Isopycnal mixing of interhemispheric intermediate waters by subthermocline eddies east of the Philippines

Both sporadic observations and modelling studies suggest that subthermocline eddies (SEs) exist east of the Philippines, where interhemispheric waters meet. However, effects of SEs on water mass mixing have never been observed. Here, using data from mooring and buoy deployed in the frontal region of the interhemispheric water masses, we show for the first time that the SEs act as an “underwater mixer” of intermediate waters from north and south Pacific oceans. The SEs have typical swirl speeds of 0.1~0.4 m s−1 between 200 and 800 m depth with a dominant period of ~90 days. Variation in intermediate water salinity also had a period of ~90 days, lagging eddy speed by ~8 days. Horizontal eddy diffusivity representative of eddy mixing rate was quantified using a mixing-length framework. Horizontal eddy diffusivity had both surface and subthermocline maxima. The vertically varying eddy diffusivity can be used to improve parameterization of eddy stirring in the tropical Pacific by coarse-resolution ocean climate models. The effect of the SEs on mixing of intermediate water masses seems not resolved by available eddy-resolving ocean models typically used for this region.

www.nature.com/scientificreports www.nature.com/scientificreports/ studies suggest the existence of SEs with dominant intraseasonal variability [18][19][20][21][22] . It is known that eddies can mix water masses and influence the distribution of tracers such as heat and salt 32,33 . The effect of SEs on isopycnal mixing of interhemispheric intermediate waters in this region remains unclear.
Understanding mixing by mesoscale eddies is of both theoretical interest and practical importance 34 . In this study, based on data from mooring and buoy deployed in the frontal region of interhemispheric water masses, we show the characteristics of the SEs and their effects on isopycnal mixing of intermediate waters from north and south Pacific oceans. Horizontal eddy mixing length and diffusivity are quantified using a mixing-length framework based on observations, which can be used to improve parameterization of eddy stirring in this region for coarse-resolution ocean climate models.

Data and Methods
Mooring and buoy measurements. To investigate water mass mixing, a subsurface mooring system (M8) was deployed in the frontal region of interhemispheric water masses at 129°E, 8°N (Fig. 1a) from January 2016 to April 2017. Similar to earlier studies 24,29 , two (one upward looking and one downward looking) 75 kHz Acoustic Doppler Current Profilers (ADCPs) manufactured by Teledyne RD Instruments were installed on the main float of M8 at 400 m water depth to measure current velocity in the upper 800 m. The ADCPs were configured to measure hourly with a standard bin size of 8 m. A 1-day running mean was applied to the raw ADCP velocity data to remove tidal and synoptic signals and linearly interpolated at 8-m intervals profiles 31 . Six conductivity-temperature-depths sensors (CTDs) manufactured by Sea-Bird Scientific were installed at depths of 350, 400, 500, 600, 700, and 800 m. These CTDs were configured to measure temperature and salinity every 10 minutes, and a 1-day running mean was applied to the data. www.nature.com/scientificreports www.nature.com/scientificreports/ The buoy (T14) was used to obtain temperature and salinity data from surface to 750 m. Buoy T14 is part of the Triangle Trans-Ocean Buoy Network (TRITON) that is located at 130°E, 8°N (Fig. 1a). The TRITON buoys were deployed since 1998 by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) as part of the Tropical Ocean Climate Study (TOCS) 35  Estimation of mixing length and horizontal diffusivity. Horizontal eddy mixing rate can be quantified through eddy diffusivity (κ h ) 36 . Until now, research on the spatial and temporal distribution of eddy diffusivity has been limited. The mixing length (λ) represents the distance a fluid parcel is transported before significant irreversible mixing occurs 32 . A mixing-length framework that relates observed salinity anomalies on isopycnal surfaces to gradients of the mean salinity field is used to estimate horizontal λ and κ h as follows 32,33 : where S′ is the isopycnic salinity anomaly, = 〈 ′ + ′ 〉 u u v rms 2 21/2 is the root-mean-square eddy velocity, c 0 is the constant mixing efficient (0.16), braces {} indicate a one-year running mean, and brackets 〈〉 are for the temporal average over all years 32 . u′ and v′ are zonal and meridional velocity anomalies, respectively. Mean salinity gradient 〈|∇ |〉 S { } was calculated using the gridded Argo data set, i.e., Grid Point Value of the Monthly Objective Analysis (MOAA GPV). The MOAA GPV data is a global 1° × 1° dataset of monthly temperature and salinity starting from January 2001 to the present 37 . In this study, we used data obtained from 2008 to 2012 to correspond with data measured by buoy T14.

Model data.
To test whether the available eddy-resolving ocean models can reproduce the effects of the SEs on intermediate waters mixing, the OFES (Oceanic General Circulation Model for the Earth Simulator) model data without data assimilation and the HYCOM (Hybrid Coordinate Ocean Model) model data with data assimilation were selected. The OFES (HYCOM) model has a horizontal resolution of 0.1° × 0.1° (1/12° × 1/12°) and a total of 54 (32) levels. These two eddy-resolving models have been widely used in a number of earlier studies to investigate currents, eddies, and water masses in the northwestern Pacific Ocean 20,21,23,[27][28][29]38 . Detailed descriptions of the models are provided by those studies. In this study, daily (3-day intervals) temperature and salinity data are used from the HYCOM (OFES) output from 2008 to 2012.

Isopycnal mixing of intermediate waters.
Off the Philippine coast is a region of mixing of interhemispheric water masses. Water masses originating from the northern and southern hemispheres intersect here and are transformed through isopycnal/diapycnal mixing. In this study, we focus on isopycnal water mixing with κ h~O (10 3 ) m 2 s −1 , which is mainly induced by eddies and significantly greater than diapycnal mixing with diapycnal diffusivity κ ρ~O (10 −6 ) m 2 s −1 in this region 39 . Different water masses were often identified by salinity extremes 27,40 . Climatological salinity distribution along 130°E from gridded MOAA GPV data is shown in Fig. 1b. Low salinity water in the upper mixed layer characterizes the North Pacific Tropical Surface Water (NPTSW) that is formed in the northwestern Pacific warm/fresh pool where precipitation exceeds evaporation. In the thermocline (<26 σ θ ) and below the NPTSW, two subsurface salinity maxima exist. The southern one is the South Pacific tropical water (SPTW), with maximum salinity of ~35.4. It is saltier than the northern one, called NPTW, with maximum salinity of ~35.2. Both the NPTW and SPTW are formed in the middle of the Pacific subtropical gyres where evaporation is far higher than precipitation 40 . In the subthermocline (>26 σ θ ), there exist two subsurface salinity minima. The southern one, named the AAIW, has a minimum salinity of ~34.6 and is saltier than the northern one, called the NPIW with minimum salinity of ~34.2. Note that the depths of the SPTW and the AAIW in the southern hemisphere are greater than those of the NPTW and NPIW in the northern hemisphere.
The MC is strongest near the Philippine coast and can extend approximately 250 km offshore (128.75°E) 28 . The MUC is 50-80 km off the Philippine coast. Meridional MC and MUC have significant semiannual variability and play an important role in interhemispheric intermediate water exchange 29 . Most moorings have been deployed at ~127°E near the Philippine coast to observe variability of the MC and MUC 23,28,29,31 . Here, to observe eddy effects on water masses mixing, a subsurface mooring was deployed at 129°E, 8°N off the Philippine coast, where the eddy activity is dominant [20][21][22] . It is also in the frontal region of the interhemispheric water masses (Fig. 1a,b).
Salinity is a good indicator for investigation of intermediate water mixing [31][32][33] . Figure 1c shows a scatterplot of θ-S measured from six CTDs deployed on mooring M8. The magenta and red curves show representative θ-S diagrams at 15°N and 2°N, respectively. The minimum salinity of the AAIW is ~34.55 at 2°N at ~27.2 σ θ , while the minimum salinity of the NPIW is ~34.3 at 15°N at 26.6 σ θ . The most striking feature of Fig. 1c is that the observed salinity of intermediate water at 8°N is almost between salinities of AAIW and NPIW, which reflects the strong mixing of these two water masses. Data are not available above 350 m, and so water mass characteristics above 26.8 σ θ are not shown. There was a nearby TRITON buoy (T14) at 130°E, 8°E. Figure 1d shows a scatterplot of θ-S measured by 11 CTDs deployed on buoy T14. The results suggest that the AAIW and the NPIW are sufficiently mixed between 26.3 and 27.1 σ θ off the Philippine coast.
Observed SEs off the Philippine coast. From Fig. 2, the mean current measured by mooring M8 is south-westward with maximum velocity of ~0.15 m/s above 200 m. This is caused by cyclonic currents of the Mindanao Eddy (ME), since the mooring M8 is deployed in the northwestern of the ME based on satellite altimeter data (Figure not shown). The mean current below 200 m is near zero. Interestingly, the standard deviations of both zonal and meridional currents are larger than the mean currents, indicating that eddy activity is predominant. The maximum standard deviations of both zonal and meridional currents reach 0.25 m/s above 200 m. This is because the ME is a recirculation of the MC rather than a quasi-stationary eddy confined above the thermocline 38,41 . Below 200 m, the maximum standard deviation appears at ~400 m where the maximum speed of the SEs appears (Fig. 3a). The standard deviation of the zonal current is approximately 0.1 m/s, which is a little smaller than that of the meridional current (0.15 m/s).   (Fig. 3e,f), which is consistent with occurrence frequency of the SEs. Variations of intermediate water and the eddy current as measured by mooring M8 seem opposite (Fig. 4a). The correlation coefficient between eddy speed and salinity is greatest (r = −0.32, exceeding 99% significance level) when salinity lags eddy speed by 8 days (Fig. 4b). The 8-day lag reflects the response time of local salinity to the eddy. Thus, we can conclude that  www.nature.com/scientificreports www.nature.com/scientificreports/ intermediate water mixing off the Philippine coast observed by mooring M8 and buoy T14 is caused by water mass mixing from the SEs. The SEs seem to act as an "underwater mixer" of interhemispheric intermediate water masses.
Eddy stirring effects cannot be resolved by most coarse-resolution ocean climate models, and instead have to be parameterized. The value for parametrization of eddy diffusivity (κ h ) is typically of 1000 m 2 s −1 42,43 , which sometimes is higher in the upper ocean and lower below the thermocline 44 . Thus, quantification of vertical variation of eddy diffusivity will improve parameterization of mixing in coarse-resolution models. Based on observed variation in salinity by buoy 14 and eddy activities by mooring M8, horizontal eddy mixing length (λ) and diffusivity (κ h ) are estimated using Equations (1) and (2), respectively (Fig. 5). According to Fig. 5a-d, salinity variance, gradient, and current velocity are largest in the surface layer and decrease with depth. There exists a subthermocline maximum in the intermediate layer at ~26.8 σ θ . Figure 5e,f show the vertical distributions of estimated λ and κ h as a function of potential density and depth. λ and κ h are not uniform but vary vertically. They both have surface and subthermocline (at ~26.8 σ θ ) maximum. λ varies from 120 to 350 km, which is consistent with the eddy diameter in this region 22 . κ h varies with depth by about an order of magnitude from 3 × 10 3 m 2 s −1 to 1.2 × 10 4 m 2 s −1 . The averaged κ h is 6.4 × 10 3 m 2 s −1 , and the subthermocline maximum κ h is 1.1 × 10 4 m 2 s −1 . The results based on observations can be used to improve parameterization of eddy stirring in this region for ocean climate models.

Discussion
In this study, based on data from mooring and buoy deployed in the frontal region of the interhemispheric water masses, we show for the first time that SEs act as an "underwater mixer" of isopycnal mixing of North Pacific and South Pacific intermediate waters off the Philippine coast. The SEs have typical swirl speeds of 0.1-0.4 m s −1 between 200 and 800 m depth and dominant intraseasonal variability. Horizontal eddy mixing length and diffusivity are estimated using a mixing-length framework. Maximum eddy mixing length and diffusivity appear at the surface, decrease with depth, and exhibit a subthermocline maximum. Ocean models for climate studies are still too coarse to resolve mesoscale eddies and have to be parameterized 36,[42][43][44] . Arguably, ocean models are limited by observations of eddy diffusivity with spatial and temporal variations 36 . With more and more subthermocline/ intrathermocline/mode-water/subsurface eddies being observed around the global ocean [45][46][47] , the typical value of eddy diffusivity (1000 m 2 s −1 ) which is higher in the upper ocean and lower below the thermocline is being questioned. Significant deviation from the mean eddy diffusivity in the subsurface will induce more mixing, a process missed in oceanic models. The results presented here that are based on observations can be used to improve parameterization of eddy mixing in this region for climate and ocean models.
Mesocale eddies can be resolved in high-resolution ocean models. The eddy-resolving OFES and HYCOM model outputs have been extensively analyzed by a number of earlier studies to investigate ocean circulation and eddies in the northwestern Pacific Ocean 20,21,23,27-29,38 . However, the effect of SEs on intermediate waters mixing www.nature.com/scientificreports www.nature.com/scientificreports/ is not reproducible by these eddy-resolving ocean models. According to Fig. 6, the OFES model without data assimilation reproduces water mass structure within the thermocline. However, in the subthermocline, the AAIW and the NPIW are not mixed, which is not in accordance with observations presented here. The result of the HYCOM model with data assimilation is even less accurate. The salinity varies larger than the observed results. Meanwhile, the subthermocline maximum is not clear. According to Equations (1) and (2), vertical distributions of λ and κ h at 130°E, 8°N were derived from OFES and HYCOM outputs during 2008-2012 (Fig. 7). It can be seen that eddy mixing length and diffusivity are largest in the surface layer and decrease with depth. Compared www.nature.com/scientificreports www.nature.com/scientificreports/ with observational results, there are not subthermocline maxima in the intermediate layer at ~26.8 σ θ for the two models. The poor representation of eddy-mixing effects is the major cause of these biases. It is recommended that these two models should be applied with caution when investigating water mass mixing and transformation in the tropical Pacific Ocean.
In addition, this study focuses on the effects of eddies on isopycnal water mixing since horizontal mixing is more influential than vertical mixing in determining meso-to large-scale salinity distribution 44 . It has been observed that the ME can enhance diapycnal mixing at its flanks. The diapycnal diffusivity (κ ρ ) is elevated by an order of magnitude due to eddy-induced geostrophic shear 39 . The SEs off the Philippine coast can also induce strong vertical current shear in the subthermocline (Fig. 2b), which is suggested to enhance diapycnal mixing. Carefully designed observations may provide more insights on the effects of SEs on diapycnal mixing.

Data Availability
Buoy T14 data can be downloaded from the TRITON Web site http://www.jamstec.go.jp/jamstec/TRITON/real_ time/. The MOAA GPV data are publicly available at http://www.jamstec.go.jp/ARGO/argo_web/argo/?lang=en. The OFES data and the HYCOM data are publicly available at http://apdrc.soest.hawaii.edu/data/ and https:// hycom.org/, respectively. The mooring M8 data that support the findings of this study are available from the corresponding author upon request.