Study of Dissolved Oxygen Responses to Tropical Cyclones in the Bay of Bengal Based on Argo and Satellite Observations

Science of the Total Environment 659 (2019) 912–922

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Study of dissolved oxygen responses to tropical cyclones in the Bay of Bengal based on Argo and satellite observations

Huabing Xu a,c, Danling Tang a,c,⁎, Jinyu Sheng b,YupengLiua,c,YiSuib a Guangdong Key Laboratory of Ocean Remote Sensing, State Key Laboratory of Tropical Oceanography, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China b Department of Oceanography, Dalhousie University, Halifax, Nova Scotia, Canada c University of the Chinese Academy of Sciences, Beijing, China

HIGHLIGHTS GRAPHICAL ABSTRACT

• Three different types of DO responses to tropical cyclones were observed in the central Bay of Bengal • Storm-induced temporal variability of DO was affected by the shallow oxycline, mesoscale eddies and biochemical processes • Tropical cyclones with slow translational speeds can result in DO reduction and OMZ intensiﬁcation over the Bay of Bengal

article info abstract

Article history: Effects of tropical cyclones (TCs) on dissolved oxygen (DO) in subsurface waters (20–200 m) over the Oxygen Received 20 October 2018 Minimum Zones (OMZs) in the Bay of Bengal (BoB) are examined based on Argo and satellite data. Five TCs Received in revised form 19 December 2018 (Hudhud, Five, Vardah, Maarutha and Mora) during 2013–2018 are considered. Analyses reveal three types of Accepted 24 December 2018 DO temporal variability caused by the storm-induced mixing and upwelling. The first type features temporal Available online 26 December 2018 DO increases in subsurface waters (37–70 m) caused mainly by intense vertical mixing and downwelling. The Editor: Jay Gan second type features DO reductions in subsurface waters after the storms attributed to storm-induced upwelling. The third type features temporal DO increases at depths between 40 and 79 m and decreases at depths between Keywords: 80 and 150 m due to the combined effect of strong vertical mixing and upwelling. These three types of DO re- DO sponses can occur in different areas, depending on TC intensity, translational speed and Ekman pumping. The Tropical cyclone temporal DO variability is also influenced by the shallow oxycline (58.3 ± 16.7 m), mesoscale eddies and bio- Mixing chemical processes. Due to TC intensification, a pre-existing oceanic cyclonic eddy produced a large upwelling Upwelling and induced a long time of DO decrease in the subsurface layer. This study suggests three different types of DO Mesoscale eddy responses along the TC track in the OMZ, which is useful to evaluate the influence of TCs on the OMZ. Oxygen Minimum Zone © 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

⁎ Corresponding author at: Guangdong Key Laboratory of Ocean Remote Sensing, State Dissolved oxygen (DO) is critical for sustaining marine animal life. Key Laboratory of Tropical Oceanography, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China. Mobile macro-organisms are stressed or even die when oxygen concen- −1 E-mail address: [email protected] (D. Tang). trations drop below ~60 to 120 μmol kg (Stramma et al., 2008). Low

https://doi.org/10.1016/j.scitotenv.2018.12.384 0048-9697/© 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). H. Xu et al. / Science of the Total Environment 659 (2019) 912–922 913

DO restricts the maximum living depth of animals to the oxygenated level of QC is the delayed-mode quality control system (Thierry et al., near-surface layer (Stramma et al., 2012). The Oxygen Minimum 2016). Zones (OMZs) are known to contribute significantly to the oceanic pro- Based on Argo data, the following four indices were used. The iso- duction of N2O, a greenhouse gas which is more efficient than CO2 thermal layer depth (ILD) is defined as the depth at which the potential (Paulmier et al., 2011). water temperature is 1 °C lower than the near-surface temperature The OMZ is defined as vast areas of depleted oxygen (b20 μmol kg−1 (Vissa et al., 2013). The mixed layer depth (MLD) is defined as the of DO), due mainly to a combination of high primary productivity and depth at which the potential water density is 0.125 kg m−3 higher weak ventilation (Al Azhar et al., 2017). The Bay of Bengal (BoB) is than the surface density (Girishkumar et al., 2014; Levitus, 1983). The one of the four well-known tropical OMZs (Paulmier and Ruiz-Pino, barrier layer (BL) is defined as the layer between the bottom of the 2009). The immense freshwater influx from rivers and heavy precipita- mixed layer and the top of the thermocline (Thadathil et al., 2007). tion generate a strong salinity stratification and inhibit vertical ex- The oxycline depth is defined as a subsurface layer with the maximum change of oxygen in the upper ocean of the BoB (Thushara and gradient of DO in the vertical direction. Vinayachandran, 2016), creating a sharp and more intense OMZ in the BoB. The occurrence of the OMZ in the BoB was investigated in the past (Madhu et al., 2006; Sardessai et al., 2007; Sarma et al., 2013). Trop- 2.2. Tropical cyclones and satellite data ical cyclones (TCs) were found to induce intense vertical mixing of ocean waters and result in strong entrainment and upwelling (Price, There were 18 TCs passing through the BoB for the period – 1981). Several studies examined the main physical processes associated 2013 2018 (Joint Typhoon Warning Center). Only 6 TCs swept areas oc- with the TC-induced cooling of the sea surface temperature (Lin et al., cupied by the Argo DO sensors in the BoB. Since the focus of this study is 2003; Vissa et al., 2013; Warner et al., 2016) and phytoplankton blooms on the DO variability over the offshore ocean waters, TCs close to the fi (Madhu et al., 2002; Chacko, 2017; Vidya et al., 2017) in the BoB. Only a coast are not considered. As a result, only ve TCs (Hudhud, Five, few studies, however, were made on the DO responses to TCs in the sub- Vardah, Maarutha and Mora) were chosen to examine the DO responses surface waters of the OMZ over the central BoB. to TCs over the central BoB (Table 1 and Fig. 1). The intense TCs were found to play a very important role in DO dis- TC Hudhud originated from a low pressure system in the Andaman tributions over coastal waters and in the surface waters of deep oceans Sea on 6 October 2014. Hudhud became a severe cyclonic storm on 9 (Chen et al., 2012; Feng et al., 2012; Lin et al., 2014; Wang et al., 2017). October and reached its peak strength with the maximum sustained Chen et al. (2012) reported that the observed DO concentrations were wind speed (MWS) of ~185 km/h on 12 October (Fig. 1). TC Five, at a high level for about 6–8 days after typhoon Muifa due to mixing which is also named deep depression BOB 04 according to India Meteo- in the Changjiang Estuary. In the South China Sea (SCS), Lin et al. rological Department, formed over the central and adjoining southeast – (2014) documented that the DO concentrations increased one week BoB during 5 8 November 2014. It initially moved northwards on 5 No- after a typhoon. Prakash et al. (2012) found that a TC caused the DO de- vember 2014, took a loop on 6 November and turned westwards on 7 cline in the subsurface layer and shoaled the oxycline in the central Ara- November morning. TC Vardah originated as a low pressure system bian Sea. Over coastal waters of the BoB, Mitra et al. (2011) documented near the Malay Peninsula on 3 December 2016 and was designated as fi that the surface DO concentrations decreased in a short time due to the a tropical depression on 6 December. It gradually intensi ed into a cy- intrusion of saline waters. clonic storm on 8 December. Vardah had a westward movement and Although TCs have strong impacts on DO over coastal waters and in reached its maximum strength of ~130 km/h on 11 December as a the sea surface waters of deep oceans, knowledge is sparse on how the very severe cyclonic storm. It made a landfall close to Chennai and DO in the subsurface waters (20–200 m) responds to a TC in the OMZ degenerated into a remnant low on 13 December. TC Maarutha devel- over the central BoB. The primary objective of this study is to test the hy- oped over southeast BOB on 15 April 2017. It moved northeastwards fi pothesis whether DO responses to TC-induced mixing and upwelling and intensi ed into a deep depression over central BOB. Further moving depend on the intensity and translational speeds of storms and Ekman northeastwards, Maarutha reached its peak intensity on 15 April 2017. pumping velocities along the storm track in the subsurface waters of TC Mora developed from an area of low pressure over the southeastern the OMZ over the central BoB. The second objective is to examine the ef- BoB on 28 May. Mora reached peak strength with the MSW of fect of pre-existing oceanic conditions (such as the oxycline and meso- ~110 km/h, and then steadily weakened and dissipated early on 31 May. fi scale eddies) on the DO distributions under the influence of TCs. The storm track data of these ve TCs were obtained from the Joint Typhoon Warning Center (http://weather.unisys.com/hurricanes/). This data set consists of 6 hourly time series of locations of the TC center 2. Data and methods and the MWS at 10 m above the mean sea level. The translational speeds of each TC were estimated from time-varying positions of its storm cen- fi 2.1. Argo data ter. Daily elds of geostrophic velocities at the sea surface and merged sea level anomaly (SLA) were extracted from www.aviso.oceanobs. Observations from Argo floats (www.argodatamgt.org/)wereused com. The 8-day MODIS sea surface Chl-a with a horizontal resolution in this study to investigate temporal changes of DO in the subsurface of 4 km was obtained from oceancolor.gsfc.nasa.gov/. The daily remote fi – sensing fields of sea surface winds (at 10 m above the mean sea level) layers of the BoB during ve TCs for the period 2013 2018. Temperature ! τ and salinity measurements from four Bio-Argo floats (2902086, andwindstress( ) with a 0.25° × 0.25° resolution were obtained 2902114, 2902189 and 5903712) were also used to examine the hydro- from ftp.ifremer.fr/ifremer/cersat/products/gridded/MWF/L3/ASCAT/. fl The Ekman pumping velocity (EPV) was calculated from the surface graphic responses in the subsurface waters to the TCs. These four oats ! measured vertical profiles of DO, temperature, salinity and chlorophyll a wind stress vector ( τ )asfollows(Ye et al., 2017): (Chl-a) from ~5 to 2000 m (with different vertical resolutions) with fi ! time intervals of 5 days or 7 days. The Bio-Argo vertical pro les at ! fl τ depths from ~5 to 200 m were used in this study. These Argo oats ¼ − ð Þ EPV Curlz ρ 1 were equipped with DO sensors (Aanderaa Optode 4330). The claimed 0 f accuracy of the factory calibration is ~5% (or 8 μM) for the optode. The sensors were calibrated between 0% and 120% saturation (D'Asaro and −3 McNeil, 2013). The Bio-Argo data had two levels of quality control where ρ0 is the sea water density set to 1025.0 kg m and f is the (QC). The first level of QC is a set of automatic checks. The second Coriolis parameter. 914 H. Xu et al. / Science of the Total Environment 659 (2019) 912–922

Table 1

Values of the barrier layer (BL) thickness before the storm; the wind speed (U10), translational speed and Ekman pumping velocity (EPV) when the storm swept the Argos; the recorded time of Argos after the storm passage, the oxycline depth change; and the integrated DO change and DO change in the top 200 m after the storm. Information is also given about whether there were eddies and phytoplankton bloom after the storm over the areas occupied by Argos and the predominant processes to affect DO in the Oxygen Minimum Zone of the Bay of Bengal.

TC name Hudhud Five Vardah Maarutha Mora

Argo 5903712 (Ha) 2902087 (Hb) 2902114 (Hc) 5903712 (Fa) 2902086 (Fb) 2902114 (V) 2902189 5903712 TC time 08–12 Oct 2014 05–07 Nov 2014 07–12 Dec 2016 15–16 Apr 2017 27–30 May 2017 BL (m) 10 31 7 21 19 12 4 3 −1 U10 (m s ) 15.4 13.5 17.5 8.4 10.2 17.6 13.8 10.8 EPV (10−4 ms−1) −0.2 0.5 2.2 0.7 0.3 1.02 1.6 0.8 Translational speed (m s−1) 1.8 2.7 5.9 3–7.6 2.8–6.5 The recorded time after the During During ~6 ~2 ~3 ~3 ~2 ~4 storm (day) The oxycline depth change 39 → 49 81 → 70 51 → 56 40 → 30 65 → 61 46 → 56 65 → 65 79 → 64 (m) The integrated DO change 1.3 −2.2 −1.3 −2.5 −2 0.4 −2.1 −1.6 (μmol m−2) DO change Increase (37–70 Decrease Decrease Decrease Decrease Increase (40–79 Decrease Decrease m) (0–109 m) (11–51 m) (7–82 m) (7–91 m) m) (21–110 m) (47–95 m) Decrease (80–150 m) Eddy No No Cyclonic eddy Cyclonic eddy Cyclonic eddy Anticyclonic No No eddy Phytoplankton bloom Bloom No Bloom No No Bloom No No data The predominant processes Mixing + Upwelling Upwelling Upwelling Upwelling Mixing + Upwelling Upwelling downwelling upwelling

Note: (a) Since TC Five was looped over the areas occupied by Argos for three days, the 3-day average values are listed for the wind speed, translational speed and EPV for this storm. (b) Since the ASCAT wind website did not provide the wind speed data during TCs, the WindSat 3-day average wind data were used on 28–30 May 2017. (c) Since Argo 5903712 only recorded the temperature and DO during TC Mora, observed temperature and salinity made by a nearby Argo were used to calculate its BL.

3. Results (~90 mm day−1) would dilute the sea surface salinity, which is the case for the decrease of the sea surface salinity with time during the pas- 3.1. In situ observations of temperature, salinity and DO from Bio-Argos sage of Hudhud on 9 October. The potential density of sea water decreased with time in the subsurface layer between ~35 and 100 m. 3.1.1. TC Hudhud The MLD for the upper water column was about 18 m before Hudhud Fig. 2 presents three vertical proﬁles of observed temperature, salin- and increased to 29 m during Hudhud on 9 October. The observations ity, and DO in the top 200 m of the water column made by three Argo demonstrate very strong vertical mixing caused by Hudhud in the top ﬂoats over different areas before, during and after TC Hudhud. At water column up to the depth of 100 m (Fig. 2a). The DO concentrations

Argo_Ha, the observed temperature decreased with time in the top decreased with time slightly at the sea surface and increased signiﬁ- 20 m, but increased in the subsurface layer between ~20 and 100 m cantly in the shallow subsurface layer between 37 and 70 m (with the on 9 October during the passage of Hudhud. Heavy rainfall maximum increase of 79.8 μmol kg−1 at 46 m) from 2 to 9 October 2014. After 7 days, the DO concentrations gradually returned to the pre-storm level in the top 70 m.

At Argo_Hb, the vertical proﬁles of observed temperature, salinity and potential density indicate upwelling during TC Hudhud. This upwelling brought deep subsurface waters of low temperature and high salinity to the shallow layer on 9 October (Fig. 2c–d). The DO concentrations decreased signiﬁcantly at depths b 109 m (with the maximum decrease of 59 μmol kg−1 at 72 m) from 4 to 9 October 2014. After 5 days, the DO concentrations returned to the pre-storm level in the top ~98 m.

At Argo_Hc, the vertical proﬁles in the top 200 m show a similar upwelling on 15 October (6 days after TC) compared to the pre-storm level (Fig. 2e–f). The DO concentrations decreased at depths between 11 and 51 m (with the maximum decrease of 49 μmol kg−1 at 28 m) from 4 to 15 October 2014. It should be noted that the temperature and salinity in the top 200 m did not return to the pre-storm values at 11 days after TC Hudhud. The DO concentrations in the top 50 m did not return to the pre-storm values until 25 October (16 days after TC, not shown).

3.1.2. TC Five After TC Five, two vertical profiles of temperature, salinity and DO in fl Fig. 1. Map of the Bay of Bengal (BoB), storm tracks of five tropical cyclones (TCs). The the top 200 m of the water column were observed by two Argo oats storm tracks of tropical cyclones and Argo positions are marked by black lines and (Argo_Fa and Argo_Fb) over different areas (Fig. 3a–d). The MLD was different symbols respectively for Cyclone-4 Hudhud (2014), Tropical Storm Five about 8 m before TC Five and increased to about 16 m after TC Five (2014), Cyclone-1 Vardah (2016), Tropical Storm Maarutha (2017), and Cyclone-1 Mora over the area occupied by Argo_Fa on 7 November 2016. The MLD was (2017). The 6 hour positions of storm centers are marked by colored dots along the 35 m before TC Five and decreased slightly to 34 m after TC Five over storm track, of which sizes and colors represent the storm intensity. TD, TS and T1–T4 stand for respectively Tropical depression, Tropical Storm and Tropical Cyclone 1 to 4 the area occupied by Argo_Fb on 8 November 2014. The two vertical based on Saffir-Simpson scale. profiles reveal the upwelling which brought the deeper subsurface H. Xu et al. / Science of the Total Environment 659 (2019) 912–922 915

Fig. 2. Vertical proﬁles of (a, c, e) temperature (°C) and salinity (psu), (b, d, f) dissolved oxygen (DO) (μmol kg−1) and potential density (D) (kg m−3) in the upper 200 m observed by Argo

ﬂoat 5903712 (Ha), 2902087 (Hb) and 2902114 (Hc) before, during and after the passage of TC Hudhud. Due to the data gap of Argo_Hc on 10 October, observations made on 15 October were used.

water to the shallow layers. The observed DO concentrations decreased to ~43 m after Vardah on 13 December. These observed vertical proﬁles with time in the subsurface layers between 7 and 82 m (with the max- suggest very strong vertical mixing caused by Vardah in the upper water imum decrease of 137.2 μmol kg−1 at ~32 m) and between 7 and 91 m column up to 90 m and storm-induced upwelling in the deep subsurface −1 (with the maximum decrease of 62.2 μmol kg at ~62 m) at Argo_Fa layer between 90 and 200 m. Compared to the DO responses of TCs and Argo_Fb respectively after TC Five. Hudhud and Five, the observed DO concentrations during TC Vardah The observed changes measured by Argo_Fb suggest a continuous show the combined changes in the upper water (Fig. 3f). The observed upwelling which brought the deep subsurface waters of low DO to the DO decreased slightly with time in the subsurface layer between 18 and shallow subsurface layer on 13 November. The observed DO concentra- 39 m, increased signiﬁcantly in the subsurface layer between 40 and tions were also low in the subsurface layers between 29 and 95 m on 13 79 m (~52 m, the maximum increase of 55.5 μmol kg−1), and decreased November (Fig. 3d) and did not return to the pre-storm values until 23 in the deep subsurface layer between 80 and 150 m (~122 m, the max- November 2014 (18 days after the storm, not shown). imum decrease of 27.6 μmol kg−1).

3.1.3. TC Vardah For TC Vardah, the observed temperature exhibits a three-layer ther- 3.1.4. TC Maarutha and Mora mal structure, with a cooling layer near the surface, a warming layer The storm-induced temporal changes of DO observed by Argo_Maa right below and another cooling in the deep subsurface layer (Fig. 3e). and Argo_Mora during TCs Maarutha and Mora were similar to the ob-

The MLD of the water column was ~38 m before Vardah and increased servations made by Argo_Hb during TC Hudhud. 916 H. Xu et al. / Science of the Total Environment 659 (2019) 912–922

Fig. 3. Vertical proﬁles of (a, c, e) temperature (°C) and salinity (psu), (b, d, f) dissolved oxygen (DO) (μmol kg−1) and potential density (D) (kg m−3) in the upper 200 m observed by Argo

ﬂoat 5903712 (Fa), 2902086 (Fb) and 2902114 (V) before, during and after the passage of TCs Five and Vardah. Due to the data gap of Argo_V on 8 December, observations made on 3 December were used.

3.2. Wind and EPV affected by TCs (Fig. 4c). Vardah swept the area occupied by Argo float with the daily wind speed of 17.6 m s−1 and high EPV of 1.02 × 10−4 ms−1. TC Hudhud swept the areas occupied by the Argo floats on 9 October − 2014 (Fig. 4a), with translational speeds of ~1.8 m s 1. The daily wind 3.3. Geostrophic circulation and eddy fields speeds reached 15.4, 13.5 and 17.5 m s−1, and EPVs were −0.2 − − − − ×10 4, 0.5 × 10 4 and 2.2 × 10 4 ms 1 over the areas occupied by Satellite data of SLA and surface geostrophic currents show a small-

Argo_Ha,Argo_Hb and Argo_Hc respectively (Table 1). These large and size cyclonic eddy (counter-clockwise currents associated with negative positive EPVs indicate occurrence of strong wind-induced upwelling. sea level anomalies) and a small-size anticyclonic eddy (clockwise cur- The negative EPV indicates occurrence of wind-induced downwelling. rents associated with positive sea level anomalies) before Hudhud on 7 ﬂ TC Five swept the areas occupied by Argo oats on 5 November 2014 October 2014 (Fig. 5a). Argo_Ha moved southward during Hudhud, and looped over the areas for three days. The average translational with the above-mentioned cyclonic eddy enlarging its size and moving −1 speed was 2.7 m s . The 3-day average wind speeds reached about to the area occupying by Argo_Ha. During the same period, Argo_Hb − − − − 8.4 and 10.2 m s 1 and EPVs were 0.7 × 10 4 and 0.3 × 10 4 ms 1 re- moved southwestward from the edge of the anticyclonic eddy and spectively over the areas occupied by Argo_Fa and Argo_Fb from 5 to 7 was located outside of the anticyclonic eddy after TC Hudhud. Argo_Hc November 2014 (Table 1). TC Vardah swept the area occupied by moved eastward to reach the southeastern part of the cyclonic eddy. −1 Argo_V on 10 December, with translational speeds of ~5.9 m s The SLA around Argo_Hc changed from −12 cm (7 October) to H. Xu et al. / Science of the Total Environment 659 (2019) 912–922 917

Fig. 4. Satellite remote sensing data of ASCAT wind vectors (m s−1) and Ekman pumping velocity (EPV, m s−1) when TCs Hudhud (a), Five (b) and Vardah (c) swept the areas occupied by Argo ﬂoats respectively.

Fig. 5. Satellite remote sensing data of sea level anomaly (SLA, cm) and associated surface geostrophic current vectors (m s−1) before (a, d, g), during (b, e, h) and after (c, f, i) the passage of TCs Hudhud, Five and Vardah. The inverted black triangles represent Argo positions. Dashed lines and color dots represent the storm track and intensity of TCs. The pink (black) dotted line represents the SLA = −5 (+5) cm. Black arrows represent the directions of Argo's movements. 918 H. Xu et al. / Science of the Total Environment 659 (2019) 912–922

−14 cm on 9 October and then declined to −16 cm on 16 October, dem- on 6 December to 12 cm on 13 December and then declined to 8 cm onstrating the intensification of the cyclonic eddy. on 18 December, indicating that the anticyclonic eddy was weakening. There was a large-size cyclonic eddy over the area under the direct The storm reduced the size and strength of the anticyclonic eddy, influence of TC Five on 4 November 2014 (Fig. 5d). This cyclonic eddy which affected the downwelling caused by this eddy. strengthened and enlarged its size due to the storm (Fig. 5d–f). Argo_Fa For TCs Maarutha and Mora, satellite data show that no eddies oc- moved northeastward during TC Five and reached the southeastern curred before and during the storms. edge of the cyclonic eddy after the storm. Argo_Fb moved westward during TC Five and was located at the southern edge of the cyclonic 3.4. In-situ and satellite observations of Chl-a eddy after the storm (Fig. 5d–f). The SLA around Argo_Fb decreased from −4 cm on 4 November to −6 cm on 8 November and to −9cm After TC Hudhud (October 08–15, 2014), the Chl-a concentrations on 13 November, indicating intensification of the above-mentioned cy- increased significantly and reached up to 2.8 mg m−3 (Fig. 6b). In the clonic eddy, which was similar to the eddy under the influence of TC second week after Hudhud, high Chl-a concentrations occurred along

Hudhud. the track of Hudhud (Fig. 6c). Except for Argo_Hb,highChl-a concentra- For TC Vardah, there was a large-size anticyclonic eddy over the area tions were observed at both Argo_Ha and Argo_Hc. The vertical Chl-a occupied by Argo_V (Fig. 5g). A small-scale cyclonic eddy strengthened proﬁles around Argo_Hc show an increase of Chl-a concentrations up signiﬁcantly in size and moved gradually northeastward during TC to 1.6 mg m−3 on 10 October 2014 (Fig. 7b). The Chl-a concentrations Vardah (Fig. 5g–i). The anticyclonic eddy reduced its size and moved increased furthermore and reached up to 4.6 mg m−3 (20–30 m gradually northeastward. The SLA around Argo_V changed from 16 cm depth) on 20 October 2014. The Chl-a concentrations returned to the

Fig. 6. Satellite remote sensing data of chlorophyll a (mg m−3) before, during and after the passage of TCs Hudhud, Five and Vardah. The inverted black triangles represent the Argo positions. Dashed lines and color dots represent the storm track and intensity of TCs. H. Xu et al. / Science of the Total Environment 659 (2019) 912–922 919

−3 Fig. 7. Vertical profiles of chlorophyll a (mg m ) in the upper 100 m from four Bio-Argo floats (Argo_Hb and Argo_Hc during TC Hudhud, Argo_Fb during TC Five and Argo_V during TC Vardah). The red arrows represent the day when TCs swept the areas occupied by Argos. pre-storm values on 25 October 2014. Similarly to the surface Chl-a subsurface waters of low-oxygen to the shallow subsurface waters around Argo_Hb,theverticalChl-a profiles also show no phytoplankton two or a few days after TCs. The positive EPVs over the area occupied bloom after Hudhud (Fig. 7a). After TC Five, there was no phytoplankton by these Argos all showed the storm-induced upwelling. This response bloom in the surface waters under the influence of the TC (Fig. 6d–f). also occurred in the central Arabian Sea where tropical storm 05A The vertical Chl-a profiles also demonstrate no phytoplankton bloom transported the suboxic waters to a much shallower layer (Prakash after TC Five (Fig. 7c). But after TC Vardah, high Chl-a concentrations oc- et al., 2012). In the open ocean of the SCS, however, a different DO re- curred over the area under the influence of TC. The vertical Chl-a pro- sponse was found. Lin et al. (2014) reported that, one week after TC files around Argo_V show high concentration of the subsurface Chl-a Nanmadol, the layer of high DO concentrations in the SCS extended (≥1mgm−3). Three days after TC Vardah, the subsurface Chl-a concen- from the surface to a depth of 35 m, for two main reasons. Firstly, the trations in the upper 40 m increased to 1.03 mg m−3 and enhanced to OMZ did not exist in the subsurface layer below 100 m in the central 1.69 mg m−3 in the next three measuring cycles (Fig. 7d). SCS. The storm-induced upwelling could not bring the low-oxygen waters in the deep subsurface layer to the shallow subsurface layer. Sec- 4. Discussions ondly, the Kuroshio Current transported rich DO waters from the Western Pacific Ocean to the surface water in the SCS. 4.1. Effects of TCs on DO responses in subsurface waters It should be noted that a number of seismic events were recorded in theregionduringTCHudhud(Akilan et al., 2017). The volcanic mate- The satellite data and in-situ observations presented above suggest rials were emitted into the troposphere and were transported to a three different types of temporal changes of DO concentrations in re- long distance due to TC Hudhud. The volcanic materials are rich in or- sponse to TCs in subsurface layers over the central BoB. ganic components. When these organic materials mix with stream or The first type of DO responses occurred over the area occupied by sea water, the materials demand more oxygen to decompose the or-

Argo_Ha and features temporal increases of DO in subsurface waters ganic materials. Furthermore, the heavy rainfall would help these or- (37–70 m) during the passage of TC Hudhud. This was mainly due to ganic materials enter the sea water during TC Hudhud. This chemical the intense storm-induced vertical mixing and downwelling (EPV of process would most likely decrease the DO concentrations in the sea −0.2 × 10−4 ms−1) that blended the surface waters of rich DO with water during Hudhud. Further studies are needed to quantify the effect subsurface layers of low DO (Fig. 2a–b). This type also occurred in the of volcanic materials on the DO in the BoB. Changjiang Estuary. Chen et al. (2012) reported the intense vertical The third type of DO responses occurred during TC Vardah (Argo_V). mixing induced by TC Muifa which made the DO concentrations at The storm-induced changes in DO concentrations in this type is charac- 50 m increased in a short time of about 1–2 h after the storm. Ni et al. terized as a three-layer vertical structure, with DO decrease in time be- (2016) also documented the strong vertical mixing induced by TC tween 18 and 39 m, increase between 40 and 79 m and decrease Morakot which made the DO concentrations to reach maximum between 80 and 150 m. This type was due mainly to the combination (~5.5 mg/L) at the bottom of Changjiang Estuary (around 50 m) on of strong vertical mixing and upwelling induced by the storm. For this 7–11 August 2009. It should be noted that the observed DO concentra- type, the strong storm-induced vertical mixing blended the rich- tions had smaller temporal variability in the surface layer than in the oxygen surface waters with the low-oxygen waters in the shallow sub- subsurface waters during TCs. The air-sea flux of O2 would increase dur- surface layer between 40 and 79 m. The storm-induced upwelling ing TCs due to stronger-than-normal mixing at the sea surface. Owing to brought the low-oxygen waters in the deep subsurface layer up to the the OMZ existed in the subsurface, however, the intense vertical mixing intermediate subsurface waters between 80 and 150 m. This is the blended the low-oxygen subsurface waters to the surface layer. The DO first report about the three-layer vertical structure of DO variability in increase introduced by the air-sea flux compensated the DO decrease of response to a TC in the OMZ over the central BoB. surface waters, resulting in smaller temporal DO variability in the surface waters than the subsurface waters during the passage of TCs. 4.2. Effects of wind speeds and translational speeds on DO variability during The second type of DO responses occurred during TCs Hudhud TCs

(Argo_Hb and Argo_Hc), Five, Maarutha and Mora. This type is charac- terized as the decrease of DO in the subsurface layer during the storm, For TC Hudhud, the wind speeds of the storm were ~15.4 m s−1 over due mainly to the TC-induced upwelling that advected the deep the area occupied by Argo_Ha (Table 1). The strong wind speeds led to 920 H. Xu et al. / Science of the Total Environment 659 (2019) 912–922

intense vertical mixing and downwelling, which increased the DO con- For TCs Hudhud and Five, Argo_Hc (TC Hudhud), Argo_Fa and Argo_Fb centrations in the shallow subsurface layer between 37 and 70 m during (TC Five) were located in the cyclonic eddies. Fig. 7 presents vertical pro- the passage of Hudhud. The storm-induced DO decrease in subsurface files of differences in DO after and before the storm passage for five TCs in waters over the areas occupied by the Argo floats during TCs Hudhud the top 200 m. These differences are used to represent the storm-induced

(Argo_Hb and Argo_Hc), Five, Maarutha and Mora, owing to strong up- changes in DO. The vertically integrated DO change in the top 200 m gen- welling (positive EPVs) after a few days of TCs (Table 1). Several previ- erated by TC Five was the largest among the five storms and about −2.5 −2 ous studies demonstrated that a relatively slow-moving storm (with μmol m over the areas occupied by Argo_Fa (Table 1).TheBLbeforeTC −1 translational speeds ≤ 4ms ) can generate strong upwelling to bring Five over the area occupied by Argo_Fa, however, was thick to 21 m, the subsurface cold waters to the surface layer (Price, 1981; Zhao which would inhibit the storm-induced upwelling. Moreover, the wind et al., 2008; Sun et al., 2010; Lin, 2012; Chacko, 2017). The translational speeds of TC Five were weakest among the five TCs. In addition to the speeds of these four TCs were b4ms−1 near the areas occupied by the slow translational speeds during TC Five, two plausible reasons can ex- Argo floats (Table 1). Hence, these four TCs could generate relatively plain the largest storm-induced DO change induced by TC Five. Firstly, strong upwelling to bring cold and low-oxygen waters in the deep sub- there was a pre-existing cyclonic eddy over the areas under the direct in- surface layers to the shallow subsurface layers. fluence of TC Five. Xu et al. (2017) reported that a pre-existing cyclonic In addition, the strongest winds of Vardah (17.6 m s−1) caused vig- eddy produced the largest upwelling due to cyclone intensification off orous vertical mixing during the storm passage. This intense mixing northeastern Taiwan. Therefore, the pre-existing cyclonic eddy during injected warm surface waters downward and made the upper ocean TC Five could generate strong upwelling and bring the deep subsurface waters colder (Emanuel, 2001). The intense vertical mixing also brought waters with low DO to the shallow subsurface layer. Secondly, TC Five lin- subsurface waters with lower oxygen to the upper layers and injected gered over the area occupied by Argo_Fa for three days. This looping the surface waters with higher oxygen downward. However, the trans- storm would also generate strong upwelling (Chen and Tang, 2012). In −1 lational speed (~5.9 m s ) was relatively fast during TC Vardah. The addition, the observed DO concentrations around Argo_Hc and Argo_Fb storm-induced upwelling was only able to transport the deep subsur- remained low in the subsurface layers and did not return to the pre- face waters up to 80 m and reduce the DO concentrations in the subsur- storm values until 16 and 18 days after TCs Hudhud and Five respectively. face layer between 80 and 150 m three days after Vardah. The storm- After TCs, the cyclonic eddies were strengthened by TCs, as shown by the induced upwelling caused the vertical movement of water mass. The SLA decreases around Argo_Hc and Argo_Fb. Hence, the strengthened cy- vertical velocity of the movement was slow (EPV = 1.02 clonic eddies induced the continuous upwelling, transporting deep sub- ×10−4 ms−1). After 3 days, the water movement yielded an estimation surface waters of low-oxygen upwards to the shallow subsurface layers. of the wind-driven upwelling of 26 m. The combined effect of mixing The DO recovery time (16.7 ± 1.2 days) to the pre-storm level over the and upwelling resulted in the three-layer vertical structure of DO vari- cyclonic eddies (Argo_Hc,Argo_Fa and Argo_Fb)wassignificantly longer ability, which has not been reported in the literature. Two relevant but than the recovery time (6.7 ± 1.5 days) at other three Argos (Argo_Ha, different studies were made previously about the three-layer vertical Argo_Hb and Argo_V; p b 0.05). structure of temperature owing to this combined effect after TCs In order to quantify the similarities/dissimilarities for the storm- Rammasun (2008) and Kalmaegi (2014) in the top 200 m (Pei et al., induced DO changes among five TCs, we used the cosine similarity to 2015; Zhang et al., 2016). The similar characteristics of these three TCs analyze the DO changes in the top 100 m after and before TCs based were the fast translational speeds (about 4.9 m s−1 for Rammasun, on (Liu et al., 2018): 8.3 m s−1 for Kalmaegi, and 5.9 m s−1 for Vardah). The fast-moving P TCs were suggested to induce the three-layer vertical structure. nðÞ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 Ai qBiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi cosθ ¼ P P ð2Þ n 2 n 2 1 Ai 1 Bi 4.3. Effect of shallow oxycline on the DO variability where, cosθ is the cosine similarity of the DO changes between two The oceanographic observations reveal that the depths of oxycline Argos in the top 100 m, Ai (Bi) is the DO changes after and before TCs (58.3 ± 16.7 m; mean ± SD) varied between 39 and 81 m over the in the top 100 m over 8 Argos. The storms-induced DO changes in the areas occupied by Argo floats before the passages of five TCs (Table 1). subsurface waters between 20 and 40 m around Argo_H and Argo_F The observed temperature and salinity of water columns reveal that c a are presented in Fig. 8. Table 2 shows a high similarity value of 0.86 the storm-induced vertical mixing blended the near-surface waters with the subsurface waters (N80 m) during TCs Hudhud and Vardah (Figs. 2aand3e). The storm-induced vertical mixing resulted in increas- ing of DO in the subsurface waters and weakening of the OMZ. This was demonstrated by the observed DO increases at depths between 37 and 70 m during Hudhud and between 40 and 79 m during Vardah. Due to the shallow oxycline (39–81 m), the storm-induced upwelling during storms can transport the low-oxygen waters in the deep layers upwards to the relatively shallower subsurface layers. Hence, the shallow oxycline in the OMZ of the central BoB facilitated the variability of subsurface DO under the influence of TCs.

4.4. Effect of mesoscale eddies on DO variability

Mesoscale eddies are ubiquitous features in the ocean. Anticyclonic eddies were found to transport surface waters with rich oxygen to the subsurface layers, leading to weakening of the OMZ over the BoB (Sarma and Udaya Bhaskar, 2018). By comparison, lower oxygen levels were observed over regions influenced by cyclonic eddies (Sarma et al., Fig. 8. Vertical profiles of differences in observed DO concentrations in the top 200 m made fi 2016). Therefore, the mesoscale eddies have signi cant impact on the by each Argo float after and before the storm passage for five TCs Hudhud, Five, Vardah, OMZ (Sarma et al., 2018). Maarutha (Maa) and Mora. H. Xu et al. / Science of the Total Environment 659 (2019) 912–922 921

Table 2 pump” effects (mainly the storm-induced mixing and upwelling) in- Values of the cosine similarity for storm-induced DO changes in the top 100 m after and duced by TCs. These responses can occur in different areas, depending before the passage of TCs Hudhud, Five, Vardah, Maarutha (Maa) and Mora among the on the TC's intensity, translational speed and Ekman pumping velocity eight Argos. along the TC track. Careful considerations of three different types of Ha Hb Hc Fa Fb Vardah Maa Mora DO responses are needed in order to evaluate the inﬂuence of TCs on

Ha 1 −0.47 −0.18 −0.39 −0.85 0.76 −0.51 −0.27 the DO variability of the OMZ in the BoB. The major conclusions from − Hb 1 0.17 0.18 0.73 0.66 0.93 0.89 the present study include: Hc 1 0.86 0.25 −0.04 0.16 0.02 F 1 0.4 −0.18 0.18 −0.03 a 1. The intense vertical mixing and downwelling induced by TC Hudhud Fb 1 −0.93 0.82 0.6 Vardah 1 −0.79 −0.64 increased the DO in the subsurface layer between 37 and 70 m Maa 1 0.92 around Argo_Ha. The DO reductions in subsurface waters were attrib- Mora 1 uted to storm-induced upwelling two to four days after TCs Hudhud

(Argo_Hb and Argo_Hc), Five, Maarutha and Mora. A storm with a relatively slow translational speed would induce the DO decrease and between Argo_Hc and Argo_Fa, associated mainly with the pre-existing intensify the OMZ to reduce the living space of animals. The combi- cyclonic eddy which shoaled the oxycline at these two sites. At Argo_Hb, nation of strong vertical mixing and upwelling induced by TC Vardah Argo_Fb, Argo_Maa and Argo_Mora, however, the large storm-induced generated a three-layer structure of DO variability in the vertical di- DO changes between 50 and 80 m without a pre-existing cyclonic rection, with DO decreases between 18 and 39 m, increases between eddy (Fig. 8). The cosine similarity values between Argo_Hb and 40 and 79 m and decreases between 80 and 150 m. Argo_Fb, Argo_Maa, Argo_Mora are high and up to 0.73, 0.93 and 0.89 2. The pre-existing oceanographic conditions were found to play an im- respectively (Table 2). The DO changes around these Argos were mainly portant role in regulating the DO concentrations. The shallow due to the storm-induced upwelling. The storm-induced mixing en- oxycline in the OMZ of the BoB facilitated the temporal variability hanced the DO concentrations between 40 and 60 m during TCs of subsurface DO under the influence of TCs. Due to TC intensifica- Hudhud and Vardah over Argo_Ha and Argo_V. As a result, the cosine tion, a pre-existing oceanic cyclonic eddy produced a large upwelling similarity value in the top 100 m between Argo_Ha and Argo_V was and induced a long time of DO decrease in the subsurface layer. high (0.76). 3. The storm-induced physical processes were found to play a significant role in determining the boundary of the OMZ. The phytoplank- fl 4.5. In uence of biological processes on DO variability ton bloom caused by TCs plays a dominant role in determining the absolute concentration of DO in the OMZ. Prakash et al. (2013) observed a significantly large correlation between oxycline and thermocline, suggesting that physical processes played a significant role on the variations in oxycline depth than biolog- In our future work, we will collect more data during the TCs and use ical processes in the BoB. The storm-induced intense vertical mixing and a coupled physical and biogeochemical model to get better understand- upwelling affect the vertical movement of the thermocline to govern ing processes affecting the DO changes over the study region. the boundary of the OMZ in the study region. It should be noted that the biological processes play a dominant role in determining the abso- Acknowledgments lute concentration of DO and defining the strength of the OMZ. Based on the satellite data and the vertical profiles of Chl-a concentrations, We sincerely thank two anonymous reviewers for their constructive after TC Hudhud and Five, phytoplankton bloom did not occur over suggestions and insightful comments to improve the quality of the man- uscript. This study was funded by Key Project (41430968 and the areas occupied by Argo_Hb and Argo_Fb, due mainly to the thick 52102730) of the National Natural Science Foundation of China and BLs around Argo_Hb and Argo_Fb and the weak intensity of TC Five (Table 1). Hence, the physical processes during TCs play a major role the Key Project (2015HS05) of Collaborative Innovation Centre for in determining the DO change in the subsurface layers. After passages 21st Century Mari-time Silk Road Program (GDUFS). This study was supported by LORS and LTO Overseas Visiting Fellowship Program of TCs Hudhud and Vardah, however, the areas occupied by Argo_Ha, (LTOOVFP 1601). Argo floats data were acquired freely by the Interna- Argo_Hc and Argo_V showed several fold increases in primary production due to TCs. At the early stage, the increased phytoplankton would tional Argo Program (http://www.argodatamgt.org/). enhance the DO because of the photosynthesis. In the SCS, Lin et al. (2014) reported that TC induced upwelling which transported the References nutrient-rich water to stimulate phytoplankton growth. The photosyn- Akilan, A., Azeez, K.K.A., Schuh, H., 2017. Atmospheric storm triggered and intensified by thesis of phytoplankton generated oxygen to increase the DO concen- geodynamics: case studies from Andaman Sea and Bay of Bengal region in the Indian trations. But after the phytoplankton reached peak phase, oxygen Ocean. Pure Appl. Geophys. 174, 2173–2194. Al Azhar, M., Lachkar, Z., Lévy, M., Smith, S., 2017. Oxygen minimum zone contrasts be- consumption would enhance due to the increase in organic matter sink- tween the Arabian Sea and the Bay of Bengal implied by differences in ing and subsequent decomposition. Hence, the DO variability over these remineralization depth. Geophys. Res. Lett. 44, 11,106–11,114. areas resulted from the combined effect of the storm-induced physical Chacko, N., 2017. Chlorophyll bloom in response to tropical cyclone Hudhud in the Bay of Bengal: Bio-Argo subsurface observations. Deep-Sea Res. I Oceanogr. Res. Pap. 124, processes and biochemical processes. Owing to the limited data, how- 66–72. ever, it is not easy to separate quantitatively these two kinds of pro- Chen, Y., Tang, D., 2012. Eddy-feature phytoplankton bloom induced by a tropical cyclone cesses. More future work is needed to quantify the roles of physical in the South China Sea. Int. J. Remote Sens. 33, 7444–7457. and biochemical processes. Chen, J., Ni, X., Mao, Z., Wang, Y., Liang, L., Gong, F., 2012. Remote sensing and buoy based effect analysis of typhoon on hypoxia off the Changjiang (Yangtze) Estuary. In: Bostater, C.R., Mertikas, S.P., Neyt, X., Nichol, C., Cowley, D.C., Bruyant, J.P. (Eds.), Re- 5. Conclusion mote Sensing of the Ocean, Sea Ice, Coastal Waters, and Large Water Regions. 8532, p. 2012. D'Asaro, E.A., McNeil, C., 2013. Calibration and stability of oxygen sensors on autonomous This study investigated how tropical cyclones (TCs) over the Oxygen floats. J. Atmos. Ocean. Technol. 30, 1896–1906. Minimum Zone (OMZ) in the central Bay of Bengal (BoB) affected the Emanuel, K., 2001. Contribution of tropical cyclones to meridional heat transport by the dissolved oxygen (DO) concentrations in the subsurface waters during oceans. J. Geophys. Res.-Atmos. 106, 14771–14781. fi Feng, Y., DiMarco, S.F., Jackson, G.A., 2012. Relative role of wind forcing and riverine nu- ve TCs from 2013 to 2018 based on Argo and satellite data. The analy- trient input on the extent of hypoxia in the northern Gulf of Mexico. Geophys. Res. ses of observations reveal three types of DO responses to the “wind Lett. 39. 922 H. Xu et al. / Science of the Total Environment 659 (2019) 912–922

Girishkumar, M.S., Suprit, K., Chiranjivi, J., Bhaskar, T.V.S.U., Ravichandran, M., Shesu, R.V., Sarma, V.V.S.S., Krishna, M.S., Viswanadham, R., Rao, G.D., Rao, V.D., Sridevi, B., et al., 2013. et al., 2014. Observed oceanic response to tropical cyclone Jal from a moored buoy in Intensified oxygen minimum zone on the western shelf of Bay of Bengal during sum- the south-western Bay of Bengal. Ocean Dyn. 64, 325–335. mer monsoon: influence of river discharge. J. Oceanogr. 69, 45–55. Levitus, S., 1983. Climatological Atlas of the World Ocean. U.S. Dept. of Commerce, Na- Sarma, V.V.S.S., Rao, G.D., Viswanadham, R., Sherin, C.K., Salisbury, J., Omand, M.M., et al., tional Oceanic and Atmospheric Administration. 2016. Effects of freshwater stratification on nutrients, dissolved oxygen, and phyto- Lin, I.I., 2012. Typhoon-induced phytoplankton blooms and primary productivity increase plankton in the Bay of Bengal. Oceanography 29, 222–231. in the western North Pacific subtropical ocean. J. Geophys. Res. Oceans 117. Sarma, V.V.S.S., Jagadeesan, L., Dalabehera, H.B., Rao, D.N., Kumar, G.S., Durgadevi, D.S., et Lin, I.I., Liu, W.T., Wu, C.C., Chiang, J.C.H., Sui, C.H., 2003. Satellite observations of modula- al., 2018. Role of eddies on intensity of oxygen minimum zone in the Bay of Bengal. tion of surface winds by typhoon-induced upper ocean cooling. Geophys. Res. Lett. Cont. Shelf Res. 168, 48–53. 30. Stramma, L., Johnson, G.C., Sprintall, J., Mohrholz, V., 2008. Expanding oxygen-minimum Lin, J., Tang, D., Alpers, W., Wang, S., 2014. Response of dissolved oxygen and related ma- zones in the tropical oceans. Science 320, 655–658. rine ecological parameters to a tropical cyclone in the South China Sea. Adv. Space Stramma, L., Prince, E.D., Schmidtko, S., Luo, J., Hoolihan, J.P., Visbeck, M., et al., 2012. Ex- Res. 53, 1081–1091. pansion of oxygen minimum zones may reduce available habitat for tropical pelagic Liu, J., Russell, LM., Ruggeri, G., Takahama, S., Claflin, MS., Ziemann, PJ., et al., 2018. Re- fishes. Nat. Clim. Chang. 2, 33–37. gional Similarities and NOx-Related Increases in Biogenic Secondary Organic Aerosol Sun, L., Yang, Y.-J., Xian, T., Z-m, Lu, Fu, Y.-F., 2010. Strong enhancement of chlorophyll a in Summertime Southeastern United States. J. Geophys. Res. Atmospheres 123, concentration by a weak typhoon. Mar. Ecol. Prog. Ser. 404, 39–50. 10620–10636. Thadathil, P., Muraleedharan, P.M., Rao, R.R., Somayajulu, Y.K., Reddy, G.V., Revichandran, Madhu, N.V., Maheswaran, P.A., Jyothibabu, R., Sunil, V., Revichandran, C., C., 2007. Observed seasonal variability of barrier layer in the Bay of Bengal. J. Geophys. Balasubramanian, T., et al., 2002. Enhanced biological production off Chennai trig- Res. Oceans 112. gered by October 1999 super cyclone (Orissa). Curr. Sci. 82, 1472–1479. Thierry, V., Bittig, H., Argo-Team, 2016. Argo Quality Control Manual for Dissolved Oxygen Madhu, N.V., Jyothibabu, R., Maheswaran, P.A., Gerson, V.J., Gopalakrishnan, T.C., Nair, Concentration. K.K.C., 2006. Lack of seasonality in phytoplankton standing stock (chlorophyll a) Thushara, V., Vinayachandran, P.N., 2016. Formation of summer phytoplankton bloom in and production in the western Bay of Bengal. Cont. Shelf Res. 26, 1868–1883. the northwestern Bay of Bengal in a coupled physical-ecosystem model. J. Geophys. Mitra, A., Halder, P., Banerjee, K., 2011. Changes of selected hydrological parameters in Res. Oceans 121, 8535–8550. Hooghly estuary in response to a severe tropical cyclone (Aila). Indian J. Geo-Mar. Vidya, P.J., Das, S., Murali, M.R., 2017. Contrasting Chl-a responses to the tropical cyclones Sci. 40, 32–36. Thane and Phailin in the Bay of Bengal. J. Mar. Syst. 165, 103–114. Ni, X., Huang, D., Zeng, D., Zhang, T., Li, H., Chen, J., 2016. The impact of wind mixing on Vissa, N.K., Satyanarayana, A.N.V., Kumar, B.P., 2013. Response of upper ocean and impact the variation of bottom dissolved oxygen off the Changjiang Estuary during summer. of barrier layer on Sidr cyclone induced sea surface cooling. Ocean Sci. J. 48, 279–288. J. Mar. Syst. 154, 122–130. Wang, B., Chen, J., Jin, H., Li, H., Huang, D., Cai, W.-J., 2017. Diatom bloom-derived bottom Paulmier, A., Ruiz-Pino, D., 2009. Oxygen minimum zones (OMZs) in the modern ocean. water hypoxia off the Changjiang estuary, with and without typhoon influence. Prog. Oceanogr. 80, 113–128. Limnol. Oceanogr. 62, 1552–1569. Paulmier, A., Ruiz-Pino, D., Garcon, V., 2011. CO2 maximum in the oxygen minimum zone Warner, S.J., Becherer, J., Pujiana, K., Shroyer, E.L., Ravichandran, M., Thangaprakash, V.P., (OMZ). Biogeosciences 8, 239–252. et al., 2016. Monsoon mixing cycles in the Bay of Bengal: a year-long subsurface Pei, Y., Zhang, R., Chen, D., 2015. Upper ocean response to tropical cyclone wind forcing: a mixing record. Oceanography 29, 158–169. case study of typhoon Rammasun (2008). Sci. China Earth Sci. 58, 1623–1632. Xu, F., Yao, Y., Oey, L., Lin, Y., 2017. Impacts of pre-existing ocean cyclonic circulation on Prakash, S., Nair, T.M.B., Bhaskar, T.V.S.U., Prakash, P., Gilbert, D., 2012. Oxycline variability sea surface chlorophyll-a concentrations off northeastern Taiwan following episodic in the central Arabian Sea: an Argo-oxygen study. J. Sea Res. 71, 1–8. typhoon passages. J. Geophys. Res. Oceans 122, 6482–6497. Prakash, S., Prakash, P., Ravichandran, M., 2013. Can oxycline depth be estimated using Ye, H., Sheng, J., Tang, D., Siswanto, E., Kalhoro, M.A., Sui, Y., 2017. Storm-induced changes sea level anomaly (SLA) in the northern Indian Ocean? Remote Sens. Lett. 4, in pCO2 at the sea surface over the northern South China Sea during Typhoon Wutip. 1097–1106. J. Geophys. Res. Oceans 122, 4761–4778. Price, J.F., 1981. Upper ocean response to a hurricane. J. Phys. Oceanogr. 11, 153–175. Zhang, H., Chen, D., Zhou, L., Liu, X., Ding, T., Zhou, B., 2016. Upper ocean response to ty- Sardessai, S., Ramaiah, N., Kumar, S.P., de Sousa, S.N., 2007. Influence of environmental phoon Kalmaegi (2014). J. Geophys. Res. Oceans 121, 6520–6535. forcings on the seasonality of dissolved oxygen and nutrients in the Bay of Bengal. Zhao, H., Tang, D., Wang, Y., 2008. Comparison of phytoplankton blooms triggered by two J. Mar. Res. 65, 301–316. typhoons with different intensities and translation speeds in the South China Sea. Sarma, V.V.S.S., Udaya Bhaskar, T.V.S., 2018. Ventilation of oxygento oxygen minimum Mar. Ecol. Prog. Ser. 365, 57–65. zone due to anticyclonic eddies in the Bay of Bengal. J. Geophys. Res. Biogeosci. 123, 2145–2153.