1 Introduction

The concentration of dissolved oxygen (DO) in oceanic water, particularly subsurface water, is a useful indicator of physical and biogeochemical changes in the ocean (Kalvelage et al., 2013). Studies have shown that under conditions of global warming there is ongoing decline in DO concentration and expansion of intermediate oxygen minimum zones (OMZs) because of limited surface ventilation, decrease in oxygen solubility, and changes in global circulation patterns (Matear and Hirst, 2003; Schmidtko et al., 2017; Breitburg et al., 2018). Over the past 50 years, DO in the open ocean has decreased by an estimated 2%, or 77 billion tonnes, and the volume of water believed to be completely devoid of oxygen (anoxia) has more than quadrupled (Schmidto et al., 2017).

Hypoxia, which means low oxygen concentration, describes the condition where oxygen concentration is below the level at which higher organisms start to suffer from oxygen deprivation (Vaquer-Sunyer and Duarte, 2008; Ekau et al., 2010). Accordingly, a DO concentration of 2 mg L⁻¹ (62.5 μmol L⁻¹) is commonly applied as an upper limit of hypoxia in fisheries and ecology (Rixen et al., 2020). Deutsch et al. (2011) reported that hypoxic conditions occupy approximately 5% of the total global ocean volume (7.6 × 10¹⁶ m³), and that they are particularly pervasive in the thermocline (depth: 100 – 1000 m) of the eastern Pacific Ocean and northern Indian Ocean, where warm surface water transitions to the cold abyss and most respiration occurs. If areas of hypoxia expand, nutrient cycles and marine habitats could be affected with detrimental consequences for fisheries and coastal economies. For instance, shoaling of the tropical subsurface hypoxic zone might restrict the maximum living depth of animals (e.g., pelagic fish) to the oxygenated near-surface layer, making them more vulnerable to overexploitation by surface fishing gear (Prince and Goodyear, 2006; Stramma et al., 2010).

The northern Indian Ocean is one of the four largest permanent hypoxic zones in the global ocean, and the OMZs (O₂ < 20 µmol kg⁻¹) in both the Arabian Sea and the Bay of Bengal (BoB) have been extensively and thoroughly studied (McCreary et al., 2013; D'Asaro et al., 2019; Shenoy et al., 2020). However, the description of the hypoxic zone or OMZ in the Indian Ocean usually refers to data from the World Ocean Atlas, and focuses on the variation in intensity and the range of expansion of the OMZ in the north (i.e., north of 5°N) (e.g., Keeling et al., 2010; Stramma et al., 2010). In comparison, the eastern equatorial Indian Ocean (EEIO) is an area that is not well sampled. For example, the WOA2005 (World Ocean Atlas, 2005) database contains approximately three times more data for the coastal and northern Arabian Sea and BoB than it does for the EEIO. Thus, the hypoxic constitution of the EEIO area remains unclear.

The EEIO experiences highly variable current systems and water mass interaction in response to the semiannual reversal of the monsoon, which might lead to the unique hydrological and biogeochemical regimes in this region (Schott and McCreary, 2001; Schott et al., 2009; Garcia et al., 2018; Baer et al., 2019). Previous studies have shown that hydrological processes (e.g., eddies, vertical mixing, and horizontal advection) contribute substantially to the maintenance and spatial variation of the hypoxic zone (McCreary et al., 2013; Thomsen et al., 2016; Sarma and Udaya, 2018; Sarma et al., 2018). Based on the available data of the upper 300 m in the spring/fall intermonsoon and summer monsoon periods, the seasonal variation of the hypoxic zone in the EEIO was preliminarily investigated in a previous paper (Wang et al., 2018). However, it was not reported the hypoxia in the deep water as well as in the winter monsoon due to data limitations. Armed with the complete continuous data of DO for the down depth of CTD (the upper 1500 m) in the four seasons, we will redo the study by taking into consideration the hypoxia zone in the deep water as well as in winter monsoon. This study focuses on the vertical structure, seasonal variations in, and the factors that controlled, the spatial distribution of the hypoxic zone in the EEIO during monsoon and monsoon transformation.

2 Materials and Methods

Four cruises were conducted with a single meridional section (section A along 88°E) and 3 zonal sections between 2013 and 2017/2018 (Fig. 1). With a rosette of Niskin bottles attached to a conductivity-temperature-depth (CTD) profile (SBE 911 Plus, Sea-Bird Inc.), discrete seawater samples were collected at 88 stations during the spring intermonsoon period (April – May, 2013), 83 stations during summer monsoon (July – August, 2016), 102 stations in fall intermonsoon period (October – November, 2016), and 101 stations in winter monsoon (December 2017 to January 2018).

Water samples for nutrient, pH and chlorophyll a (Chl-a) determinations were taken at 2, 30, 75, 100, 150, and 300 m. Profiles of temperature and salinity were measured using a CTD (SBE 911 Plus, USA) at each station. The DO profile was measured at each station using a sensor (SBE 43, Sea-Bird Inc.) and Winkler titration. The sensors were calibrated before and after the cruise, and duplicate sensors were cross-referenced for precision. Discrete measurements of DO were also measured with a compact titrator and a sensor using the Winkler method (Metrohm 848 Titrino, Titrando Metrohm, Zofingen, Switzerland) (Chen et al., 2001). The DO profiles presented in this paper were derived from the CTD DO data calibrated using the Winkler titration method. The slope of the linear regression was close to 1.0 shown as Fig. 2 (i.e., R² value of 0.98 for the spring intermonsoon period, 0.96 for the summer monsoon, 0.98 for the fall intermonsoon period, and 0.95 for the winter monsoon). The apparent oxygen utilization (AOU) was calculated as [O₂] eq − [O₂] mean, where [O₂] eq is the DO solubility at equilibrium with the atmosphere and [O₂] mean is the in situ DO concentration.

The pH values were determined with a potentiometric pH sensor (Thermo Fisher Orion, America). Before determining the dissolved inorganic nutrient concentrations, the water samples were filtered through 0.45-μm cellulose acetate filters and stored at −20℃. Phosphate (PO₄-P) was measured following standard spectrophotometric procedures with a nutrient automatic analyzer (QuAAtro, Bran + Luebbe Gmbh) in an onshore laboratory. The Chl-a concentrations were determined by a Turner Design 10-AU fluorometer after 24 h extraction in 90% acetone at 4℃ in darkness.

The mixed layer depth was defined as the depth at which the temperature differed by 0.8℃ from the 10-m value (Levitus et al., 2000). The depth of the 20℃ isotherm is considered the representative depth of the thermocline (Sarma et al., 2016). WindSat sea surface temperature (APDRC, http://apdrc.soest.hawaii.edu) and SMAP sea surface salinity data (http://data.remss.com/smap/) were used to identify the general hydrological features (Yuan et al., 2021). OSCAR sea surface current data (http://www.oscar.naaa.gov) were extracted to identify the main surface currents (Praveen et al., 2014; Yuan et al., 2021).

3 Results 3.1 Seasonal Variation in Hydrological Conditions

Sea surface temperature in the EEIO exceeded 27℃ with a trend of decrease from north to south (Figs. 1a–d). Sea surface temperature was highest during the spring intermonsoon period owing to the clear sky and light winds, and lowest during the summer monsoon owing to the influx of low-temperature water in the south (Figs. 1a and b). Sea surface salinity decreased from west to east. High-salinity water (S > 35.0) was observed only west of 80°E during the spring intermonsoon period, flowing eastward to about 86°E and 89°E during the summer monsoon and the fall intermonsoon, respectively, and expanding to about 97°E during the winter monsoon. Sea surface salinity was lower at the mouth of the BoB and at the southern stations, reflecting the influence of BoB low-salinity water (BBW) and Indonesian throughflow water (ITFW) (Figs. 1e–h).

The spatial distribution of the surface currents showed that the Wyrtki Jet flowed eastward along the equator and turned southward to the west of Sumatra during the spring / fall intermonsoon and winter monsoon periods (Figs. 1i, k, and l). The Southwest Monsoon Current flowed northeastward into the BoB around the south of Sri Lanka during the summer monsoon, and the Northeast Monsoon Current flowed westward during the winter monsoon (Figs. 1j and l). There was an obvious cyclonic circulation between 5°S and 10°S, in which the current flowed eastward at 4° – 8°S, turned southward, and fed the westward South Equatorial Current near 10°S.

On the basis of the distribution of hydrological parameters and a potential temperature-salinity (T-S) diagram, the different water masses present along section A are shown in Fig. 3. The characteristics of these water masses are listed in Table 1. Three stations at the northern end of section A were characterized by BBW (S < 34.2), which was fresher and thicker in the summer monsoon. The 34.2 isohaline extended vertically to the depth of 80 m and horizontally southward to near the equator (1.2°N). In the spring and fall intermonsoon periods, BBW was observed in the upper 50 m, whereas it was not visible in the winter monsoon. Another low-salinity water (i.e., ITFW; S < 34.2) dominated the upper 40 m in the southern part of section A, and its northward extent varied from 10°S to 3°S in the spring intermonsoon period, from 2°S to 4°S during the summer monsoon, and to beyond 7.5°S in the fall intermonsoon and winter monsoon periods. Moreover, high-salinity Arabian Sea water (ASW; S > 35.2) was observed in the thermocline (90–150 m) around the equatorial regime (2°S – 1.5°N) in the spring intermonsoon period, but it was not visible during the summer monsoon. In the fall intermonsoon period, ASW was present in the mixed layer depth (0 – 140 m) and its core (S > 35.4) expanded from the surface to approximately 140 m between 0.5°N and 3.5°N but was concentrated to within 50 – 100 m between 2.5°S and 4.5°S. During the winter monsoon, ASW occupied the subsurface layer and expanded southward to 6°S and upward to the surface from 6°S to 2°S. High-salinity subtropical subsurface water (SSW; S > 35.4) was observed in the thermocline between 2°S and 8°S and the area it influenced was largest during the summer monsoon, followed by the spring and fall intermonsoon periods, and smallest during the winter monsoon. Additionally, high-salinity Indian Central Water (ICW; S > 35.0) occupied most of the intermediate water in the north and showed little seasonal variation. The Antarctic Intermediate Water (AAIW; S < 34.8) was below the ITFW, and found at depths of between 25 – 270 and 400 – 1500 m south of 4°S (Figs. 3e–h). Meanwhile, on the basis of the T-S-DO scatter diagram (Figs. 3i–l), hypoxic water was observed to originate from the low-temperature high-salinity ICW and existed with temperatures varied from 5℃ to 20℃ and salinity of 34.75 – 35.25. The deeper hypoxic waters (26.5 < σ₀ < 27.5) had a high and constant salinity (about 35.0), indicating that they were a single water mass, while the subsurface hypoxic waters had a lower salinity, indicating that they were mixed with other water masses.

3.2 Seasonal Variation in Dissolved Oxygen Distribution 3.2.1 Seasonal variation in dissolved oxygen and the oxycline along section A

The seasonal distribution of DO along section A is shown in Fig. 4. The DO concentration was in the range of 0.91 – 7.50 mg L⁻¹. The DO concentration was highest (DO > 6.5 mg L⁻¹) in the upper mixed layer, but then decreased sharply through the thermocline and was < 3.0 mg L⁻¹ in the intermediate water. The maximum DO gradient was sharpest (0.395 mg L⁻¹ m⁻¹) during the summer monsoon, followed by the intermonsoon periods, and it was lowest during the winter monsoon (Table 2). The oxycline (marked by the 3.5 mg L⁻¹ isoline) also showed considerable seasonal fluctuation. It was located between the depth of 50 and 140 m, and its meridional variation was consistent with that of the thermocline; that is, when the thermocline became shallower, the oxycline also became shallower, and vice versa (Fig. 3 and Fig. 4). Generally, the oxycline showed a V-shaped tendency from south to north, but its depth varied in different seasons. It was at the depth of approximately 100 m in the equatorial belt in the spring intermonsoon period, but it was approximately 20 m deeper in the other seasons (Fig. 4). The depth of the oxycline was also affected by the water masses present in different seasons. For example, in the equatorial belt, the oxycline was close to the upper boundary of the ASW in the spring intermonsoon period, but it was at the bottom of the ASW in the fall intermonsoon and winter monsoon periods. The oxycline dropped to the depth of approximately 140 m between 2°S and 6°S during the summer monsoon, but it shoaled to the depth of approximately 50 m in the winter monsoon, in line with the intrusion range of SSW.

3.2.2 Seasonal variation in the hypoxic zone along section A

Below the oxycline, the hypoxic zone (DO < 2.0 mg L⁻¹) extended southward from the north in a double tongue-like shape: one extension was in the subsurface layer centered at the depth of 150 m and the other was in the intermediate water centered at the depth of 800 m (Fig. 4). The hypoxic zone tongues extended furthest south in the spring intermonsoon period, followed by the fall intermonsoon and winter monsoon periods, and they extended least during the summer monsoon. The hypoxic zone in the subsurface layer expanded further south and crossed the equator during the winter monsoon, but it was observed only to the north of the equator (0.5°N) in the other seasons. The hypoxic tongue in the intermediate water expanded much further than the one in the subsurface layer; that is, it reached 7°S in the spring intermonsoon period, 4.5°S during the winter monsoon, 3.5°S in the fall intermonsoon period, and 3°S during the summer monsoon.

The vertical thickness of the hypoxic zone also varied spatially and seasonally. The maximum thickness was largest in the spring intermonsoon period and lowest during the summer monsoon (Table 2, Fig. 4). During the spring – summer transition, the hypoxic core (DO < 1.5 mg L⁻¹) contracted vertically by 25% (from 725 to 540 m), associated with shoaling of the hypoxic core from 125 to 110 m of the upper-core limit. Horizontally, the hypoxic zone presented contraction of 40% (from 2.2 × 10³ to 1.3 × 10³ km²). During the summer-fall-winter-spring transitions, the hypoxic core presented thickening of 25%; that is, from 540 to 700 m, then to 680 m, and finally to 725 m, respectively, for the above four seasons, with concurrent deepening of the hypoxic core upper limit by 40 m (from 110 to 150 m depth). In the vertical section, the hypoxic zone expanded by 40%, from 1.3 × 10³ to 2.2 × 10³ km² during the above seasonal transitions. Meanwhile, the vertical thickness of the hypoxic zone thinned from north to south. During the summer monsoon, it decreased from 870 m at 5°N to 20 m at 0.5°N in the subsurface water and to 200 m at 3°S in the intermediate water. During the winter monsoon, it was at the depth of approximately 930 m in the north, and at the depth of approximately 20 m at the equator in the subsurface water and at 4.5°S in the intermediate water (Figs. 4b and d).

3.2.3 Seasonal variation in horizontal distribution of the hypoxic zone

The horizontal distribution of the hypoxic zone also varied seasonally. At the subsurface core depth (150 m), the hypoxic zone extended southward from the north to the equator, but it was more intense (DO < 1 mg L⁻¹) in the spring intermonsoon period than in the other seasons. The hypoxic zone extended further south in the east than in the west during the spring intermonsoon period, and vice versa in the fall intermonsoon period (Figs. 5d and p). However, the range of extension of the hypoxic zone in the intermediate water (800 m) showed little variation from west to east (Figs. 5f, l, r and x). The southern boundary of the hypoxic zone extended to approximately 6°S in the spring intermonsoon period, 4°S during the winter monsoon, and 2.5°– 3.0°S in the summer monsoon and the fall intermonsoon periods. The hypoxic zone was smallest at the depth of 300 m, and it extended further southward during the winter monsoon (about 0.5°N) than in the other seasons (about 3°N). Comparison of the horizontal distribution of salinity and DO revealed that high-salinity oxygenated water was stronger along the equatorial belt at the depth of 150 m, which blocked the southward expansion of the hypoxic zone. The salinity was lower (S < 35.2) at the depths of 300 and 800 m, and it decreased from north to south but without variation in the east – west direction, indicating that the intermediate hypoxic zone was affected by the low-salinity oxygenated water from the south.

4 Discussion 4.1 Characteristics of the Hypoxic Zone in the Eastern Equatorial Indian Ocean

Our study area was located in the south of the BoB, which is a region found to experience intense depletion in DO levels at intermediate depths (100 – 1000 m), with a mean DO level of 16 ± 2 µmol L⁻¹ in the OMZ cores (Sarma, 2002; Paulmier and Ruiz-Pino, 2009; Chimmi et al., 2019). The hypoxic zone in the EEIO originates from the hypoxic intermediate water of the BoB, and it becomes two tongue-like extensions of hypoxic conditions in the subsurface and intermediate layers in the process of southward expansion (Fig. 4 and Fig. 5). The spatial distribution of the hypoxic zone is consistent with that reported in previous studies (Chinni et al., 2019; Twining et al., 2019). The mean DO concentration in the hypoxic zone (0.73 mg L⁻¹) in the EEIO is less intense than that found in the BoB (about 0.2 mg L⁻¹) (Table 2; Paulmier and Ruiz-Pino 2009). Both the extent of the horizontal southward expansion and the vertical thickness of the hypoxic zone in the EEIO are at a minimum during the summer monsoon (Table 2), consistent with the OMZ in the BoB (Paulmier and Ruiz-Pino, 2009). However, both are at a maximum in the spring intermonsoon period in the EEIO in comparison with the values in the fall intermonsoon period in the BoB (Table 2; Paulmier and Ruiz-Pino, 2009).

4.2 Effect of Monsoon Circulation on the Expansion of the Hypoxic Zone

The EEIO is a dynamically complex and highly variable system under monsoonal influence (Schott and McCreary, 2001; Iskandar et al., 2009; Hood et al., 2017; Ruma and Shaji, 2020). Hydrological processes, such as advection and vertical eddy mixing, governs the boundary of OMZ (Sarma, 2002; McCreary et al., 2013; Sarma et al., 2013, 2018; Sarma and Udaya, 2018). The mid-depth hypoxia in the EEIO is present owing to the net southward transportation of water in the depth range of 100 – 1000 m from the BoB (Sarma, 2002). Because the annual average meridional circulation returns southward in the Ekman layer (0 – 120 m) and the intermediate layer (500 – 1000 m) (Hu et al., 2005; Horri et al., 2013; Li and Chao, 2013), the hypoxic zone expands southward in a double tongue-like shape, with one extension in the subsurface layer and the other in the intermediate water (Fig. 4). Sarma (2002) reported that the southward water flux from the BoB is 20-fold stronger in the intermediate layer (500 – 1000 m, 17 × 10¹² m³ yr⁻¹) than that in the subsurface layer (100 – 250 m, (0.35 – 0.73) × 10¹² m³ yr⁻¹). This is one of the reasons why the hypoxic zone in the intermediate layer is much larger than that in the subsurface layer (Fig. 4).

Multiple zonal circulations near the equator also affect the distributional characteristics of the mid-depth hypoxic zone during its southward expansion. Although the intensity of the OMZ in the BoB is weaker than that in the Arabian Sea, the southward expansion of the hypoxic zone in the EEIO extends further than that in western areas owing to the supplement of oxygenated water (e.g., the Persian Gulf water mass and the Red Sea water mass) carried by the swift Somali and Omani coastal currents (Paulmier and Ruiz-Pino, 2009; McCreary et al., 2013; Shenoy et al., 2020). Between 6°N and 10°S, the subsurface DO level is lower in the EEIO than that in the western Indian Ocean (Stramma et al., 2010). The Southwest Monsoon Current, which is stronger during the summer monsoon, carries subsurface water with slightly higher oxygen concentration (DO > 2.0 mg L⁻¹) eastward to the EEIO. Thus, the core of the hypoxic zone (DO < 1.5 mg L⁻¹) occupies a smaller area during the summer monsoon than it does in the fall intermonsoon period (Figs. 1g, k; Figs. 4b and c; Unger et al., 2003; Schott et al., 2009). Similarly, the core of the hypoxic zone occupies a greater area during the winter monsoon than it does in the spring intermonsoon period because the Northeast Monsoon Current is stronger during the winter monsoon and it transports subsurface water with slightly lower oxygen concentration westward (Figs. 1i and l; Figs. 4a and d).

In the equatorial area, high-salinity ASW occupies a greater area around the thermocline (50 – 110 m) in the fall intermonsoon period, but it occupies a smaller area below the thermocline in the spring intermonsoon period, indicating that the fall Wyrtki Jets is more intense than the spring Wyrtki Jets (Figs. 1e and g), which is consistent with previous research (Iskandar et al., 2009; Nyadjro and Mcphaden, 2014; Duan et al., 2016; Wang et al., 2018). Therefore, the oxycline in the fall intermonsoon period is deeper in comparison with that in the spring intermonsoon period along the equatorial zone (2°N – 3°S) (Figs. 4a and c). Except of the surface Wyrtki Jets, the eastward Equatorial Undercurrent is quasi permanent at depths of 50 – 170 m (Schott and McCreary, 2001; Swapna and Krishman, 2008; Iskandar et al., 2009; Gnanaseelan and Deshpande, 2017). The high-salinity (S > 35.15) oxygen-rich (DO > 2.0 mg L⁻¹) ASW transported by the eastward Equatorial Undercurrent/WJs is stronger at the depth of 150 m in the equatorial belt (Figs. 5a–d); therefore, the subsurface hypoxic zone is limited north of 2°S in section A (Figs. 5e–h). Because this mass is stronger in the west in the spring intermonsoon period, the hypoxic zone extends further southward in the east than in the west, and vice versa in the fall intermonsoon period (Figs. 5a–c, m–o, d–f, and p–r). No high-salinity water (S > 35.2) was observed at deeper layers (e.g., 350 and 800 m, Fig. 5). The intermediate hypoxic zone, largely unaffected by the low-salinity oxygenated water from the south, extends further than the subsurface hypoxic zone and shows little east – west variation (Fig. 5). Huang et al. (2019) also reported that the maximum intensity of the zonal currents is located in the near-surface layer and that it weakens with increasing depth.

4.3 Contribution of Biological Activities to the Hypoxic Water

Except of hydrological processes, biological activities regulate the strength of the hypoxic water (Sarma, 2002; Sarma and Udaya, 2018; Ward et al., 2018). Low oxygen concentrations in intermediate layers of the northern Indian Ocean in all seasons (Paulmier and Ruiz-Pino, 2009; McCreary et al., 2013) was interpreted to be a result of higher inputs of organic matter from the surface layers. Higher primary production, increased organic matter export, and subsequent enhanced remineralization and DO consumption at depth (Matear and Hirst, 2003; Stramma et al., 2008). During the study season in 2013, chlorophyll maximum was observed at depths between 75 m and 100 m, thus the concentration of total Chl-a over the top 100 m in the warter column (total Chl-a (0 – 100 m)) was used to represent primary productivity. The total Chl-a (0 – 100 m) was higher in the BoB and decreased from the BoB to the EEIO (Fig. 6a). A linear regression analysis of the total Chl-a (0 – 100 m) and DO at the depth of 150 m showed significant negative correlation (R² = 0.52, P < 0.001), suggesting that subsurface hypoxia may be largely influenced by primary productivity (Fig. 6b). Organic matter produced in the euphotic zone mostly were oxidized, and removed oxygen from the water column in the upper 1000 m (Suess, 1980). As the oxygen is consumed by bacterial respiration / remineralization, nutrients increase and the pH decreases (Teng et al., 2014; Rixen et al., 2019; Wei et al., 2021). Paulmier and Ruiz-Pino (2009) found that a low-oxygen domain was accompanied by high PO₄-P. We also found that the AOU was positively correlated with PO₄-P but negatively correlated with pH throughout the seasonal cycle (Fig. 7), indicating the contribution of biological activities to oxygen consumption.

5 Conclusions

The occurrence of hypoxia depended on the balance between oxygen inputs and its consumption, biology may play a dominant role in defining the strength/intensity of the hypoxia, whereas hydrological processes (e.g., advection and vertical mixing) governs the expansion boundary in the EEIO. A simple conceptual model of the vertical pattern of the hypoxic zone based on hydrological processes is proposed (Fig. 8). The hypoxic zone from the BoB extends southward in a double tongue-like shape, with one extension in the subsurface layer centered at the depth of 150 m and the other in the deep layer centered at the depth of 800 m.

The spatial distribution of the hypoxic zone in the EEIO is controlled by the southward output flux of mid-deep water (100 – 1000 m) from the BoB. The hypoxic zone in the subsurface layer is much smaller in extent than that in the intermediate layer primarily because of the following two reasons: more hypoxic water is exported from the BoB in the intermediate layer than in the subsurface layer, and the hypoxic zone is mixed with high-salinity oxygenated waters carried by multiple zonal circulations (e.g., WJs, and Equatorial Undercurrent) near the equator (5°N – 2°S) in the subsurface layer (100 – 500 m); it is only mixed with low-salinity oxygenated water (DO < 3.5 mg L⁻¹) from the south (south of 3°S) in the intermediate layer (500 – 1000 m).

Acknowledgements

We thank the captain and crew of the 'RV Xiangyanghong 01' for their assistance with the field investigations. We thank Drs. Guangbing Yang and Guang Yang for their valuable comments. This work was supported by the National Natural Science Foundation of China (No. 41806099), and the Global Change and Air-Sea Interaction Project of China (No. GASI-04-HYST-06).