The category 5 (on the Saffir-Simpson Hurricane Wind Scale) super typhoon Hagibis developed from a tropical depression in the northwestern Pacific on 4 October 2019 ( Figure 1 ). It moved across a CE to the south of Japan with a MSW of ~48.9 m/s and a translation speed of ~7.3 m/s. Ten days later after TC Hagibis passage, TC Neoguri, a category 2 typhoon, passed over the same CE with a MSW of ~20.6 m/s and a translation speed of ~6.8 m/s. Figure 1 shows the location of the CE to the south of Japan, denoted by a patch of low and negative SLA, on October 5 before the two TCs arrived. This CE was quasi-stationary with almost no change in its central position during the influence period of the TCs ( Figure S1 ).

Most of previous studies defined the effects of TCs as the differences of post-TCs and pre-TCs SST or Chl-a concentrations (Siswanto et al., 2009; Zhao et al., 2013; Huang et al., 2016; Li et al., 2021b). However, the background SST in the study area decreased continuously during the two sequential TCs because of the seasonal variation of air temperature. Hence, we chose the differences between daily SST anomalies and pre-TCs SST anomalies to investigate the impacts of Hagibis and Neoguri ( Figure 2 ). Here, SST anomaly on a day was calculated by removing the climatic mean SST state on the corresponding date in a year from the SST, eliminating SST seasonal change. We defined the pre-TCs status as the average SST anomalies during the period of October 3-9 just before the TCs affected the study area. On October 10, two days before Hagibis arriving at the CE, remarkable surface cooling happened under the track of Hagibis with the maximum cooling of 4.9°C ( Figure 2A ). After Hagibis passed over the CE ( Figure 2B ), the average SST anomaly within the CE region dropped to about -1.0°C, but smaller in magnitude than the maximum cooling (-4.9°C) south of the CE. A notable feature of SST cooling was rightward bias as the same as past researches (e.g. Price et al., 1994). The maximum SST anomaly was gradually reduced to -3.8°C in the two days after Hagibis passage while the area-averaged SST cooling in the CE was still -1.0°C. The subsequent TC Neoguri induced a smaller SST anomaly mainly because the intensity of Neoguri was weaker than Hagibis ( Figure 1 ). The area-averaged and maximum values of SST anomaly in the CE were -0.72°C and -2.0°C, respectively. The surface cooling almost disappeared one day after Neoguri passage.

An interesting feature in the Figure 2B is that the SST cooling along Hagibis track were separated by the Kuroshio indicated by ADT derived geostrophic currents in Figure 3A . Temperature profiles in the north of the Kuroshio, the Kuroshio, and the south of the Kuroshio were obtained by Argo floats 2902754, 2903336 and 2902972, respectively, on October 9, 5 and 7 before Hagibis passage ( Figure 3B ). Among these areas, the temperature below the ML in the Kuroshio was the highest with the weakest vertical temperature gradient. This is not beneficial to surface temperature drop due to TC-induced vertical mixing and upwelling, resulting in a weak surface cooling during Hagibis passage ( Figure 2B ). In contrast, the ML was shallow and the temperature gradient below the ML was strong in the area south of the Kuroshio, which was mainly responsible for the prominent TC-induced surface cooling there ( Figure 2B ). In the area north of the Kuroshio, the vertical temperature gradient was also strong but the ML was deeper than that in the south of the Kuroshio ( Figure 3B ) so the TC-induced surface cooling was weaker. Some studies have reported that stronger thermal stratification below shallower ML corresponds to stronger SST cooling induced by a TC (Schade and Emanuel, 1999; Mei et al., 2015).

Figure 4 demonstrates the variation of surface Chl-a concentration during the influence period of the TCs. Before Hagibis passage, the surface Chl-a concentration was very low in open ocean, however, within the CE it reached 0.12 mg/m³, probably attributed to the CE-induced upwelling (Gaube et al., 2014). After Hagibis passing over the CE, the surface Chl-a concentration increased to a peak value of 0.33 mg/m³ within the CE ( Figure 4B ). Same as the SST cooling, the increase of surface Chl-a was restricted in the Kuroshio which is identified as low-nutrient water. Figure 4C shows that about three days after Hagibis passage, the surface Chl-a concentration in the CE increased obviously with the maximum value of 0.68 mg/m³ while that was only 0.35 mg/m³ to the south of Kuroshio. Different from the pattern of SST response, the most striking increasing of surface Chl-a concentration was located at the CE center probably due to eddy pumping lifting the nutricline (Falkowski et al., 1991). Besides, upwelling within the CE was strengthened by Hagibis with the maximum SLA varied from -121 to -131 cm ( Figure S1 ), implicating the further enhancement of nutrients for phytoplankton growth in the upper ocean. On October 17, five days after Hagibis, the increased phytoplankton biomass began to decline, and the maximum values of Chl-a concentration in the CE and to the south of the Kuroshio decreased to 0.56 and 0.19 mg/m³, respectively.

The surface Chl-a within the CE increased again soon after TC Neoguri passage with the area-averaged and maximum values reaching to 0.23 and 0.8 mg/m³, respectively. About one week later, it diminished rapidly with its maximum concentration decreasing to 0.51 mg/m³. Hagibis and Neoguri generated comparable and significant surface Chl-a increases although their intensity difference is conspicuous. Note that the translation speed of Neoguri (6.8 m/s) is slower than Hagibis (7.3 m/s), but Neoguri is not a slow-moving TC according to Li et al. (2021a). Li et al. (2021a) indicated that a TC that causes a significant ocean response usually has a translation speed<= 4 m/s. Chacko (2019) investigated the variability of TC-induced ocean biological response in the Indian Ocean from 1999 to 2016. Their results demonstrated that all TCs which had MSWs and translation speeds close to the corresponding values of Neoguri did not generate an obvious surface bloom. Similar results were also found by Pan et al. (2018) in the Northwest Pacific and South China Sea.

Seen from Figures 2 , 4 , the increase of surface Chl-a concentration apparently lagged the SST cooling because the phytoplankton need some time to grow by utilizing the nutrients injected from subsurface water. Pan et al. (2017) indicated that when the uplifted nitrate causes a phytoplankton bloom, Chl-a concentration reaches a peak three days after nitrate concentration begins to obviously increase. Zheng and Tang (2007) also found that the offshore bloom along the track of TC Damrey exhibited a Chl-a concentration peak five days after the lowest SST occurred.

Before the super typhoon Hagibis arrival, the vertical distribution of Chl-a obtained by BGC-Argo float 2902754 was characterized by a pronounced deep Chl-a maximum (DCM) reaching up to 0.48 mg/m³ ( Figure 5C ). The DCM is a ubiquitous phenomenon in worldwide oceans, formed by biological-chemical-physical interaction. In the stratified open ocean, the vertical stratification below the ML suppresses the supply of nutrients from subsurface layer, generally leading to a DCM beneath the MLD and at the top of nutricline within the euphotic layer where there are adequate nutrients and light (Furuya, 1990; Cullen, 2015; Wang, 2020; Chai et al., 2021; Zhuang et al., 2021). On October 9, 2019, The DCM occurred at the depth of 59.9 m just above the nutricline. Meanwhile, the EZD and MLD were 65.7 and 58.9 m, respectively ( Figure 6 ), which reflects the above relationship.

On October 14, two days after Hagibis passage, the MLD decreased by ~14 m to 44.9 m probably due to strong upwelling induced by Hagibis with no distinct change in stratification intensity ( Figures 5A , 6A , S2 ). A similar result was also observed by a Bio-Argo after TC Hudhud in the Bay of Bengal (Chacko, 2017). Shoaling of thermocline caused an intense cooling in the subsurface with the maximum temperature drop of 6°C at the depth of 57 m, while a slight temperature rise happened below the 100 m depth. At the same time, nitrate concentration increased distinctly in the upper layer but decreased below the 100 m depth. The temperature-salinity (T-S) property below the 100 m depth (near 25 kg/m³ sigma contour) was consistent with that of the Kuroshio water with high temperature and low nitrate ( Figure S3 ), indicating the invasion of the Kuroshio water due to the deep convergence accompanying with the Hagibis-induced upwelling. The daily absolute geostrophic currents derived from altimeter ADT data also certified the invasion of the Kuroshio, showed in Figure 7 .

The DCM disappeared and the maximum Chl-a concentration occurred in the ML on October 14 ( Figure 5C ). The Chl-a concentration was remarkably higher than that before Hagibis passage in the ML, but the opposite was true below the ML. To further explore the reason of these Chl-a variations, here we define the upper ocean Chl-a content integrated from the surface to 160 m depth as depth-integrated Chl-a since the Chl-a concentration below 160 m depth is very low and negligible. The depth-integrated Chl-a on October 14 was 20.4 mg/m² which was close to 19.8 mg/m² on October 9 before the influence of Hagibis ( Figure 6C ), indicating the increase of surface Chl-a was mostly the result of Chl-a redistribution due to TC-induced physical processes (upwelling and mixing) rather than phytoplankton growth.

One week after Hagibis passage, Chl-a concentration decreased a little in the upper 44 m but increased significantly below it on October 19 ( Figure 5C ), compared with that on October 14. And a small Chl-a concentration peak of 0.35 mg/m³ was formed at the depth of 28 m. The depth-integrated Chl-a rose to 24.9 mg/m² with a net increase of 4.5 mg/m², indicating a great growth of phytoplankton. The growth of phytoplankton consumed much nutrients injected from the subsurface layer, resulting in the reduction of nitrate concentration within the euphotic zone above 80 m depth ( Figure 5D ). The salinity in the ML was much lower than that 5 days ago because of heavy precipitation ( Figure 8A ). These results manifest that Hagibis induced strong upwelling and mixing, resulting in an apparent cooling and Chl-a increase in the upper ocean, and the phytoplankton growth lagged the response of temperature and nitrate by approximately one week.

The vertical distribution of Chl-a was almost uniform in the ML soon after TC Neoguri passage ( Figure 5C ). The temperature profile on October 24 also presented an obvious mixing process with the MLD increasing about 10 m and temperature increasing in subsurface layer. Differing from Hagibis, Neoguri raised the depth-integrated Chl-a immediately, especially in the ML. The increase of depth-integrated Chl-a reached up to 2.4 mg/m² despite the weak intensity of Neoguri. This unusual increase in phytoplankton biomass was attributed to both the uplift of nutrients due to the previous TC Hagibis and the influence of Neoguri. The abnormal profile of salinity on October 24 was affected by typhoon-induced flood-related hyperpycnal flow as reviewed by Liu et al. (2013) so that the salinity in the subsurface was very low compared with that before Neoguri passage ( Figure 5B ). An intense precipitation happened after Neoguri passage in the south of Japan ( Figure 8B ), which would bring flood-related fresh water with rich terrestrial sediment to offshore ocean, resulting in the low salinity in the subsurface layer. Meanwhile, this hyperpycnal flow might also contribute to the increase of subsurface nutrients and then Chl-a concentration on October 24 ( Figures 5C, D ). Five days later, the salinity profile on October 29 was restored to its pre-typhoon state ( Figures 5B , S3 ).

A noteworthy feature in this case is that the weaker TC Neoguri rapidly generated an obvious augment of phytoplankton biomass in the upper ocean. A category 5 typhoon Trami in late September 2018 with average translation speed of 6.1 m/s barely caused net increase of primary production in the upper ocean observed by BGC-Argo float 2902750, in spite of the stronger intensity and slower translation speed compared to Neoguri (Chai et al., 2021). Huang et al. (2016) reported that TC Jangmi caused a weaker surface bloom owing to the thick ML induced by TC Sinlaku that passed earlier. In our study, the surface Chl-a concentration following Neoguri was slightly less than that on October 14 ( Figure 5C ), but a dramatical increase happened in the subsurface layer. Hence, the sequential TC-induced primary production increase would be underestimated if only satellite sensing surface data were used.