Freshening of the upper ocean in the South China Sea since the early 1990s

Deep–Sea Research I 118 (2016) 20–29

Contents lists available at ScienceDirect

Deep–Sea Research I journal homepage: www.elsevier.com/locate/dsri

crossmark

Freshening of the upper ocean in the South China Sea since the early 1990s a

a,⁎

b,c

b

b

d

Feng Nan , Fei Yu , Huijie Xue , Lili Zeng , Dongxiao Wang , Shilun Yang , Kim-Cuong Nguyene a

Institute of Oceanology, Chinese Academy of Sciences, Qingdao, China State Key Laboratory of Tropical Oceanography, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, China School of Marine Sciences, University of Maine, Orono, ME, USA d State Key Laboratory of Estuarine and Coastal Research, East China Normal University, Shanghai, China e Department of Marine Science and Technology, VNU University of Science, Hanoi, Vietnam b c

A R T I C L E I N F O

A BS T RAC T

Keywords: South China Sea Salinity Water mass Freshening Kuroshio intrusion

Ocean salinity is often used as a dynamical tracer for investigating the Kuroshio intrusion into the South China Sea (SCS). In this study, we found that the upper-ocean water in the SCS had a freshening trend since the early 1990s. Salinity in the upper 100 m of the SCS (SSCS) decreased by ~0.24 psu in the period 1993–2012, yielding a negative trend of −0.012 psu yr−1. The maximum freshening occurred in the surface layer west of the Luzon Strait, and freshening gradually lessened from northeast to southwest and with depth, indicating the important influence of the Kuroshio intrusion. Quantitative analysis of salinity budget from the surface to 100 m depth in the SCS suggests that the weakened Kuroshio intrusion is the leading factor controlling the SSCS freshening, while the increased air-sea freshwater flux plays a minor role. Based on GODAS (Global Ocean Data Assimilation System) model output, the Luzon Strait transport (LST) in the upper 100 m decreased in a negative trend of −0.12 Sv yr−1 (1 Sv=106 m3 s−1) from 1993 to 2012, corresponding to a freshening trend of the SSCS at a rate of −0.011 psu yr−1. Both the LST and SSCS changes are closely related to the Pacific Decadal Oscillation (PDO). Our findings demonstrate that the strength of the Kuroshio intrusion into the SCS weakened markedly since the PDO phase shifted in 1990s, which resulted in the pronounced freshening of the SCS water.

1. Introduction Ocean salinity is one of the most fundamental parameters in physical oceanography, and it plays an important role in modulating ocean and climate variability (Katsura et al., 2013). Salinity is a key variable in computing the geostrophic circulation (e.g., Suga et al., 2000; Qiu and Chen, 2012). Because of its better conservative property compared with temperature, salinity is often used as a dynamical tracer for circulation studies (e.g., Chen and Huang, 1996; Yan et al., 2013; Nan et al., 2013). Salinity also has thermodynamic importance. The stratification of salinity can affect the mixed layer depth, and advection of salinity anomalies from subtropical regions can influence tropical climate changes (Lukas and Lindstrom, 1991; Lukas, 2001). Salinity also has climatologic importance. Any change in the hydrological cycle can be reflected in the ocean salinity field (Williams et al., 2007). Because of its importance, salinity changes have been reported on both global and regional scales. Durack and Wijffels (2010) noted that, at the global scale, salty regions get saltier and fresh regions get fresher from 1950 to 2008, which is consistent with an amplification of the

⁎

global hydrological cycle. In the northwestern Pacific, Suga et al. (2000) found that salinity of the North Pacific Tropical Water increased remarkably associated with the mid-1970s regime shift. Lukas (2001) reported a pronounced freshening (−0.15 psu) of the upper ocean in the Pacific Subtropical Gyre from 1991 to 1997. Sugimoto et al. (2013) found that the Subtropical Mode Water in the north Pacific, characterized by low potential vorticity, freshened markedly in 2009 and 2010. The South China Sea (SCS) is the largest semi-enclosed marginal sea in the north Pacific (Fig. 1). The general circulation pattern in the SCS is driven primarily by the East Asian monsoon and significantly influenced by the Kuroshio intrusion through the Luzon Strait (Qu, 2000; Qu et al., 2004; Wang et al., 2013). In past decades, much work has been done on the temperature changes, circulation, eddy activities, and Kuroshio intrusion etc. in the SCS (see Hu et al., 2000; Nan et al., 2014; and references therein). However, studies on salinity changes in the SCS are limited. Though Huang et al. (2015) investigated the response of salinity variation in the Taiwan Strait to the interannual variation of the Kuroshio intrusion using in situ measurements, the decadal variation and long-term trend of the salinity in the interior SCS

Corresponding author. E-mail address: [email protected] (F. Yu).

http://dx.doi.org/10.1016/j.dsr.2016.10.010 Received 5 November 2015; Received in revised form 20 October 2016; Accepted 21 October 2016 Available online 25 October 2016 0967-0637/ © 2016 Elsevier Ltd. All rights reserved.

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

Fig. 1. Mean Absolute Dynamic Topography (ADT, units: cm) in the SCS derived from the 20-year satellite altimeter data from 1993 to 2012.

have not been reported. Hsin (2015) related the multi-decadal change of salinity in the northern SCS to the multi-decadal changes of Kuroshio intrusion in the Luzon Strait. Nan et al. (2013) revealed that the Kuroshio intrusion into the SCS had a weakening trend from 1990s to 2000s, which resulted in a freshening in the northeastern SCS. However, the area and depth influenced by the weakened Kuroshio intrusion are not clear. On the other hand, using gridded Argo data, Yan et al. (2013) noted a sustained freshening of subsurface water in the Kuroshio east of the Luzon Strait during the period 2003–2011. Based on long-term repeated observations, Nan et al. (2015) demonstrated that both the surface and subsurface salinity in the northwest Pacific Subtropical Gyre including the Kuroshio had a freshening trend during the period 1987–2012. The effects of the long-term Kuroshio freshening on the salinity changes in the SCS remain to be quantified. Zeng et al. (2014) found that the upper-ocean salinity in the northern SCS decreased markedly (~0.4 psu) in 2012 because of abundant local freshwater flux and reduction of the Kuroshio intrusion. They also noted that the river discharge associated with abnormal precipitation played a leading role in the salinity near the Mekong River mouth. Wu et al. (2010) revealed a distinct increase in rainfall over southern China since the early 1990s, but the influences of interannual to decadal changes of the river discharge and freshwater flux on the salinity changes in the SCS remains unclear. The present study has three objectives: 1) to investigate the longterm changes of salinity in the SCS based on reanalysis data, 2) to clarify what controls the salinity variability (e.g., river discharge, airsea freshwater flux, and Kuroshio intrusion), and 3) to quantify the influence of the Kuroshio intrusion variation on the salinity changes in the SCS. The rest of the paper is organized as follows. Section 2 describes the data and methodology. Section 3 presents the results on the long-term salinity changes in the SCS. Section 4 discusses the possible driving factors in the salinity changes. Finally, Section 5 summarizes the main findings.

Fig. 2. Distribution of salinity profiles (a) binned in 1°×1°cells and yearly numbers (b) in the SCS ( < 121°E) derived from WOD13.

Program (GTSPP), the global temperature-salinity in the tropical Pacific from the IRD (L′Institut de recherche pour le development, France), the Centennial in situ Observation Based Estimates (COBE) sea surface temperature, and the Argo profiling buoy data since the early 2000s (Moon et al., 2013; Chen and Tung, 2014). The temperature-salinity profiles without quality check have been deleted. This global 1°×1° dataset at 24 levels in the upper 1500 m, called Ishii data in this study, is freely available at http://rda.ucar.edu/datasets/ds285. 3. The dataset covers the period from 1945 to 2012, and the data from 1980 to 2012 are used in this study. To check the reliability of the gridded dataset, the number of salinity profiles and their distributions are plotted in Fig. 2 based on WOD13 (http://www.nodc.noaa.gov/ OC5/WOD13/). There are a total of 14,594 verified salinity profiles (5912, 2249, and 6433 in the 1980s, 1990s, and 2000s, respectively) in the SCS (Fig. 2a). The profile numbers are larger than 200 in most years except for 1994–1996 and 2003–2006 (Fig. 2b). There is a temporal-spatial mismatch of yearly salinity profiles, which may cause interpolation error for Ishii data. For example, Fig. 2a shows that salinity profiles in the southeastern SCS are sparse. The reanalysis dataset has been widely used in several earlier studies (e.g., Carton and Santorelli, 2008; Moon et al., 2013; Chen and Tung, 2014; Hsin, 2015). Following Nan et al. (2013), surface geostrophic velocity was calculated from satellite altimeter data produced by the French Archiving, Validation, and Interpolation of Satellite Oceanographic (AVISO) as follows:

2. Data and methodology 2.1. Observational data

u g = (ug, vg ) =

To investigate the salinity changes in the SCS, we use yearly, objectively-analyzed subsurface temperature and salinity compiled by Ishii and Kimoto, 2009. The analysis is based on the World Ocean Database 2013 (WOD13), Global Temperature-Salinity Profile

g ⎛ ∂η ∂η ⎞ ⎟, ⎜− , f ⎝ ∂y ∂x ⎠

(1)

where ug (vg) is zonal (meridional) component of geostrophic flow, g is gravitational acceleration, η is the Absolute Dynamic Topography 21

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

Mekong River discharge data at Stung Treng station in Cambodia (most downstream observation station with high quality and long-term discharge data) was provided by the Mekong River Commission (http://pmfm.mrcmekong.org/). Yearly discharge data of the Pearl River was derived from the Sediment Bulletins of China Rivers issued by the Ministry of Water Resources of China (MWRC). The gauging

(ADT) and equal to mean dynamic topography plus sea surface height anomaly, and f is the Coriolis parameter. The merged ADT dataset is from the combination of Jason, Topex/Poseidon, Envisat, GFO, ERS and Geosat altimeters. Weekly average frames are interpolated onto a global grid of 1/4° resolution from October 1992 to the present. As in Nan et al. (2013), the data in areas where the water depth is less than 200 m were excluded to minimize the aliasing effects due to tides and internal waves. 2.2. The GODAS model The model output from 1980 to 2012 used in the study was based on the Global Ocean Data Assimilation System (GODAS) (http://www. esrl.noaa.gov/psd/data/gridded/data.godas.html). The model domain extends from 75°S to 65°N and has a resolution of 1°×1/3°. The model has 40 levels with a 10 m resolution in the upper 200 m. The model assimilates temperature-salinity profiles from expendable bathythermographs (XBTs), GTSPP, Tropical Atmosphere Ocean (TAO), WOD, Argo floats, currents from Triangle Trans-Ocean Buoy Network (TRITON) and Prediction and Research Moored Array in the Tropical Atlantic (PIRATA), and sea surface height (SSH) from satellite altimeter data (Behringer et al., 1998). More detailed description about this model can be seen in Huang et al. (2010). The model results have been widely used in the western Pacific, such as to show temperature changes in the equatorial Pacific, heat content changes, North Equatorial Current (NEC) variability, etc. (e.g., Carton and Santorelli, 2008; Huang et al., 2010; Hu et al., 2013; Zhai et al., 2014). From Sections 4 and 5, it can also be seen that the GODAS model can well reproduce the Luzon Strait Transport (LST), the NEC, SSH and geostrophic current trends of our interest. 2.3. Salinity budget in the upper 100 m of the SCS To investigate the mechanism of the salinity variation in the SCS, the equation relating the salinity changes to the possible forcing factors is first presented. If small-scale mixing processes are ignored, the salinity balance in the upper 100 m of the SCS is governed by the simplified equation:

∂S R×S F×S T × ΔS1 w ΔS =− + + − e 2, ∂t VSCS VSCS VSCS 100

(2)

where S is the mean SSCS in the 100 m of the SCS (105–121°E, 5– 25°N), t is time, R is total river discharge, F is air-sea freshwater flux (Evaporation minus Precipitation), T is the advection transport into the SCS above 100 m, ΔS1 is the salinity difference between the western Pacific (121–125°E, 15–25°N) and the SCS (the former minus the latter), we is vertical velocity from GODAS model output, ΔS2 is the salinity difference between 100 m and 120 m in the SCS (the former minus the latter), and VSCS is water volume in the SCS above 100 m. Taking the SCS from the surface to 100 m (deeper than the maximum mixed layer depth even in winter) as a box, the mixing process plays a minor role in the salinity changes of the total box. Here, the mixing process is ignored. Terms on the left hand side in Eq. (2) represent the salinity tendency, and those on the right hand side are the river discharge term, air-sea freshwater flux, advection term, and entrainment term, respectively. Note that the advection term only contains the LST based on GODAS model output. The LST was calculated by the difference between inflow and outflow flux across the 121°E meridian in the Luzon Strait. The water exchange via the other straits (Taiwan Strait, Mindoro Strait, and Karimata Strait) is negative (out of the SCS), according to Qu et al. (2005) and Song (2006), and does not contribute the interannual SCS salinity budget. From the analyses of Section 4, it can also be seen that the salinity changes in the SCS are mainly induced by the Kuroshio intrusion variability through the Luzon Strait. The contributions of the three largest rivers (Mekong River, Pearl River, and Red River) to the SCS salinity budget were calculated. The

Fig. 3. Salinity (units: psu) distributions at 0 m, 150 m, and 450 m depths derived from Ishii data in the SCS. Contour intervals are 0.1 psu, 0.04 psu, and 0.02 psu for (a), (b), and (c), respectively. The solid triangles labeled A–D in (c) represent four selected points in the Luzon Strait (122°E, 20°N), northeastern SCS (118°E, 20°N), central SCS (115°E, 15°N), and southern SCS (113°E, 10°N) for analysis of vertical salinity in Fig. 4.

22

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

Fig. 6. Time series of the winter (Dec.–Feb.; blue), summer (Jun.–Aug.; red), and total mean (black) salinity (units: psu) average above 100 m in the SCS. Dotted lines represent the linear trends before/after 1993 that best fit the data. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 4. Vertical salinity (units: psu) of four profiles (A–D) as indicated in Fig. 3c.

3. Long-term salinity changes in the SCS Before examining salinity changes, it is necessary to know the mean distribution of salinity in the SCS. According to the previous results (Qu et al., 2000; Liang et al., 2008; Nan et al., 2013), the T-S curves of the SCS water and the northwestern Pacific water both appear as an approximately mirrored ‘S’. The maximum and minimum salinity levels are located at ~150 m and ~450 m depth in the SCS representing the subsurface and intermediate waters, respectively. Thus, the 0 m, 150 m, and 450 m depths were selected to represent surface water, subsurface water, and intermediate water, respectively (Fig. 3). We compared the climatological salinity distributions based on Ishii data with those of WOA13. Unsurprisingly, the spatial distributions of salinity derived from the two datasets are almost alike (figure not shown). The salinity distribution in Fig. 3 shows a clear division between the SCS and the Pacific at ~121°E, i.e., the surface and subsurface salinity were lower while the intermediate salinity was higher in the SCS (see also Fig. 4). Subsurface salinity maxima and intermediate salinity minima are good tracers to delineate the SCS and Kuroshio waters (Yu et al., 2008; Nan et al., 2011; Zeng et al., 2014). The surface and subsurface salinity decreased towards the southwest in the SCS, while the intermediate salinity increased gradually (Fig. 4). To investigate the long-term changes of the salinity in the SCS, we plot the interannual variations of the average salinity in the SCS (105– 121°E, 5–25°N) at different depths from surface to 150 m in Fig. 5. It can be seen that salinity at different depths above 150 m varies almost synchronously and has notable interannual variation. Surface salinity decreased markedly (larger than 0.2 psu) in 2001, 2008, and 2012. Based on in situ observations, Zeng et al. (2014) found that the upperocean salinity in the northern SCS decreased ~0.4 psu in 2012, which was revisited in our result. The large salinity anomalies in 2001 and 2008 have not been reported. The most striking feature of Fig. 5 is that the mean salinity above ~100 m has a notable freshening trend since ~1993. The freshening trend becomes indiscernible below 100 m. Fig. 6 shows the interannual variations of the average salinity in the upper 100 m of the SCS and its long-term trends before and after 1993. Salinity average in the SCS above 100 m increased slightly (~0.006 psu) from 1980 to 1992, while it decreased markedly (~0.24 psu) from 1993 to 2012, with a freshening trend of −0.012 psu yr−1 (significant at the 95% confidence level). Note that the SCS circulation is dominated by the reversed monsoon systems. Winds typically blow strongly from the northeast during boreal winter and from the southwest during

Fig. 5. Time series of the average salinity (units: psu) in the SCS (color shaded region west of 121°E in Fig. 3) at different indicated depths from surface to 150 m based on Ishii data. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

stations used are all ranked as first-class stations in China and are operated following the national criterion established by the MWRC (Wu et al., 2012). Monthly average discharge data of the Red River at Son Tay station in Vietnam during 1956–2011 were provided by the National Center for Hydro-Meteorological Forecasting (NCHMF) of Vietnam. The Red River discharge in 2012 was calculated using the mean value of 2010 and 2011. The monthly precipitation on 1°grid during 1980–2012 is from the National Centers for Environmental Prediction's Climate Prediction Center (CPC) (Chen et al., 2002). Evaporation field is obtained from the Objectively Analyzed air-sea Fluxes (OAFlux) (Yu and Weller, 2007). Monthly mean net freshwater flux is calculated by OAFlux evaporation minus CPC precipitation. 23

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

and decreased salinity in the northwestern Pacific will all cause freshening in the SCS. Each term on the right side of the Eq. (2) can be calculated. The mean, standard deviation, and trends of each term on the right side of Eq. (2) are shown in Table 1. In this section, we will quantitatively discuss the contributions of all the aforementioned factors on the freshening in the SCS.

boreal summer (Qu, 2000). To minimize the aliasing associated with the seasonal contribution on the salinity freshening trends, interannual variations of salinity in winter and summer are also calculated in Fig. 6. The freshening trend in winter (−0.013 psu yr−1; above the 95% significance level) is a little larger than that in summer (−0.011 psu yr−1; above the 95% significance level). They all correlate well with the trend of the annual mean, suggesting that the interannual signal and long-term trend were consistent throughout the year. From Figs. 5 and 6, it can be seen that mean salinity in the SCS had a sustained freshening trend since the early 1990s. However, the spatial distribution of the salinity trends is unknown. Fig. 7 shows spatial distribution of the long-term salinity trends at depths of 0 m, 50 m, and 100 m before and after 1993. From 1980 to 1992, salinity has a small positive trend (~0.005 psu yr−1) at 0 m and a small negative trend in most areas at 100 m (~−0.005 psu yr−1). The salinity trend at 50 m is ambiguous. Note that the patterns of salinity trend at 0 m, 50 m, and 100 m depths are incoherent spatially over the period of 1980–1992. However, the spatial patterns of salinity trend are coherent from 1993 to 2012 except for some quantitative differences. Salinity at depths of 0 m, 50 m, and 100 m (except the southern tip) shows pronounced freshening trends. The maximum negative salinity trend of −0.02 psu yr−1 (above the 95% significance level) occurs in the northeastern SCS west of the Luzon Strait in the surface layer, i.e., surface salinity in the northeastern SCS decreased by 0.4 psu from 1993 to 2012. The freshening trend gradually became smaller from northeast to southwest in the SCS and decreased with depth. Fig. 8 shows vertical distribution of the mean long-term salinity trend from 1992 to 2012 in the northern SCS, central SCS, southern SCS, the total SCS, and the western Pacific (Fig. 7). It can be seen that the mean negative salinity trend can extend to 400 m in the northern SCS, while it is above 75 m in the southern SCS. The mean negative salinity trend in the central SCS can reach 125 m, which is similar to that of the total mean in the SCS. Salinity in the western Pacific also has a small negative trend lower than −0.005 psu yr−1 above 500 m. Nan et al. (2015) found that both surface and subsurface salinity (above ~600 m) in the Northwest Pacific Subtropical Gyre had a sustained freshening trend at ~−0.004 psu yr−1 during the period 1987–2012, which is consistent with this result. It is worth emphasizing that the freshening signal in the western Pacific can be advected into the SCS by the Kuroshio. However, the freshening trend in the northern SCS is much larger than that in the western Pacific (Kuroshio upstream), suggesting that the salinity freshening in the SCS is not controlled by the Kuroshio upstream freshening. We will return to this point in Section 4.3. From the above analyses based on a gridded reanalysis dataset, large and spatially coherent linear trends in salinity are found in the SCS. Since the early 1990s, the upper-ocean water in the SCS has had a marked freshening trend. Salinity in the upper 100 m of the SCS decreased by ~0.24 psu during the period 1993–2012, with a freshening trend of −0.012 psu yr−1. Note that there exist regional differences of the salinity freshening magnitude. The maximum salinity freshening trend (−0.02 psu yr−1) occurred west of the Luzon Strait and gradually became smaller from northeast to southwest in the SCS. Quantitative analyses of the dynamics on the salinity changes in the SCS will be conducted in the ensuing section.

4.1. River discharge The Mekong River, Pearl River, and Red River are the three largest rivers that debouch into the SCS, with mean volume transports of 0.013 Sv, 0.003, and 0.003 Sv, respectively (Fig. 9a). All rivers’ discharge has a decreasing trend since the early 1990s, as the rapid economic development at riparian countries has led to massive plans for hydropower construction (Lauri et al., 2012; Wu et al., 2012). The combined volume transport of the three rivers is 0.02 Sv on average. It has a decreasing trend of −0.19×10−3 Sv yr−1 during 1993–2012. From Table 1, river discharge causes a mean negative salinity anomaly of −0.08 psu. Standard deviation of the river discharge term is much smaller than those of the other terms. Decreased river discharge in 1993–2012 could cause an increasing trend of the salinity average above 100 in the SCS (SSCS) at 0.08×10−2 psu yr−1, which does not contribute to the SSCS freshening (Table 1). 4.2. Air-sea freshwater flux The mean air-sea freshwater flux is approximately 0.042 Sv, which is much larger than the total river discharge (Fig. 9b). The mean freshwater flux is positive in the northern SCS, while it is negative in the southern SCS (figure not shown). However, the basin-average airsea freshwater flux term is negative (−0.18 psu; Table 1) because average evaporation is smaller than average precipitation in the SCS (Fig. 9b). It also has a decreasing trend of −1.5×10−3 Sv yr−1 during the period 1980–1992, while it has an increasing trend of 0.44×10−3 Sv yr−1 during the period 1993–2012. Increased air-sea freshwater flux since 1993 causes a freshening trend of the SSCS at −0.19×10−2 psu yr−1, contributing approximately 15% of the total SSCS freshening (Table 1), which is the secondary factor. 4.3. The Kuroshio intrusion The advection term is positive (0.21 psu) because of saltier water input from the western Pacific, as shown in Fig. 3. Because subsurface water is more saline than the surface water, entrainment term is also positive (0.1 psu). The total salinity variation due to the four terms (river discharge term, air-sea freshwater flux term, advection term, and entrainment term) is −0.05 psu, which is approximately equal to zero (Table 1). The magnitude of the LST often represents the strength of the Kuroshio intrusion in previous studies (Qu et al., 2004; Nan et al., 2014). Compared in Fig. 9c, the interannual variation and long-term trend of the LST during the period 1993–2012 derived from GODAS model and satellite data are similar, indicating the credibility of the GODAS model result. Based on GODAS output, the mean LST above 100 m is 2.32 Sv, which is approximately 30 times the total transport of river discharge and air-sea freshwater flux. The LST has a small increasing trend of 0.004 Sv yr−1 during the period 1980–1992, while it has a notable decreasing trend of −0.12 Sv yr−1 during the period 1993–2012 (Fig. 9c). As revealed in Table 1 and Fig. 10, the interannual variation of the advection term is the largest. Calculated salinity trends based on Eq. (2) are consistent with the observed salinity trends. The correlation coefficient of yearly variations between the calculated and observed salinity anomalies is 0.69 above the 95% significance level (Fig. 10e). The decreased advection term since 1993 causes a freshening trend of the SSCS at −0.010 psu yr−1, contributing approximately 85% of the total SSCS freshening (see Table 1), which is the leading factor. It is

4. Forcing of the long-term salinity changes To interpret the observed salinity changes in the SCS, a simplified salinity budget equation is used (see Eq. (2)). From Fig. 8, the maximum freshening occurs above ~100 m. Thus, we calculated the salinity budget above 100 m in the SCS (west of 121°E). It is known that surface layer salinity in the SCS is mainly controlled by river discharge, local freshwater flux, and salt transport by large-scale ocean circulation (Zeng et al., 2014). It can be seen from Fig. 9 that the freshwater flux increased since the early 1990s, while the river discharge and the LST all decreased. Increased air-sea freshwater flux 24

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

Fig. 7. Salinity trends (units: psu yr−1) at the 0 m, 50 m, and 100 m depths from 1980 to 1992 (left) and from 1993 to 2012 (right).

25

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

Fig. 8. Vertical distribution of the mean long-term salinity trend (units: psu yr−1) from 1993 to 2012 in the different regions.

noteworthy that the advection term contains the salinity difference between the western Pacific and the SCS, and the advection transport (see Eq. (2)), i.e., both the weakening Kuroshio intrusion and freshening of the Kuroshio upstream, can contribute to the SSCS freshening. The Kuroshio east of Luzon Island also has a freshening trend. To estimate the contributions of the LST changes and salinity changes in the Kuroshio upstream, three calculations are made using annually varying T and ΔS, annually varying T and climatological ΔS, and climatological T and annually varying ΔS (Fig. 11). The LST changes contribute ~95% of the freshening caused by the advection term, which is the dominant factor. The effect of salinity changes in the Kuroshio upstream is limited. From Figs. 7 and 8, one can also see that the freshening trend in the Kuroshio in the Luzon Strait is much smaller than that in the northern SCS. This means that the SSCS freshening cannot be caused solely by the Kuroshio freshening.

5. Summary and discussion This paper focuses on the salinity changes in the SCS. Specifically, using gridded reanalysis data, it is found that the upper layer salinity in the SCS has had a freshening trend since the early 1990 s. From 1993– 2012, mean salinity in the upper 100 m of the SCS decreased by ~0.24 psu, i.e., a freshening trend of −0.012 psu yr−1. The freshening differed by region. The maximum salinity freshening trend occurred west of the Luzon Strait, and the trend decreased gradually with depth and from northeast to southwest in the SCS. The mean negative salinity trend extended to the 400 m depth in the northern SCS but was limited to 0– 75 m in the southern SCS. The freshening in the SCS appears to have resulted from a combination of factors, including increased air-sea freshwater flux, decreased salinity in the northwestern Pacific, and weakened Kuroshio intrusion. To explain the freshening in the SCS, the salinity budget above the 100 m depth was analyzed. Contributions of river discharge, air-sea freshwater flux term, and the Kuroshio intrusion to long-term changes of the salinity in the SCS were analyzed quantitatively. The results show that weakening Kuroshio intrusion is the leading factor controlling the freshening, while air-sea freshwater flux plays a minor role. Based on GODAS model output, the LST of the upper 100 m decreased from 1993 to 2010 at a negative trend of −0.12 Sv yr−1, corresponding to a freshening trend of SSCS at −0.011 psu yr−1, which contributed to 85% of the total SSCS freshening. It is unclear why the Kuroshio intrusion into the SCS has had a weakening trend since the early 1990s, as shown in Fig. 9c. Although Nan et al. (2013) suggested that the decreasing SSH difference between the western Pacific subtropical gyre and the SCS was the leading factor for the weakened Kuroshio intrusion, further efforts are needed to

Fig. 9. (a) Time series of the Mekong River, Pearl River, Red River, and total discharges (units: Sv). (b) Time series of basin-average net freshwater flux (units: Sv) derived from CPC precipitation data and OAFlux evaporation data. (c) Time series of the Luzon Strait Transport (LST; units: Sv) above 100 m derived from GODAS outputs (black). The geostrophic LST (blue) are calculated from the surface geostrophic current based on satellite ADT and multiplied by 100 m. The positive value means westward transport (into the SCS). Dotted lines in (a), (b), and (c) represent the linear trends before/after 1993 that best fit the data. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

clarify this. In Fig. 12a, we calculated the linear trends of the SSH and corresponding surface geostrophic currents derived from satellite altimeter data from 1993 to 2012. It can be seen that the SSH trend in the subtropical gyre was opposite to that in the tropical gyre, i.e.,

26

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

Table 1 Mean, standard deviation, and trends of integral salinity anomaly in one year (Δt) caused by each term on the right side of Eq. (2) and observed salinity anomaly above 100 m in the SCS during the period 1980–2012. Mean (psu)

River discharge term −R×S/VSCS×Δt Air-sea freshwater flux term F×S/VSCS×Δt Advection term T×ΔS1/ VSCS×Δt Entrainment term −weΔS2/100×Δt Total Observed salinity anomaly

Standard deviation (psu)

Trend (10−2 psu/yr) 1980– 1992

1993– 2012

−0.08

0.02

0.07

0.08

−0.18

0.04

0.34

−0.19

0.21

0.10

0.18

−1.10

0.10

0.03

0.05

0.11

−0.05 0

0.11 0.11

0.64 0.39

−1.11 −1.21

Fig. 11. Time series of advection term during the period 1980–2012. These three calculations are made to estimate the contributions of the LST changes and salinity changes in the Kuroshio upstream using annually varying T and ΔS (black), annually varying T and climatological ΔS (blue), and climatological T and annually varying ΔS (red). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

(2012; see their Fig. 7). Dynamics on the opposite trends of the NEC south and north of 12.5°N remain to be understood; this topic will be pursued in our future studies. To elucidate what controls the weakening Kuroshio in the Luzon Strait, we calculated the correlation for the upper 100 m average annual mean zonal velocity between the Luzon Strait and all locations in the western north Pacific based on GODAS model output (Fig. 13). The NECN variation correlates well with that of the Kuroshio variation in the Luzon Strait. The correlation coefficient between the interannual variation of the upper 100 m LST and the upper 100 m westward NECN transport average over 125–135°E and 12.5–20°N is 0.80 and above the 95% significance level (Fig. 14), suggesting that the LST changes depend on the NECN changes. The long-term trend of the LST is almost identical with that of the NECN transport. According to Fig. 14, the NECN transport also had a marked decreasing trend since the early 1990s. In the period 1993–2012, the NECN transport above 100 m decreased by ~2.6 Sv at a negative trend of −0.13 Sv yr−1, which is in close agreement with the negative trend of the LST. Hence, weakening of the NECN may determine the decreasing of the Kuroshio intrusion into the SCS. The NECN transport, the LST and SSCS changes are all closely related to the Pacific Decadal Oscillation (PDO; Fig. 15). The zero-leg correlation coefficients between the PDO index and yearly NEC transport, the LST, and the SSCS are 0.69, 0.74, and 0.82 (above the 95% significance level), respectively. Based on model results and

Fig. 10. Integral salinity anomaly (units: psu) in one year caused by each term on the right side of Eq. (2) during the period 1980–2012. (a) River discharge term, (b) air-sea freshwater flux (E-P) term, (c) advection term, (d) entrainment term, and (e) sum of the former terms. The blue line in (e) represents the observed salinity anomaly average in the SCS (units: psu). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

there existed a dipole-like pattern for the SSH trend in the northwestern Pacific. Increasing SSH in the tropical gyre and decreasing SSH in the subtropical gyre produced eastward flow anomalies between 12.5°N and 20°N. Since the GODAS assimilated the satellite-altimeter SSH data, the pattern of the modeled SSH linear trend is similar to that of the satellite result (Fig. 12b). The linear trend of mean currents above 100 m based on GODAS output in 1993–2012 shows a decreasing trend of the currents in the Luzon Strait and the NEC between 12.5°N and 20°N (NECN, Fig. 12b) and an enhancement of the NEC south of 12.5°N. This result is consistent with that of Qiu and Chen 27

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

Fig. 13. Correlation of the upper 100 m average annual mean zonal velocity between the Luzon Strait (121° E) and other locations derived from GODAS model output. Only positive correlation coefficients above the 95% significance level are shown.

Fig. 12. (a) Linear trends of the ADT (units: cm yr−1) and corresponding surface geostrophic currents (vectors; units: cm s−1 yr−1) derived from satellite altimeter data during the period 1993–2012 (smoothed by a 1°×1°box running mean filter). (b) Linear trends of the SSH (units: cm yr−1) and the mean currents (vectors; units: cm s−1 yr−1) above 100 m based on GODAS output from 1993 to 2012.

Fig. 14. Time series of the LST (units: Sv; black) and westward NECN transport average (units: Sv; blue) over 125–135°E, and 12.5–20°N above 100 m. Dotted lines represent the linear trends before/after 1993 that best fit the data. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

In conclusion, associated with the PDO phase shift from warm to cool since the 1990s, the strength of the Kuroshio intrusion into the SCS has weakened markedly, thereby resulting in the pronounced freshening of the SCS water. It is worth emphasizing that the salinity changes reported in this study may have a significant impact on the ocean stratification and density structure and thus may influence the ocean circulation in the SCS. Further studies are necessary to gain a better understanding of salinity-driven impacts on the SCS climate change. The results of this paper are based on a gridded Ishii dataset. Temporal-spatial mismatch of yearly salinity profiles may cause interpolation error, which may lead to considerable biases in the resultant salinity anomaly in some years. Regardless, the freshening trend in the SCS, which represents our best attempt to interpret the available data, is likely real for similar trends in the western north Pacific (Yan et al., 2013; Nan et al., 2015) as well as in the northeastern SCS (Nan et al., 2013). In addition, our analyses on the dynamics of weakening Kuroshio intrusion into the SCS are relatively preliminary. Carefully designed numerical model experiments may provide more insights.

theoretical analysis of the “island rule”, Yu and Qu (2013) revealed that decadal variation of the upper-layer (0–745 m) LST correlates well with the PDO. During the warm phase of the PDO, the Aleutian low and its related positive wind stress curl anomalies migrate southward, reducing the trade wind in the tropical North Pacific. Thus, the NEC bifurcation shifts northward and causes an enhancement of the Kuroshio intrusion into the SCS. The situation is reversed during the cool phase of the PDO. From Fig. 15, it can be seen that the PDO has changed phase from warm to cool since the 1990s (also indicated by Qiu et al., 2015). Qiu and Chen (2012) found that the NEC bifurcation shifted southward after 1993 because of strengthened easterly trade winds (the lower branch of the atmospheric Walker circulation). As noted by many previous studies (Qu et al., 2004; Yu and Qu, 2013; Yang et al., 2013; Hu et al., 2015), southward migrated NEC bifurcation tends to weaken the Kuroshio intrusion into the SCS or the LST. From Fig. 14, it can also be seen that the long-term trends for both the NECN transport and the LST are opposite before and after 1993. 28

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

broad-scale warming. J. Clim. 23b:date>43424362. http://dx.doi.org/10.1175/ 2010JCLI3377.1.. n, 2015 e:given-name>Y.-C., or>17921808.. u, 2015 D., orHu, b:se">. Pacific western boundary currents and their roles in climate. Nature 522b:date>299308. http://dx.doi.org/10.1038/nature14504.. Hu et al., 2000 J., orHu, , H.Kawamura, , H.Hongi, , Y., orQi, b:se">Hu . A review on the currents in the South China Sea: seasonal circulation, South China Sea Warm Current and Kuroshio intrusion. J. Oceanogr. 56b:date>607624.. Hu et al., 2013 e:given-name>Z.-Z., or>Hu . Leading modes of the upper-ocean temperature interannual variability along the equatorial Atlantic Ocean in NCEP GODAS. J. Clim. 26b:date>46494663.. Huang et al., 2010 B.Huang, , Y., orXue, , X.Zhang, , A.Kumar, , ce:givenname>M.J.McPhaden, b:se">Huang . The NCEP GODAS ocean analysis of the tropical Pacific mixed layer heat budget on seasonal to interannual time scales. J. Clim. 23b:date>49014925.. Huang et al., 2015 e:given-name>T.-H.Huang, , e:given-name>C.-T., or>W.-Z.Zhang, , e:given-name>X.-F.Zhuang, b:se">Huang . Varying intensity of Kuroshio intrusion into Southeast Taiwan Strait during ENSO events. Cont. Shelf Res. 103b:date>7987.. Ishii and Kimoto, 2009 M.Ishii, , M.Kimoto, b:se">Ishi. Reevaluation of historical ocean heat content variations with time-varying XBT and MBT depth bias corrections. J. Oceanogr. 65b:date>287299. http://dx.doi.org/10.1007/s10872-009-0027-7.. Katsura et al., 2013 S.Katsura, , E., orOka, , B., orQiu, , N.Schneider, b:se">Katsura . Formation and subduction of North Pacific tropical water and their interannual variability. J. Phys. Oceanogr. 43b:date>24002415.. Lauri et al., 2012 H.Lauri, , H.de Moel, , ce:given-name>P.J., or>T.A.Rasanen, , M.Keskinen, , M.Kummu, b:se">Lauri . Future changes in Mekong River hydrology: impact of climate change and reservoir operation on discharge. Hydrol. Earth Syst. Sci. 16b:date>46034619.. Liang et al., 2008 ce:given-name>W.D.Liang, , ce:given-name>Y.J., or>T.Y., or>W.S.Chuang, b:se">Liang . Kuroshio in the Luzon Strait. J. Geophys. Res. 113b:date>C08048. http:// dx.doi.org/10.1029/2007JC004609.. s, 2001 R.Lukas, b:se">34853488.. Lukas and Lindstrom, 1991 R.Lukas, , E.Lindstrom, b:se">Lukas a. The mixed layer of the western equatorial Pacific Ocean. J. Geophys. Res. 96b:date>33433358.. Moon et al., 2013 e:given-name>J.-H., or>Y.T., or>P.D.Bromirski, , ce:given-name>A.J.Miller, b:se">Moon . Multidecadal regional sea level shifts in the Pacific over 1958–2008. J. Geophys. Res. Oceans 118b:date>70247035. http://dx.doi.org/10.1002/2013JC009297.. Nan et al., 2014 F., orNan, , H., orXue, , F., orYu, b:se">Nan . Kuroshio intrusion into the South China Sea: a review. Prog. Oceanogr.b:date>. http://dx.doi.org/10.1016/ j.pocean.2014.05.012.. Nan et al., 2015 F., orNan, , F., orYu, , H., orXue, , R., orWang, , G., orSi, b:se">Nan . Ocean salinity changes in the northwest Pacific subtropical gyre: the quasi-decadal oscillation and the freshening trend. J. Geophys. Res. Oceans 120b:date>. http:// dx.doi.org/10.1002/2014JC010536.. Nan et al., 2011 F., orNan, , H., orXue, , F., orChai, , L., orShi, , M., orShi, , P., orGuo, b:se">Nan . Identification of different types of Kuroshio intrusion into the South China Sea. Ocean Dyn.b:date>. http://dx.doi.org/10.1007/s10236-011-0426-3.. Nan et al., 2013 F., orNan, , H., orXue, , F., orChai, , D., orWang, , F., orYu, , M., orShi, , P., orGuo, b:se">Nan . Weakening of the Kuroshio intrusion into the South China Sea over the past two decades. J. Clim.b:date>. http://dx.doi.org/10.1175/JCLI-D12-00315.1.. Qiu and Chen, 2012 B., orQiu, , ce:given-name>S.M., or>. Multidecadal sea level and gyre circulation variability in the Northwestern Tropical Pacific Ocean. J. Phys. Oceanogr. 42b:date>193206.. Qiu et al., 2015 B., orQiu, , S., orChen, , L., orWu, , S., orKida, b:se">Qiu . Wind- versus Eddy-Forced regional sea level trends and variability in the North Pacific Ocean. J. Clim. 28b:date>15611577.. Qu et al., 2005 T., orQu, , Y., orDu, , G.Meyers, , A.Ishida, , D., orWang, b:se">Qu . Connecting the tropical Pacific with Indian Ocean through South China Sea. Geophys. Res. Lett. 32b:date>L24609. http://dx.doi.org/10.1029/2005GL024698.. u, 2000 ce:given-name>T.D., or>. Upper-layer circulation in the South China Sea. J. Phys. Oceanogr. 30b:date>14501460.. Qu et al., 2000 ce:given-name>T.D., or>Qu . Intrusion of the North Pacific waters into the South China Sea. J. Geophys. Res. 105) (C3b:date>64156424. http://dx.doi.org/10.1029/1999JC900323.. Qu et al., 2004 ce:given-name>T.D., or>Y.Y., or>Qu . Can Luzon Strait transport play a role in conveying the impact of ENSO to the South China Sea? J. Clim. 17b:date>36443657.. g, 2006 ce:given-name>Y.T., or>C11S19. http://dx.doi.org/10.1029/2005JC003189.. Suga et al., 2000 T., orSuga, , A., orKato, , K.Hanawa, b:se">Suga . North Pacific tropical water: its climatology and temporal changes associated with the climate regime shift in the 1970s. Prog. Oceanogr. 47b:date>223256.. Sugimoto et al., 2013 S.Sugimoto, , N.Takahashi, , K.Hanawa, b:se">Sugimoto . Marked freshening of North Pacific subtropical mode water in 2009 and 2010: influence of freshwater supply in the 2008 warm season. Geophys. Res. Lett. 40b:date>31023105..

Fig. 15. Yearly variations of PDO index (units: ℃), westward NECN transport (units: Sv), and the LST above 100 m (units: Sv) derived from GODAS output, and observed salinity average above 100 m (units: psu) during the period 1980–2012. Time series of the PDO index are from http://jisao.washington.edu/pdo/PDO.latest. All time series were normalized, setting means to 0 and standard deviations to 1.

Acknowledgements We are most thankful to the public officers below for providing the data freely. Subsurface Temperature and Salinity Analyses by Ishii et al. are publicly available at http://rda.ucar.edu/datasets/ ds285.3/. The altimeter products were produced by Ssalto/Duacs and distributed by Aviso, with support from Cnes (http://www. aviso.oceanobs.com/duacs/). OAFlux evaporation product is from ftp://ftp.whoi.edu/pub/science/oaflux/data_v3, and CPC precipitation product is from http://apdrc.soest.hawaii.edu/ data/. GODAS outputs were downloaded from http://www.esrl. noaa.gov/psd/data/gridded/data.godas.html. The discharge data of the Mekong River, Pearl River, and Red River were provided by the Mekong River Commission (http://pmfm.mrcmekong.org/), the Ministry of Water Resources of China, and the National Centre for Hydro-Meteorological Forecasting of Vietnam, respectively. This work was jointly supported by the National Natural Science Foundation of China (Grant no. 41306020 and 41676005), the National Programme on Global Change and Air-sea Interaction (GASI-IPOVAI-01-06), the CAS Interdisciplinary Innovation Team, the NSFC Innovative Group Grant (Project no. 41421005), and the NSFC-Shandong Joint Fund for Marine Science Research Centers (Grant no. U1406401). References Behringer, D.W., Ji, M., Leetmaa, A., 1998. An improved coupled model for ENSO prediction and implications for ocean initialization. Part I: the ocean data assimilation system. Mon. Weather Rev. 126, 1013–1021. Carton and Santorelli, 2008 ce:given-name>J.A.Carton, , A.Santorelli, b:sew?>Carton an. Global upper ocean heat content as viewed in nine analyses. J. Clim. 21b:date>60156035. http://dx.doi.org/10.1175/2008JCLI2489.1.. Chen and Huang, 1996 e:given-name>C.-T., or>M.-H.Huang, b:se">Ch. A mid-depth front separating the South China Sea water and the Philippine Sea water. J. Oceanogr. 52b:date>1725.. Chen et al., 2002 ce:given-name>M.Y., or>P.P., or>J.E.Janowiak, , ce:given-name>P.A.Arkin, b:se">Chen . Global land precipitation: a 50-yr monthly analysis based on gauge observations. J. Hydrometeorol. 3b:date>249266.. Chen and Tung, 2014 X., orChen, , e:given-name>K.-K., or>C. Varying planetary heat sink led to global-warming slowdown and acceleration. Science 345b:date>897903.. Durack and Wijffels, 2010 ce:given-name>P.J.Durack, , ce:given-name>S.E.Wijffels, b:se">Durack . Fifty-year trends in global ocean salinities and their relationship to

29

Deep–Sea Research I 118 (2016) 20–29

F. Nan et al.

seasonal variation in the South China Sea throughflow transport. J. Geophys. Res. Oceans 118b:date>68546866. http://dx.doi.org/10.1002/2013JC009367.. u, 2013 K., orYu, , ce:given-name>T.D., or>97979805.. Yu and Weller, 2007 L., orYu, , ce:given-name>R.A.Weller, b:se">Y. Objectively Analyzed air-sea heat fluxes for the global ice-free oceans (1981-2005). Bull. Am. Meteorol. Soc. 88b:date>527539.. Yu et al., 2008 ce:given-name>Z.J., or>J.P.McCreary, , M.Yaremchuk, , R.Furue, b:se">Yu . Subsurface salinity balance in the South China Sea. J. Phys. Oceanogr. 38b:date>527539. http://dx.doi.org/10.1175/ 2007JPO3661.1.. Zeng et al., 2014 L., orZeng, , W.Timothy Liu, , H., orXue, , P., orXiu, , D., orWang, b:se">Zeng . Freshening in the South China Sea during 2012 revealed by aquarius and in situ data. J. Geophys. Res. Oceans 119b:date>. http://dx.doi.org/10.1002/ 2014JC010108.. Zhai et al., 2014 F., orZhai, , Q., orWang, , F., orWang, , D., orHu, b:se">Zhai . Decadal variations of Pacific North Equatorial Current bifurcation from multiple ocean products. J. Geophys. Res. Oceans 119b:date>12371256. http://dx.doi.org/ 10.1002/2013JC009692..

Wang and Coauthors, 2013 ce:given-name>D.X., or>Wang a. Progress of regional oceanography study associated with western boundary current in the South China Sea. Chin. Sci. Bull. 58b:date>12051215〈10.1007/ s11434-012-5663-4〉.. Williams et al., 2007 ce:given-name>P.D.Williams, , E.Guilyardi, , R.Sutton, , J.Gregory, , G.Madec, b:se">Williams . A new feedback on climate change from the hydrological cycle. Geophys. Res. Lett. 34b:date>L08706. http://dx.doi.org/10.1029/ 2007GL029275.. Wu et al., 2012 C., orWu, , ce:given-name>S.L., or>Wu . Quantifying the anthropogenic and climatic impacts on water discharge and sediment load in the Pearl River (Zhujiang), China (1954–2009). J. Hydrol.b:date>452453.. Wu et al., 2010 R., orWu, , Z., orWen, , S., orYang, , Y., orLi, b:se">Wu . An interdecadal change in southern China summer rainfall around 1992/93. J. Clim. 23b:date>23892403. http://dx.doi.org/10.1175/2009JCLI3336.1.. Yan et al., 2013 ce:given-name>Y.F., or>E.P.Chassignet, , ce:given-name>Y.Q., or>W.K.Dewar, b:se">Yan . Freshening of subsurface waters in the Northwest Pacific subtropical gyre: observations and dynamics. J. Phys. Oceanogr. 43b:date>27332751.. Yang et al., 2013 J., orYang, , X., orLin, , D., orWu, b:se">Yang . On the dynamics of the

30