Decadal freshening of the Antarctic Bottom Water exported from the Weddell Sea

Recent decadal changes in Southern Hemisphere climate have driven strong responses from 9 the cryosphere. Concurrently, there has been a marked freshening of the shelf and bottom waters across a wide sector of the Southern Ocean, hypothesised to be caused by accelerated 11 glacial melt in response to a greater ﬂux of warm waters from the Antarctic Circumpolar 12 Current onto the shelves of West Antarctica. However, the circumpolar pattern of changes has been incomplete: no decadal freshening in the deep layers of the Atlantic sector had been 14 observed. In this study, we document a signiﬁcant freshening of the Antarctic Bottom Water 15 exported from the Weddell Sea, which is the source for the abyssal layer of the Atlantic 16 overturning circulation, and we trace its possible origin to atmospheric-forced changes in 17 the ice shelves and sea ice on the eastern ﬂank of the Antarctic Peninsula that include an 18 anthropogenic component. These ﬁndings suggest that the expansive and relatively cool 19 Weddell gyre does not insulate the bottom water formation regions in the Atlantic sector 20 from the ongoing changes in climatic forcing over the Antarctic region. 21

the common November-December occupation, and another in February 2009 (Table 1). 102 The Drake Passage hydrographic programme is a joint effort between the National 103 Oceanography Centre in Southampton and the British Antarctic Survey in Cambridge, U.K. 104 Most section occupations were conducted using the RRS James Clark Ross. Each occupation 105 typically consists of 30 full-depth conductivity-temperature-depth (CTD) stations between 106 Burdwood Bank and Elephant Island in the eastern Drake Passage, with a typical horizontal 107 resolution of 30 nautical miles. In early cruises, the data were collected using a Neil Brown  The same analysis was done for the LCDW salinity maximum (γ n = 28.00 − 28.11 kg 175 m −3 - Fig. 4c,d) in order to investigate any inter-cruise biases. The calculated freshening 176 trend is not statistically significant at the 90% significance level (p = 0.14). The LCDW time   Tracer-based studies suggest that the AABW in the northern Weddell Sea is made up of 222 ∼ 25% of shelf water. A 5% change in the source water mixing ratio would thus be enough 223 to account for the measured freshening. However, such an increase in the proportion of 224 WDW contributing to AABW production would also drive a warming of the AABW (by 225 ∼ 0.07 • C), which is not consistent with our observations. We conclude therefore that a       268 We construct a simplified freshwater budget of the continental shelf region off the LIS 269 to estimate the additional (relative to the pre-freshening era) freshwater volume, V f w , that 270 is required to have entered the area to explain the observed AABW freshening. It is worth 271 noting that any attempt to calculate an accurate estimate of V f w and to precisely determine 272 the driving mechanism are subject to significant uncertainty, given the sparse data available 273 in the region. Therefore, we restrict our analyses to order-of-magnitude calculations, so as 274 to identify the processes most likely to play a primary role in the freshening of the AABW.

275
Salt and mass conservation respectively dictate that (1) where ρ 0 , S 0 and V 0 are the initial in situ density, salinity and volume of the shelf waters, We set ρ 0 = 1027 kg m −3 and ρ f w = 1000 kg m −3 . We define V 0 as the volume of the 1b), which reveal that S f ∼ 34.6 . Thus, we find that V f w ≈ 8 × 10 10 m 3 .  The plausibility of this contribution may be assessed by considering, once again, a simple 321 freshwater budget. Salt and mass conservation dictate that where ρ si = 920 kg m −3 and S si = 6 are the characteristic density and salinity of sea ice, 323 respectively (Eicken 1997), and V si is a change in sea ice volume. From these expressions, 324 we estimate the reduction in sea ice production required to explain the observed freshening and obtain V si ≈ 1.1 × 10 11 m 3 by substituting the appropriate values quoted above. As- were carefully analyzed in order to identify and correct for potential issues with the data.

480
Concerning the pressure and temperature measurements, cruise reports show that the data 481 accuracy is σ ctdT = 1 db in pressure and σ ctdT = 1 × 10 −3 • C in temperature.

482
CTD conductivity sensors, particularly in the early cruises, were subject to drift and 483 imprecisions in the measurements, requiring that the CTD data are calibrated against con-484 ductivity measured in water samples collected at different depths during station occupations.

485
Twelve water samples covering the whole water column were typically taken at each station.

486
Conductivity/salinity was measured for each water sample using a Guildline Autosal 8400B of these errors, we introduce in each temperature, salinity and pressure field an error: where σ ctdS , σ ctdT and σ ctdP are normally-distributed random errors with mean equal to 520 zero and standard deviation equal to 2.8 × 10 −3 , 2.2 × 10 −3• C and 1 db respectively. The 521 21 same gridding procedure is then applied using the biased fields and the mean AABW salinity 522 and thickness are then calculated and compared with our time series. The mean regridding 523 errors calculated are σ rgS = 6 × 10 −4 (std = 6 × 10 −4 ) and σ rgT h = 30.7m (std 37.5 m) for 524 the mean AABW salinity and thickness respectively.

525
The final error to consider is the standard error of the mean salinity, σ se , defined as: where s is the standard deviation of the salinity mean for a given cruise and n is the  The total error on salinity is σ S = σ 2 eS + σ 2 rgS and on thickness is σ T h = σ 2 eT h + σ 2 rgT h .

530
The errors are plotted as error bar on Figure 4. The error analysis reveals that the observed 531 salinity trend is larger than the potential errors coming from CTD measurement errors and 532 their subsequent propagation through the derived quantities presented in this study.