Water Masses in the Atlantic Ocean: Characteristics and Distributions

: A large number of water masses are presented in the Atlantic Ocean and knowledge of their distributions and properties are important for understanding and monitoring of a range of oceanographic phenomena. The characteristics and distributions of water masses in biogeochemical space are useful for, in particular, chemical and biological oceanography to understand the origin and mixing history of water samples. Here we define the c haracteristics of the major water masses in the Atlantic Ocean as Source Water Types (SWTs) from their formation areas, and map out their distributions. The SWTs are described by six properties taken from the biased adjusted data product GLODAPv2, including both conservative (conservative temperature and absolute salinity) and non-conservative (oxygen, silicate, phosphate and nitrate) properties. The distributions of these water masses are investigated with the use of the Optimal Multi-Parameter (OMP) method and mapped out. The Atlantic Ocean is divided into four vertical layers by distinct neutral densities and four zonal layers to guide the identification and characterization. The water masses in the upper layer originate from winter-time subduction and are defined as Central Waters. Below the upper layer, the intermediate layer consists of three main water masses; Antarctic Intermediate Water (AAIW), Subarctic Intermediate Water (SAIW) and Mediterranean Water (MW). The North Atlantic Deep Water (NADW, divided into its upper and lower components) is the dominating water mass in the deep and overflow layer. The origin of both the upper and lower NADW is the Labrador Sea Water (LSW), the Iceland-Scotland Overflow Water (ISOW) and the Denmark Strait Overflow Water (DSOW). The Antarctic Bottom Water (AABW) is the only natural water mass in the bottom layer and this water mass is redefined as North East Atlantic Bottom Water (NEABW) in the north of equator due to the change of key properties, especial silicate. Similar with NADW, two additional water masses, Circumpolar Deep Water (CDW) and Weddell Sea Bottom Water (WSBW), are defined in the Weddell Sea region in order to understand the origin of AABW.


Introduction 39
The ocean is composed a large number of water masses without clear boundaries but gradual 40 transformations between each other (e.g. Castro et al., 1998). Properties of the water in the ocean are 41 not uniformly distributed and the characteristics vary with regions and depths (or densities). The water 42 masses, which are defined as bodies of water with similar properties and common formation history, are 43 referred to as a body of water with a measurable extent both in the vertical and horizontal, and thus a 44 quantifiable volume (e.g. Helland-Hansen, 1916; Montgomery, 1958). Mixing occurs inevitably 45 between water masses, both along and across density surfaces, and result in mixtures with different 46 properties away from their formation areas. Understanding of the distributions and variations of water 47 masses have significance to several disciplines of oceanography, for instance while investigating the 48 thermohaline circulation of the world ocean or predicting climate change (e.g. Haine and Hall, 2002;49 Tomczak and Godfrey, 2013; Morrison et al., 2015). 50 The concept of water masses is also important for biogeochemical and biological applications, where 51 the transformations of properties over time can be successfully viewed in the water masses frame-work. 52 For instance, the formation of Denmark Strait Overflow Water in the Denmark Strait was described 53 using mixing of a large number of water masses from the Arctic Ocean and the Nordic Seas (Tanhua et  on an basin or global scale leads to additional and repetitive amount of work by redefining water masses 62 in specific regions. The goal of this study is to facilitate water mass analysis in the Atlantic Ocean and 63 in particular, we aim at supporting biogeochemical and biological oceanographic work in a broad sense. 64 Understanding the formation, transformation, and circulation of water masses has been a research topic 65 in oceanography since the 1920s (e.g. Jacobsen, 1927;Defant, 1929;Wüst and Defant, 1936;Sverdrup, 66 1942 etc.). The early studies were mainly based on (potential) temperature and (practical) salinity as 67 summarized by Emery and Meincke (1986). The limitation of the analysis based on T-S relationship 68 is obvious; distributions of more (than three) water masses cannot be analyzed at the same time with 69 only these two parameters, so physical and chemical oceanographers has worked to add more parameters 70 to the characterization of water masses (e.g. Tomczak and Large, 1989;Tomczak, 1981;1999). The 71 Optimum Multi-parameter (OMP) method extends the analysis so that more water masses can be 72 considered by adding parameters/water properties (such as phosphate and silicate) and solving the 73 equations of linear mixing without assumptions. The OMP analysis has been successfully applied in a 74 chemical potential (Jackett et al., 2006;Groeskamp et al., 2016), are used in this study because they 110 systematically reflects the spatial variation of seawater composition in the ocean, as well as the impact 111 from dissolved neutral species on the density and provides a more conservative, actual and accurate 112 description of seawater properties (Millero et al., 2008;Pawlowicz et al., 2011;Nycander et al., 2015). 113

Water Masses (WMs) and Source Water Types (SWTs) 114
In practice, defining properties of water masses (WMs) is often a difficult and time-consuming part, 115 particularly when analyzing water masses in a region distant from their formation areas. Tomczak (1999) 116 defined a water mass as "a body of water with a common formation history, having its origin in a 117 particular region of the ocean" whereas Source Water Types (SWTs) describe "the original properties 118 of water masses in their formation areas". The distinction between the WMs and SWTs is that WMs 119 define physical extents, i.e. a volume, while SWTs are only mathematical definitions, i.e. SWTs are 120 defined values of properties without physical extents. Knowledge of the SWTs, on the other hand, is 121 essential in labeling WMs, tracking their spreading or mixing progresses, since the values from SWTs 122 describe their initial characteristics and can be considered as the fingerprints of WMs. The SWT of a 123 WM is defined by the values of key properties, while some of them, like Central Waters, require more 124 than one SWT to be defined (Tomczak, 1999). In this study, the terminology "water mass" is used in 125 the discussions, realizing that the properties of the WMs used for the further analysis actually refer to 126 SWTs. 127

Principle of OMP Analysis 129
For the analysis, six key properties are used to define SWTs, including two conservative (Conservative 130 Temperature and Absolute Salinity) and four non-conservative (oxygen, silicate, phosphate and nitrate) 131 properties. In order to determine the distributions of WMs, the OMP analysis is invoked as objective 132 mathematical formulations of the influence of mixing (Karstensen and Tomczak, 1997;. The 133 starting point is the 6 key properties (Figure 1) from observations (such as CT obs is the observed 134 Conservative Temperature). The OMP model determines the contributions from predefined SWTs (such 135 as CTi that describes the Conservative Temperature in each SWT), which represent the values of the 136 "unmixed" WMs in the formation areas, through a linear set of mixing equations, assuming that all key 137 properties of water masses are affected similarly by the same mixing processes. The fractions (xi) in 138 each sampling point are obtained by finding the best linear mixing combination in parameter space 139 defined by 6 key properties and minimizing the residuals (R, such as RCT is the residual of Conservative 140 Temperature) in a nonnegative least squares sense (Lawson and Hanson, 1974)  Sii, Phi and Ni (i = 1, 2 …, n) represent the predetermined (known) values in each SWT for each property. 151 The last row expresses the condition of mass conservation. 152 OMP analysis represent an inversion of an overdetermined system in each sampling point, so that the 153 sampling points are required to be located "downstream" from the formation areas, i.e. on the spreading 154 pathway. The total number of WMs which can be analyzed simultaneously within one OMP run is 155 limited by the number of variables/key properties, because mathematically, 6 variables (x1 -x6) can be 156 solved with 6 equations. In our analysis, one OMP run can solve up to 6 WMs. The above system of 157 equations can be written in matrix notation as: 158 Where G is a parameter matrix of defined SWTs with 6 key properties, x is a vector containing the 160 relative contributions from the "unmixed" water masses to the sample (i.e. solution vector of the SWT 161 fractions), d is a data vector of water samples (observational data from GLODAPv2 in this study) and 162 R is a vector of residual. The solution is to find the minimum of the residual (R) with a linear fit of 163 parameters (key properties) for each data point with a non-negative values. In this study, the mixed layer 164 is not considered as its properties tend to be strongly variable on seasonal time-scales so that water mass 165 analysis is inapplicable. The solution is dependent on, and sensitive to, the prior assumptions of the 166 properties of the SWTs. Here we have not explicitly explored this sensitivity, but note that a common 167 difficulty in OMP analysis is to properly define the SWT properties, and that this study provide a 168 generally applicable set of SWT properties for the major water masses in the Atlantic Ocean. 169

Extended OMP Analysis 170
The prerequisite (or restriction) for using (basic) OMP analysis is that the water masses are formed close 171 enough to the water samples with short transport times within a limited ocean region, for instance an 172 oceanic front or intertidal belt, so that the mixing can be assumed not influenced by biogeochemical 173 processes (i.e. assume all the parameters to be quasi-conservative). However, biogeochemical processes 174 cannot be ignored in a basin-scale analysis (Karstensen and Tomczak, 1998). Obviously, this 175 prerequisite does not apply to our investigation for the entire Atlantic, so the "extended" OMP analysis 176 is required. In this concept, non-conservative parameters (phosphate and nitrate) are converted into 177 conservative parameters by introducing the "preformed" nutrients PO and NO, where PO and NO 178 denotes the concentrations of phosphate and nitrate in seawater by considering the consumption of 179 dissolved oxygen by respiration (in other words, the alteration due to respiration is eliminated) (Broecker,180 1974; Karstensen and Tomczak, 1998 As a result, the number of water masses should be further reduced in one OMP run if the biogeochemical 192 processes are considered and extended OMP analysis is used. In this study, a total number of 5 water 193 masses are included in each OMP run. 194

Presence of mass residual 195
The fractions of WMs in each sample are obtained by finding the best linear mixing combination in 196 parameter space defined by 6 key properties which minimizes the residuals (R) in a non-negative least 197 squares sense. Ideally, a value of 100% is expected when the fractions of all the water masses are added 198 together. However, mass residuals where the sum of water masses for a sample differ from 100%, is 199 inevitable during the analysis and is due to sample properties outside the input SWTs to the OMP 200 formulation. There are two different cases. The first is that one single water mass is larger than 100% 201 and other water masses are all 0%. This mostly happens in the Central Waters (γ < 27.10 kg m -3 , Figure  202 2). The reason is that key properties, for instance CT, of Central Waters are variable. When the CT 203 increases beyond the range of this water mass, the OMP analysis considers the fraction is over 100%. 204 In this case, all such samples are set to 100% after confirming the absence of any other water mass. The 205 second case is that none of each single water mass is more than 100%, but the total fraction is more than 206 100% when added together. In this study, the total fractions are generally less than 105% (γ > 27.10 kg 207 m -3 , Figure 2). 208 In order to map the distributions of water masses, all GLODAPv2 data in the Atlantic Ocean (below the 209 mixed layer) are analyzed with the OMP method by using 6 key properties. In order to solve the 210 contradiction between the limitation of water masses in one OMP run and the total number of 16 water 211 masses (Figure 3), the Atlantic Ocean is divided into 17 regions (Table 1) and each with its own OMP 212 formulation, by only including water masses that are likely to appear in the area. In the vertical, neutral 213 density intervals are used to separate boxes. In the horizontal direction, the division lines are 40 °N, the 214 equator and 50°S where the area south of 50 °S is one region, independent of density, and additional 215 divisions are set between equator and 40 °N (γ at 26.70 and 27.30 kg m -3 , latitude of 30 °N, Table 1). In 216 this way, we end up with a set of 17 different OMP formulations that are used for estimating the 217 fraction(s) of water masses in each water sample. The neutral density and the latitude of the water sample 218 are thus used to determine which OMP should be applied ( Table 1). Note that all water masses are 219 present in more than one OMP so that reasonable (i.e. smooth) transitions between the different areas 220 can be realized.  To define the main water masses in the Atlantic Ocean, the determination of their formation areas is the 231 first step ( Figure 5) and then the selection criteria are listed to define SWTs based on the CT-SA 232 distribution, pressure (P) or neutral density (γ) ( Table 2). For some SWTs, additional properties such 233 oxygen or silicate are also required for the definition. With these criteria, which are taken from the 234 literature and also based on data from GLODAPv2 product, the SWTs of all the main water masses can 235 be defined for further estimating their distributions in the Atlantic Ocean by using OMP analysis. 236 For the water masses in the upper layer, i.e. the Central Waters, properties cover a "wide" range instead 237 of a "narrow" point value due to their variations, especially in CT and SA space i.e. The Central Waters 238 are labeled by two SWTs to identify the upper and lower boundaries of properties (Karstenson and 239 Tomczak, 1997;. In order to determine these two SWTs, one property is taken as a benchmark 240 (neutral density in this investigation) and the relationships to the others are plotted to make a linear fit 241 and the two endpoints are selected as SWTs to label Central Waters ( Figure 6). 242 During the determination of each SWT, two figures are displayed to characterize them, including a) 243 depth profiles of the 6 key properties under consideration (same color coding), and b) bar plots from the 244 distributions of the samples within the criteria (the blue dots in Figure 6  Most water masses maintain their original characteristics away from their formation areas. However, 250 some are worthy to be mentioned as products from mixing of several original water masses (for instance, 251 North Atlantic Deep Water is the product from Labrador Sea Water, Iceland-Scotland Overflow Water 252 and Denmark Strait Overflow Water). Also, characteristics of some water masses changes sharply 253 during their pathways (namely, the sharp drop silicate concentration of Antarctic Bottom Water after 254 passing the equator). As a result, it is advantageous to redefine their SWTs. In order to distinguish such 255 water masses from the other original ones, their defined specific areas are mentioned as "redefining" 256 areas instead of formation areas, because, strictly speaking, they are not "formed" in these areas. Central Waters can be easily recognized by their linear CT-SA relationships (Pollard et al., 1996;263 Stramma and England, 1999). In this study, the upper layer is defined to be located above the neutral 264 density isoline of 27.10 kg m -3 (below the mixed layer). The formations and transports of the Central

Eastern North Atlantic Central Water (ENACW) 272
The main Central Water in the region east of the Mid-Atlantic-Ridge (MAR) is the East North Atlantic 273 Central Water (ENACW, Harvey, 1982). This water mass is formed in the inter-gyre region during the 274 winter subduction (Pollard and Pu, 1985).

Western North Atlantic Central Water (WNACW) 285
Western North Atlantic Central Water (WNACW) is another water mass formed through winter 286 subduction (Worthington, 1959;McCartney and Talley, 1982) with the formation area at the southern 287 flank of the Gulf Stream (Klein and Hogg, 1996). In some studies, this water mass is referred to as 18 °C 288 water since a temperature of around 18 °C is one symbolic feature (e.g. Talley and Raymer, 1982;Klein 289 and Hogg, 1996). In general, seawater in the Northeast Atlantic has higher salinity than in the Northwest 290 Atlantic due to the stronger winter convection (Pollard and Pu, 1985) and input of MW (Pollard et al., 291 1996;Prieto et al., 2015). However, for the Central Waters, the situation is the opposite. WNACW has 292 a significantly higher salinity (SA) by ~0.9 g kg -1 than ENACW (Table 4). In this study, the work from 293 McCartney and  is followed and the region 24-37°N, 50-70°W shallower than 500 m 294 is considered as the formation area ( Figure 5). By defining the SWT of WNACW, the neutral density 295 between 26.20 and 26.70 kg m -3 is selected due to the discrete CT-SA distribution outside this range 296 (Table 2). Besides the linear CT-SA relationship, another property of this water mass is, as the 297 alternative name suggests, a temperature of around 18 °C, which is the warmest in the four Central 298 Waters due to the low latitude of the formation area and the impact from the warm Gulf Stream (Cianca 299 et al., 2009;Prieto et al., 2015). In addition, low nutrient is also a significant property compared to other 300 Central Waters (Figure 2 in Supplement). 301

Eastern South Atlantic Central Water (ESACW) 302
The formation area of ESACW is located in area southwest of South Africa and south of the Benguela 303 Current (Peterson and Stramma, 1991). In this region the Agulhas Current brings water from the Indian 304 Ocean (Deruijter, 1982;Lutjeharms and van Ballegooyen, 1988) that mixes with the South Atlantic 305 of this water mass. For the properties, neutral density (γ) between 26.00 and 27.00 kg m -3 and oxygen 311 concentration higher than 230 µmol kg -1 are used to define ESACW (Table 2). Similar as ENACW, 312 ESACW also exhibits relative large CT and SA ranges and low nutrient concentrations (especially low 313 in silicate) compared to the AAIW below. The properties in ESACW are similar to that of WSACW, 314 although with higher nutrient concentrations due to input from the Agulhas current (Figure 3 in 315 Supplement). 316

Western South Atlantic Central Water (WSACW) 317
The WSACW is formed in the region near the South American coast between 30 and 45 °S, where 318 surface South Atlantic Current brings Central Water to the east (Kuhlbrodt et al., 2007). The WSACW 319 is formed with little directly influence from other Central Water mass (Sprintall and Tomczak, 1993; 320 Stramma and England, 1999), while the origin of other Central Waters (e.g. ESACW or ENACW) can 321 be traced back, to some extent at least, to WSACW (Peterson and Stramma, 1991). This water mass is   sections are found in Figure 9. Note that the Central Waters are found at different densities, the eastern 337 variations being denser, so that the there is significant overlap in the horizontal distribution. The vertical 338 extent of the Central Waters is clearly seen in Figure 9. 339 The ENACW is mainly found in the northeast part of North Atlantic, near the formation area in the inter-340 gyre region (Figure 8). High fractions of ENACW is also found in a band across the Atlantic at around 341 40 °N, where the core of this water mass is found at close to 1000 m depth in the western part of the 342 basin ( Figure 9). 343 The WNACW is predominantly found in the western basin of the North Atlantic in a zonal band between 344 ~10 °N and 40 °N (Figure 8). The vertical extent of WNACW is significantly higher in the western basin 345 with an extent of about 500 meter in the west, tapering off towards the east (Figure 9). 346 The ESACW is found over most of the South Atlantic, as well as in the tropical and subtropical north 347 Atlantic (Figure 8). The extent of ESACW do reach particular far north in the eastern part of the basin 348 where it is an important component over the Eastern Tropical North Atlantic Oxygen Minimum Zone, 349 roughly south of the Cape Verde Islands. In the vertical direction, the ESACW is located below 350 WSACW (Figure 9). 351 The horizontal distribution of the WSACW does reach into the northern hemisphere but is, obviously, 352 concentrated in the western basin (Figure 8). In the vertical scale, the WSACW also tends to dominate 353 the upper layer of the South Atlantic above the ESACW (Figure 9). is considered to distinguish AAIW from surrounding water masses, including SACW in the north and 375 NADW in the deep. Piola and Georgi (1982) and Talley (1996) define AAIW as potential densities (σ θ ) 376 between 27.00/27.10 and 27.40 kg m -3 and Stramma and England (1999) define the boundary between 377 AAIW and SACW at σ θ = 27.00 kg m -3 and the boundary between AAIW and NADW at σ 1 = 32.15 kg 378 m -3 . The following criteria are used as selection criteria to define AAIW: neutral density between 26.95 379 and 27.50 kg m -3 and depth between 100 and 300 m. In addition, high oxygen (> 260 µmol kg -1 ) and 380 low temperature (CT < 3.5 °C) are used to distinguish AAIW from Central Waters (WSACW and 381 boundary to differentiate AAIW from AABW ( Table 2). The AAIW is distributed across most of the 383 Atlantic Ocean up to ~30 °N and the water mass fraction shows a decreasing trend towards the north 384 (Kirchner et al., 2009). AAIW is found at depths between 500 and 1200 m (Talley, 1996)  For defining the spatial boundaries we followed Arhan (1990) and selected the region between 35 and 398 55 °W and 50 and 60 °N, i.e. the region along the Labrador Current and north of the NAC as the 399 formation area of SAIW ( Figure 5). Within this area, neutral densities higher than 27.65 kg m -3 and CT 400 higher than 4.5 °C is selected to define SAIW by following Read, (2000). Samples in the depth range 401 from the MLD to 500 m are considered as the core layer of SAIW, which included the formation and 402 subduction of SAIW (Table 2). 403

Mediterranean Water (MW) 404
The along the European coast and its influence can be observed as far north as the Norwegian Sea (Reid, 414 1978;. The impact from MW is significant in almost the entire Northeast Atlantic in the 415 Intermediate Layer (east of the MAR, Figure 7 in Supplement), with high temperature and Salinity but 416 low nutrients compared to other water masses. 417 Here we followed Baringer and Price (1997) and define the SWT of MW by the high salinity (SA 418 between 36.5 and 37.00 g kg -1 , Table 2) samples in the formation area west of the Strait of Gibraltar 419 ( Figure 5). 420

Atlantic Distributions of Intermediate Waters 421
A schematic of the main currents in the intermediate layer (γ between 27.10 and 27.90 kg m -3 ) is shown 422 in Figure 10 (left panel). 423 The SAIW is mainly formed north of 30 °N in the western basin by mixing of two main sources, the 424 warmer and saltier NAC and the colder and fresher Labrador Current and characterized with relative 425 low CT (< 4.5 °C), SA (< 35.1 g kg -1 ) and silicate (< 11 μmol kg -1 ). The SAIW and MW can be easily 426 The AAIW has a southern origin and is found at slightly lighter densities (core neutral density ~27.20 439 kg m -3 , Figure 10

Labrador Sea Water (LSW) 467
As an important water mass that contributes to the formation of North Atlantic Deep Water (NADW), 468

Labrador Sea Water (LSW) is predominant in mid-depth (between 1000m and 2500m depth) in the 469
Labrador Sea region (Elliot et al., 2002). This water mass was firstly noted by (Wüst and Defant, 1936)  Atlantic-Ridge in the Iceland Basin. The following criteria, Conservative Temperature between 2.2 and 500 3.3 °C and Absolute Salinity higher than 34.95 g kg -1 , are used to define the SWT of ISOW, and neutral 501 density higher than 28.00 kg m -3 is added order to distinguish ISOW from LSW in the region west of 502 MAR (Table 2 and Figure 9 in Supplement). 503

Denmark Strait Overflow Water (DSOW) 504
A number of water masses from the Arctic Ocean and the Nordic Seas flows through Denmark Strait 505 west of Iceland. At the sill of the Denmark Strait and during the descent into the Irminger Sea, these 506 water masses undergo intense mixing. This overflow water mass is considered as the coldest and densest 507 component of the sea water in the Northwest Atlantic Ocean and constitute a significant part of the 508 southward flowing NADW (Swift, 1980). Samples from the Irminger Sea ( Figure 5) with neutral density 509 higher than 28.15 kg m -3 (Table 2 and Figure 10 in Supplement) are used for the definition of DSOW 510 (Rudels et al., 2002;Tanhua et al., 2005). 511

Upper North Atlantic Deep Water (uNADW) 512
The uNADW is mainly formed by mixing of ISOW and LSW and considered as a distinct water mass 513 south of the Labrador Sea as this region is identified as the redefining area of upper and lower NADW 514 (Dickson and Brown, 1994). The region between latitude 40 and 50 °N, west of the MAR is selected as 515 the redefining area of NADW ( Figure 5) and the criteria of neutral density between 27.85 and 28.05 kg 516 m -3 and CT < 4.0 °C within the depth range from 1200 to 2000 m (Table 2 and Figure 11 in Supplement) 517 are used to define the SWT of uNADW (Stramma et al., 2004). As a mixture from LSW and ISOW, the 518 uNADW obviously inherits many properties from LSW, but is also significantly influenced by the 519 ISOW. The relative high temperature (~3.3 °C) is a significant feature of the uNADW together with 520 relatively low oxygen (~280 µmol kg -1 ) and high nutrient concentrations, which is a universal symbol 521 of deep water (Table 4). 522

Lower North Atlantic Deep Water (lNADW) 523
The same geographic region is selected as the formation area of lNADW ( Figure 5). In this region, the 524 ISOW and DSOW, influenced by LSW, mix with each other and form the lower portion of NADW 525 (Stramma et al., 2004). Water samples between depths of 2000 and 3000 m with CT higher than ~2.5°C 526 and neutral densities between 27.95 and 28.10 kg m -3 are selected to define the SWT of lNADW (Table  527 2 and Figure 12 in Supplement). 528

Atlantic Distributions of Deep and Overflow Waters 529
The water masses dominate the neutral density interval 27.90 -28.10 kg m -3 in the Atlantic Ocean north  The LSW is commonly characterized as two variations, "upper" and "classic" although in this study we 535 consider this as one water mass in the discussion above. Our analysis indicates that the LSW dominates 536 the North West Atlantic Ocean in the characteristic density range. In Figure 12, we choose to display γ 537 = 27.95 that corresponds to the main property of the LSW (Kieke et al., 2006;. The LSW spreads 538 east and southward in the North Atlantic Ocean, but is less dominant in the area west of the Iberian 539 Peninsula where the presence of MW from the Gulf of Cadiz tends to dominate that density level. Note 540 that although the LSW is slightly denser than the MW, their density ranges do overlap (Figure 12 and  541

13). 542
The ISOW is mainly found in the Northeast Atlantic north 40 °N between Iceland and Iberian Peninsula 543 with core at γ = ~28.05 kg m -3 . The ISOW is also found west of the Reykjanes Ridge, in the Irminger 544 and Labrador Seas between the DSOW and LSW (Figure 12 and 13). 545 The DSOW is mainly found in the Irminger and Labrador Seas as the densest layer close to the bottom 546 ( Figure 11). Our analysis indicates a weak contribution of DSOW also east of the MAR. South of the 547 Grand Banks the DSOW is already significantly diluted and only low to moderate fractions are found 548 (Figure 12 and 13). 549 After passing 40 °N, the upper and lower NADW are considered as independent water masses and 550 dominate the most of the Atlantic Ocean in this density layer. The map in Figure 12 shows that upper 551 NADW covers the most area, while the lower NADW is found mainly found in the west region near the 552 Deep Western Boundary Current (DWBC), especially in South Atlantic. In the vertical view based on 553 sections (Figure 14), the southward transports of both upper and lower NADW can be seen until ~ 554 50 °S where they meets AABW in the ACC region. 555

The Bottom Layer and the Southern water masses 556
The Bottom Waters are defined as the densest water masses that occupy the lowest layers of the water 557 column, typically below 4000 m depth and with neutral densities higher than 28.10 kg m -3 . These water 558 masses have an origin in the Southern Ocean (Figure 15, left)  The definition of AABW is all water samples formed south of the Antarctic Circumpolar Current (ACC), 577 i.e. south of 63 °S in the Weddell Sea ( Figure 5), with neutral density (γ) larger than 28.27 kg m -3 (Weiss 578 et al., 1979;Orsi et al., 1999). As an additional constraint, AABW is defined as water samples with 579 silicate higher than 120 µmol kg -1 to distinguish from other water masses in this region as high silicate 580 is a trade mark property of AABW (Table 2). 581 The formation process of AABW is a mixture of another two original water masses, CDW and WSBW, 582 which are referred to as southern water masses, in the Weddell Sea region, consistent with Orsi et al.  (Figure 17). AABW is found from 1000m to 586 5500m depth (Figure 16 and 17) with low temperature (CT < 0 °C), salinity (SA < 34.68) but high 587 nutrient, especially silicate, concentrations (Figure 13 in Supplement). 588

Northeast Atlantic Bottom Water (NEABW) 589
Northeast Atlantic Bottom Water (NEABW), also called lower Northeast Atlantic Deep Water 590 (lNEADW, Garcia-Ibanez et al., 2015), is mainly found below 4000m depth in the eastern basin of the 591 North Atlantic (Figure 5). This water mass is an extension of AABW during the way to the north, since 592 the properties of AABW change significantly on the slow transport north. A new SWT is redefined for 593 this water mass in north of the Equator, similar as the redefinition of NADW south of the Labrador Sea. 594 The region east of the MAR and between the equator and 30 °N, i.e. before NEABW enters the Iberian 595 Basin, is selected as the redefining area of NEABW. The criteria of depth deeper than 4000 m and CT 596 above 1.8 °C are also used (Table 2). In the CT-SA diagram in Figure 3, similar T-S distribution 597 between NEABW and AABW can be seen but with higher CT and SA of ~1.95 °C and ~35.060 g kg -1 . 598 Most NEABW samples have a neutral density higher than 28.10 kg m -3 and NEABW is characterized 599 by low CT and SA, but high silicate concentration (Figure 14 in Supplement). This further suggests that 600 NEABW originates from AABW, although most properties have been changed significantly from the 601 origin in the South Atlantic. In this study, the SWTs of CDW is defined by considering the lower branch and the region between 55 616 and 65 °S is selected as the formation area ( Figure 5). To define SWT of CDW (lower branch), water 617 samples are selected from depth between 200 and 1000m in this region and additional constraints are 618 SA higher than 34.82 g kg -1 and CT between -0.5 and 1.0 °C (Table 2). Similar to other bottom/southern 619 SWTs, CDW is also defined by high nutrient (silicate, phosphate and nitrate) and low oxygen 620 concentrations (Figure 15 in Supplement). 621

Weddell Sea Bottom Water (WSBW) 622
The Weddell Sea Bottom Water (WSBW) is the densest water mass in the bottom layer. As mentioned 623 in the above section, part of CDW from the upper branch cools down rapidly by mixing with extremely 624 cold shelf water and sinks down to the bottom along the continental slop (Gordon, 2001). WSBW is 625 This work is based on the comprehensive and detailed data from GLODAP data set throughout the past 685 few decades. In particular, we are grateful to the efforts from all the scientists and crews on cruises, who 686 generated funding and dedicated time on committing the collection of data. We also would like to thank 687 the working groups of GLODAP for their support and information of the collation, quality control and   Conservative Temperature (°C), Absolute Salinity (g kg -1 ), Neutral Density (kg m -3 ), Oxygen and Nutrients (µmol kg -1 ) The red Gaussian fit shows mean value and standard deviation of selected data.