Spatial and temporal variability of Atlantic Water in the Arctic from observations

Atlantic Water (AW) is the largest reservoir of heat in the Arctic Ocean, isolated from the surface and sea-ice by a strong halocline. In recent years AW shoaling and warming are thought to have had an increased influence on sea-ice in the Eurasian Basin. In this study we analyse 59000 profiles from across the Arctic from the 1970s to 2018 to obtain an observationally-based pan-Arctic picture of the AW layer, and to quantify temporal and spatial changes. The potential temperature maximum of the AW (the AW core) is found to be an easily detectable, and generally effective metric for assessments of AW properties, although temporal trends in AW core properties do not always reflect those of the entire AW layer. The AW core cools and freshens along the AW advection pathway as the AW loses heat and salt through vertical mixing at its upper bound, as well as via likely interaction with cascading shelf flows. In contrast to the Eurasian Basin, where the AW warms (by approximately 0.7°C between 2002 and 2018) in a pulse-like fashion and has an increased influence on upper ocean heat content, AW in the Canadian Basin cools (by approximately 0.1°C between 2008 and 2018) and becomes more isolated from the surface due to the intensification of the Beaufort Gyre. These opposing AW trends in the Eurasian and Canadian Basins of the Arctic over the last 40 years suggest that AW in these two regions may evolve differently over the coming decades.

to obtain a broad picture of the AW and its variability. The AW temperature maximum 38 is found to be an easily observable, generally effective way to assess how much heat is 39 stored in the AW layer. Over the period studied, the AW in the eastern Arctic warmed 40 and had an increasing influence on the amount of heat in the surface layer, whereas AW 41 heat became increasingly isolated from the surface in the west due to changes in local  The AW layer is the most significant reservoir of heat in the Arctic Ocean (Carmack 59 et al., 2015), therefore changes in its temperature could have a significant impact on the 60 Arctic region. The AW layer currently contains enough heat to melt all Arctic sea-ice 61 within just a few years if this heat were brought to the surface in that time (Turner, 2010), 62 -2-manuscript submitted to JGR: Oceans although across most of the Arctic the AW is isolated from the sea-ice and surface mixed 63 layer by a strong halocline. Observations suggest that AW temperature variations are 64 dominated by low-frequency oscillations with a period of 50-80 years, linked to changes 65 in the Nordic Seas which are advected through the Fram Strait (Polyakov et al., 2004). 66 Superimposed on these low-frequency oscillations are inter-annual pulse-like tempera- ing trend in AW temperature over the twentieth century (Polyakov et al., 2004(Polyakov et al., , 2012, 71 and AW in the Fram Strait is now unprecedentedly warm compared to the last two mil-72 lennia, with a rapid temperature increase in the upper AW layer over the last 120 years 73 (Spielhagen et al., 2011). 74 In the eastern Eurasian Basin, recent increases in AW temperature, along with as-75 sociated shoaling of the AW and a weakening halocline, have enhanced vertical heat trans-76 fer from the AW to the surface layer and have resulted in a substantial reduction in win-77 ter sea-ice formation (Lind et al., 2018;Polyakov et al., 2010Polyakov et al., , 2017. This "Atlantifica-  Understanding how AW heat is likely to change in the future is therefore a key part 105 of predicting what will happen to the Arctic in the years to come. There is large discrep-106 ancy and bias amongst coupled climate model representations of AW in the Arctic, with 107 the AW layer generally being too deep and thick. The AW temperature biases are pri-108 marily due to inaccurate representation of sea ice coverage and surface cooling in the Bar-109 ents Sea, formation of cold and dense water in the Barents Sea, and AW inflow temper- observations has increased in recent years. This study aims to synthesise data from var-118 ious sources across the Arctic from the 1970s to 2018 to give a pan-Arctic, up-to-date 119 description of the AW layer and its impact on the water column. Diagnostics derived from 120 these observations, such as the temperature, salinity and depth at the AW temperature 121 maximum (the AW core) and AW heat content are used to characterise the spatial and 122 temporal variability of the AW and are described in section 2. The spatial variability of 123 the AW properties is investigated in section 3, with temporal variability in both the east-124 ern and western Arctic described in section 4. Observed changes in AW and heat dis-125 tribution within the water column at moorings and at repeated CTD transects are dis-126 cussed in section 5. Section 6 explores regional correlations between AW core metrics 127 and vertically integrated AW layer properties to investigate both how representative the 128 AW core temperature is of AW heat content, and regional differences in mixing. Con-129 clusions are given in section 7. Observational System (NABOS, https://uaf-iarc.org/nabos), and data from the NOAA 137 World Ocean Database (WOD, https://www.ncei.noaa.gov/products/world-ocean 138 -database). The WOD collates oceanic observational data from a wide range of sources.

139
The WOD data used here are those from CTD profiles, drifting buoys, and ocean sta-  Throughout the paper, salinity is given in Practical Salinity Units, and potential 144 temperature (when not available directly from the observational data product), heat con-  All data used in this study are processed versions of the raw data gathered in the 148 field. Details of these procedures can be found in the sources referenced above, but all 149 involved calibration, sensor-correction and the removal or flagging of obviously erroneous 150 data. In addition to this initial processing, further routines were applied to the data and 151 profiles were smoothed for much of our analysis -details of which are given below. Pro-152 files with more than 10 % of data masked or flagged as suspicious were omitted from the 153 analysis and, unless otherwise stated, monthly mean data from moorings were used so 154 as not to bias any regional analysis to the mooring locations due to the relative high sam-  Here we define the Atlantic Water layer as that portion of the water column that 158 lies below the halocline and has potential temperature above 0°C. The top and bottom 159 of the layer are the 0°C crossing points, either side of the potential temperature max-160 imum, and the AW layer thickness is the distance between them. As density is driven 161 by salinity in the Arctic, potential temperature is effectively a passive tracer. The po-162 tential temperature maximum, referred to here as the AW core, is commonly used to fol-    To ensure the AW core identified in each profile was not an artefact of limited sam-171 pling, profiles were required to start above 100 m and cover a depth range of at least 500 172 m before being used to identify the AW core (this also eliminated data from the surround-173 ing shelf seas, allowing the study to focus on AW within the Arctic basin only). This lat-174 ter step resulted in about 44000 AW core data points. Before identifying the core, pro-  This quantity is proportional to the average temperature over that depth range.   Building on the regional differences in AW shown in Figure  308 The year-to-year spatial variation in data distribution in the eastern Arctic makes   that the AW that reaches the shelf is that which is fresher. As above, the core temper- has a comparatively salty, deep AW core. This non-coincidence of AW core salinity and 393 temperature changes suggests that even enhanced mixing due to a weaker halocline does 394 not mask the signal of these warm AW pulses.

395
The fourth panel of Figure 9 shows the "heat content density" (the heat content is also reflected in the dissimilarity between variations in AW layer heat content and UHC 410 seen in Figure 9. idea that this AW core salinity increase is related to AW core heat loss, and is not re-439 lated to changes in the AW layer as a whole, is also supported by a lack of trend in AW ings (data in black in Figures 4-7) are more clearly shown in the time series in Figure   460 -19-manuscript submitted to JGR: Oceans Table 1. Trends in monthly mean AW core potential temperature, salinity, potential density anomaly, and depth from each of the four Canada Basin moorings. R-squared values for the fit of each trend are given in parentheses. Note that mooring C data only cover five years (see Figure   11).  Table 1. Again, the west-461 ern AW core temperature and depth trends in Figure 11 and Table 1 oppose those ex-462 pected from the Atlantification reported in the east.

463
As seen in the maps in Figure 4, the warm anomaly that entered the Canada Basin  Figure 11 and Table 1), and the pre-2010 warming at moorings C and D ( Figure 11).

467
There is an increase in salinity at all moorings, with significant trends at moorings A, 468 C and D. As discussed above, this seems to be related to the cooling of the core after 469 the spread of the warm anomaly. The resultant increase in AW core density can be clearly 470 seen in Figure 11, with significant trends at moorings A, B and D (note that C has a shorter 471 time series). The trends in depth in both Figures 7 and 11 are also significant (Table 1), 472 and could be attributed to the increase in AW core salinity. However, comparing the pat-473 terns of change in AW core depth and salinity in Figure 10 suggests that the salinity in-

494
The increase in upper layer heat content density seen in Figure 10 can also be at- Although spatial and temporal patterns of AW core potential temperature and AW 510 layer heat content in Figure 10 are similar, they do not match as closely as might be ex-511 pected. We explore this further in the following section. Much of the analysis in this study has involved the use of AW core properties to 514 infer AW layer properties within the Arctic. It is therefore important to investigate how 515 representative AW core properties are of the AW layer in general. Here we compute cor-516 relations between AW core and integrated AW layer metrics, which also shed some light 517 on how the AW layer loses heat in each region.

518
The maps in Figure 13 show total AW layer heat content during different time pe-519 riods, chosen to give a roughly even data distribution between panels. As most obser- tic. The heat content maps in Figure 13 are very similar to the AW core potential tem-535 perature maps in Figure 2, further suggesting that the AW core temperature captures 536 the general pan-Arctic spatial pattern of AW heat content variability well. However, the 537 cooling of the AW core in the Canada Basin (Figures 4 and 11) does not seem represen-538 tative of the steady western AW heat content values in Figure 13, providing more ev-539 idence that the decrease in temperature (and increase in salinity) of the AW core do not 540 reflect temporal changes in the AW layer itself.

541
To be more quantitative in assessing how well the AW core temperature represents 542 AW heat content, correlations between these two metrics were computed. Figure 14 shows 543 scatter plots between pairs of variables computed from profiles in both the eastern (re-544 stricted to EEB only due to constraints on spatial data coverage -see Figure 13) and west- AW mean salinity and AW heat content, and (e) AW core salinity and AW core potential temperature. Blue data is from the eastern Arctic, with red data from the western Arctic (regions defined in Figure 1). R-squared values and regression lines are shown for each scatter plot.
are given, along with regression lines. The relationship between total AW heat content 547 and the potential temperature of the AW core is shown in Figure 14a. There is a strong 548 correlation between these two variables in both the EEB and western Arctic, highlight-549 ing the general effectiveness of the AW core temperature as an easily measurable met-550 ric for assessing changes in AW heat content. As noted above, however, caution must 551 be taken when using AW core properties to infer temporal trends in the AW layer, as 552 the cooling (and resultant increase in salinity) of the AW core does not necessarily cor-553 respond with a cooling (or salinification) of the AW layer more generally.
554 Figure 14b shows AW layer thickness against AW layer heat content, with a mod-555 erate correlation in the EEB and a weak correlation in the west. This is an important 556 reminder that although the potential temperature of the AW core will give a good idea 557 of how AW heat content may vary (see Figure 14a), AW heat located away from the AW 558 core also affects AW heat content. This is particularly true in the EEB where BSBW 559 heat in the deep AW (which does not vary in concert with FSBW/the AW core, as seen 560 in Figure 3) affects AW heat content. 561 Figure 14c shows the relationship between AW core depth and the depth of the up-562 per boundary of the AW layer (i.e. the 0°C crossing point above the AW core depth give an idea of the role that mixing with fresher waters plays in AW heat loss. Figure   574 14d shows that, while there is a relatively high correlation between these variables in the 575 EEB, the correlation is negligible in the west. This implies that although mixing with 576 fresher waters is important for AW heat loss in the Eurasian Basin -as would be expected 577 given that the AW subducts beneath the cooler, fresh polar waters here, losing a lot of 578 heat -it is not as important in the western Arctic.

579
In Figure 14e, AW core salinity is compared to AW core potential temperature. In 580 the EEB this reflects what is seen in the integrated AW layer (Figure 14d). In the west 581 however, a stronger (moderate) correlation exists between the AW core data than that 582 between the AW layer data. This could be related to the thermohaline intrusions which focused on the properties of the AW core (the depth at which the maximum potential 591 temperature occurs). This was found to be a generally effective and easily detectable met-592 ric to assess the heat content of the AW layer. However, the depth of the AW core does 593 not always reflect the depth of the top of the AW layer, particularly in the eastern Arc-594 tic, and care must be taken when using temporal trends in AW core properties to assess 595 trends in the AW as a whole -a cooling or increase in salinity of the core does not nec-596 essarily translate to a cooling or increase in salinity of the entire AW layer.

597
In general, as the AW is advected around the Arctic the potential temperature and 598 salinity of its core decrease. Despite freshening, the AW core density increases along its 599 advection pathway. This is partially due to the preferential loss of heat and salt from 600 the top of the AW layer to the fresher, cooler water above through vertical mixing along 601 the AW advection pathway. This likely deepens the core without the AW layer as a whole 602 getting denser -upper AW cools such that the AW core (temperature maximum) is found 603 on deeper (denser) isopycnals. Interaction with dense shelf flows formed by brine rejec-604 tion during sea-ice formation may also play an important role in the cooling and fresh-605 ening of the AW core during its advection around the basin.

606
The evolution of AW has differed between the eastern and western basins of the 607 Arctic. In the Eurasian Basin, AW core temperature and AW heat content increased from

624
Despite the limitation of sparse, temporally inhomogeneous oceanographic mea-625 surements in the Arctic, pan-Arctic observational analysis can give useful insights into 626 the overall temporal and spatial patterns of heat distribution in the Arctic Ocean. Given 627 the challenges of realistically representing the AW layer in forced ocean-sea-ice and cou-628 pled climate models, and the stark regional differences emerging in the Arctic Ocean, 629 the use of pan-Arctic observations for model validation and benchmarking will be essen-630 tial. Only by combining insight from observations and models will we be able to accu-631 rately determine what the future Arctic will look like under a changing climate, which 632 is important both for the region itself as well as for the wider climate system.