −3 . Except for the influence of translation speed and intensity of TCs, two other mechanisms may be responsible for this. (1) The depth of water in areas tens of kilometers offshore is usually only tens of meters [47], which make it easier to bring the nutrient from the lower layer to the surface. (2) The heavy rainfall caused by TCs can trigger the enrichment of terrigenous nutrients in near-shore waters.
The peak value of chl-a rise in Luzon Strait (region R3) reached up to 0.58 mg m −3 . The chl-a concentration even increased from 2.01 to 5.16 mg m −3 at the land-sea junction, demonstrating a prominent enhancement. It is noteworthy that the intensity of Fung-Wong had been continuously strengthened as it passed through this region during September 19 to 20. In addition to the mixing effect induced by sequential TCs, it can be observed from Figure 10 that both Kalmaegi and Fung-Wong caused strong upwelling with maximum EPV larger than 1.5×10 4 m s −1 , which is able to transport subsurface nutrients upward to the euphotic zone as well. Compared with the period from September 14 to 21, the response of chl-a during September 22 to 29 was more remarkable, probably related to the lag response of chl-a and the superposition effect of sequential TCs [11,21,48].

Influence of Sequential TCs on the CE
As mentioned in Sections 3.1 and 3.2, the pre-existing CE has a significant impact on both the thermodynamic and biological responses of the upper ocean to TCs. Conversely, TCs could also influence the structure and intensity of the eddy. In this study, it is of interest that the CE region interacted with two TCs successively. Before Kalmaegi, the SST inside the CE was about 29.5 • C, not obviously lower than that of background waters (Figure 11a). On September 13 and 14 when Kalmaegi passed over the eddy, there was a significant decrease of 3.18 • C in SST (Figure 11b). During this period, the MSW ranged from 28.3 to 36.0 m s −1 . The translation speed was higher than 6.0 m s −1 , which is defined as a "fast-moving" TC [3]. The conditions were favorable for the formation of strong vertical mixing, especially to the right side of the TC's track. After that, the water temperature gradually recovered but the process was then disturbed when Fung-Wong passed by. On September 19 when Fung-Wong left this area, a distinctly extreme cold patch was observed again in the CE region with a much stronger cooling of up to 3.28 • C compared to that on September 9 (Figure 11c). By September 30, more than 10 days, this cold patch was still evident (Figure 11d).
intensity of TCs, two other mechanisms may be responsible for this. (1) The depth of water in areas enrichment of terrigenous nutrients in near-shore waters.

318
The peak value of chl-a rise in Luzon Strait (region R3) reached up to 0.58 mg m -3 . The chl-a 319 concentration even increased from 2.01 to 5.16 mg m -3 at the land-sea junction, demonstrating a 320 prominent enhancement. It is noteworthy that the intensity of Fung-Wong had been continuously 321 strengthened as it passed through this region during September 19 to 20. In addition to the mixing 322 effect induced by sequential TCs, it can be observed from Figure 10 that both Kalmaegi and 323 Fung-Wong caused strong upwelling with maximum EPV larger than 1.5×10 4 m s -1 , which is able to 324 transport subsurface nutrients upward to the euphotic zone as well. Compared with the period 325 from September 14 to 21, the response of chl-a during September 22 to 29 was more remarkable, 326 probably related to the lag response of chl-a and the superposition effect of sequential TCs 327 [11,21,48].  Kalmaegi passed over the eddy, there was a significant decrease of 3.18 °C in SST (Figure 11(b)).

340
During this period, the MSW ranged from 28.3 to 36.0 m s −1 . The translation speed was higher than 341 6.0 m s −1 , which is defined as a "fast-moving" TC [3]. The conditions were favorable for the

357
These isotherms then rose slowly until the following TC Fung-Wong interacted with the eddy.

358
Under the influence of the sequential TCs, the isotherms tended to descend as a whole, which 359 means the vertical mixing effect plays a dominant role in the surface cooling of the CE region. Figure 11. Evolution of satellite observed SST (Top, (a-d)) and negative SSHA (Bottom, (e-h)) in the CE region before, during and after sequential TCs. The expression of TC tracks is the same as Figure 2. The red and green dots denote the center locations of Kalmaegi and Fung-Wong at 00:00 UTC, respectively. Figure 12 shows the variation of upper layer temperature at the eddy center. It can be seen that, the 29 • C isotherm, which was initially at a depth of~50 m, rose rapidly to the surface on September 14 and outcropped after that until September 26. From September 14 to 15, all isotherms exhibited a downward trend. For example, the 27 • C isotherm deepened from the depth of~60 m to~80 m. These isotherms then rose slowly until the following TC Fung-Wong interacted with the eddy. Under the influence of the sequential TCs, the isotherms tended to descend as a whole, which means the vertical mixing effect plays a dominant role in the surface cooling of the CE region. The intensity of a CE can be depicted by negative values of SSHA [49]. As shown in Figure 11,

371
In additional to the significant modulation on the ocean environment, sequential TCs usually 372 result in an unusual track, as indicated by various studies [50][51][52]. As shown in Figure 1 Kurihara et al. (1993Kurihara et al. ( , 1995 [53,54], can objectively identify an asymmetric TC 381 component to extract a smooth large-scale environmental field [55,56]. Using this TC-removal 382 method, the background wind field, which represents the steering flow in the middle troposphere, 383 is extracted to trace the TC's movement.

384
Variations of environment geopotential height and wind in the middle troposphere from 385 September 18 to 21 are shown in Figure 13. On September 18, the western North Pacific subtropical 386 high (WNPSH) was significantly strengthened with the 5880 gpm contour penetrating southward 387 to approximately 10°N and dominated the entire tropical Northwestern Pacific (Figure 13(a)).

388
Easterly prevailed to the south of the WNPSH and steered the TC Fung-Wong to the west. West of 389 Fung-Wong, a significant ridge and a trough appeared over the SCS and the Philippine Island 390 (Figure 13(a)). Such ridge and trough were related to the Rossby energy dispersion radiating from 391 the prior TC Kalmaegi which had dissipated over the Indochina Peninsula [57,58]. Affected by the 392 energy dispersion, the 5880 gpm contour to the south of the WNPSH wriggled sharply which The intensity of a CE can be depicted by negative values of SSHA [49]. As shown in Figure 11, the SSHA at eddy center dropped by~10 cm after the passage of Kalmaegi. Subsequently, Fung-Wong passed by and the SSHA dropped to the lowest value (~15 cm) by September 20. The northwestern part of the pre-existing CE was visibly strengthened by Fung-Wong and moved westward slightly. It is noted that the CE was intensified by both TCs although with varying degrees of impact, mainly because the MSW of Kalmaegi was twice that of Fung-Wong during this period.

Possible Causes of Sudden Track Change in Sequential TCs Scenario
In additional to the significant modulation on the ocean environment, sequential TCs usually result in an unusual track, as indicated by various studies [50][51][52]. As shown in Figure 1a, Fung-Wong shifted westward to the Luzon Strait following the prior TC Kalmaegi before September 20. However, Fung-Wong suddenly turned northward passing the eastern coast of the Taiwan Island. Is the sudden change of Fung-Wong's track related to the prior TC Kalmaegi? This study tries to answer this question by analyzing both atmospheric and oceanic environments. First of all, tracks of TCs are mostly driven by large-scale steering flow in the middle troposphere. However, the TC-related circulation obscures the identification of the steering flow as the TCs are always embedded in complex environmental circulations. A TC-removal method, which was developed by Kurihara et al. (1993Kurihara et al. ( , 1995 [53,54], can objectively identify an asymmetric TC component to extract a smooth large-scale environmental field [55,56]. Using this TC-removal method, the background wind field, which represents the steering flow in the middle troposphere, is extracted to trace the TC's movement. Variations of environment geopotential height and wind in the middle troposphere from September 18 to 21 are shown in Figure 13. On September 18, the western North Pacific subtropical high (WNPSH) was significantly strengthened with the 5880 gpm contour penetrating southward to approximately 10 • N and dominated the entire tropical Northwestern Pacific (Figure 13a). Easterly prevailed to the south of the WNPSH and steered the TC Fung-Wong to the west. West of Fung-Wong, a significant ridge and a trough appeared over the SCS and the Philippine Island (Figure 13a). Such ridge and trough were related to the Rossby energy dispersion radiating from the prior TC Kalmaegi which had dissipated over the Indochina Peninsula [57,58]. Affected by the energy dispersion, the 5880 gpm contour to the south of the WNPSH wriggled sharply which resulted in a breakdown of the WNPSH. Thus, the trade easterly turned to northward since September 19 (Figure 13b-d). Meanwhile, Fung-Wong turned northward following the transition of the environment steering flow. steering flow changed from easterly to westerly during the period from September 18 to 20 (Figure.   This pattern aroused the conceptual model of Carr et al. (1977) [59]. However, the critical differences exist between this event and Carr's model. According to Carr's model, the WNPSH was located to the north of the subsequent TC. Combined with the Rossby wave dispersion-induced anticyclone to the east and equatorward of the prior TC, it will impose an equatorward environmental flow steering the following TC southwest. In this case, the trade easterly to the south of the WNPSH penetrated further equatorward. Modulated by the Rossby wave, the northward steering flow formed and steered the subsequent TC poleward. The differences in the location of the WNPSH are the key for the opposite moving direction of the subsequent TC between Carr's model and this event.
In addition to the favorable atmospheric conditions, oceanic factors are also likely to affect the northward turning of Fung-Wong. Figure 14 displays the distributions of some ocean environmental variables on September 19, i.e., before Fung-Wong turned to the north. It is found that SST in the SCS dropped obviously after Kalmaegi. It was only~28.4 • C and lower than SST in Luzon Strait and the coastal region of Taiwan Island. Simultaneously, lower UOHC, shallower MLD and D26 were also observed in the SCS. It is believed that the maintenance of a mature TC requires heat energy supplement pumping from the warm ocean. Apparently, the ocean conditions in the SCS were not favorable for the maintenance of a TC. Thus, we speculate that Fung-Wong tended to move to areas with warmer ocean conditions other than shifting further westward into the colder SCS.

424
In addition to the favorable atmospheric conditions, oceanic factors are also likely to affect the 425 northward turning of Fung-Wong. Figure 14

Conclusions
The combination of satellite data, numerical model outputs and in situ observations provides a unique opportunity to study the influence of TCs on the upper ocean. In this paper, multi-platform datasets were used to investigate the thermodynamic and biological responses of the upper ocean to two sequential TCs over the NWP, i.e., Kalmaegi and Fung-Wong in 2014. The possible causes of the sudden change of Fung-Wong's track were also explored considering both atmospheric and oceanic environments.
After two TCs passed by, the SST significantly decreased due to the strong oceanic entrainment and upwelling. Three distinct cold patches were observed along their tracks. The magnitude and extent of sea surface cooling are not only highly related to the intensity and translation speed of the TC, but also associated with the ocean background environments (e.g., MLD, D26 and UOHC). Double TCs had superimposed effects on the upper ocean. They deepened the mixing layer and lower the MLT at the same time. The change of MLD or MLT on the right side of the TC's track was stronger than that on the left. The MLS showed a different trend, with a decrease on the left side of the TC but an increase on the right. Meanwhile, it is found that the surface chl-a concentration was enhanced by the TCs. The areas with particularly pronounced chl-a enhancement coincided well with surface cooling regions.
The pre-existing CE plays a notable role in modulating both thermodynamic and biological responses of the upper ocean owing to its unique and relatively unstable thermodynamic structure. On the other hand, it was significantly influenced by the TCs. The CE was enhanced with SST drop of 3.28 • C and SSHA decline of~15 cm.
The present work indicated that the prior TC can affect the track of the subsequent TC by changing the atmospheric and ocean environments. The mature TC Kalmaegi radiated a Rossby wave energy southeastward which wriggled the environmental flow to the south of WNPSH. Consequently, the steering flow changed from westward to northward, leading to the abrupt northward turning of Fung-Wong.
For another aspect, the possible impact of ocean environment on the movement of the TC was also pointed out. With lower UOHC, shallower MLD and D26, the colder SCS could not supply Fung-Wong with enough energy, which can help explain why Fung-Wong did not continue to move westward into the SCS but suddenly turned to the north.
Although under different ocean background conditions, the upper ocean responses to TCs may differ, TC-induced strong upwelling and vertical mixing are the main factors that influence upper ocean conditions. Thus the main findings of this study, including sea surface cooling, chlorophyll enhancement, and subsurface warming, could shed light on one of the general behaviors of the upper ocean responses to TCs. Furthermore, the effect of atmospheric and oceanic environmental changes accompanying the prior TC on the characteristics of the following TC is highlighted, which may provide important insights into the forecasting of the TC's track in sequential TCs scenario. Nevertheless, further studies are needed to improve our understanding of the relationship between the two sequential TCs and their interaction with the ocean by combining the coupled ocean-atmosphere numerical model with current observations.