4 Discussion and conclusions
Most widely-used sea-ice and ocean models represent AIS meltwater fluxes by adding spatially uniform freshwater fluxes along the Antarctic coast, a protocol suggested by OMIP1. However, such uniform freshwater flux representation contrasts with the spatially varying melting patterns observed for the AIS [Rignot et al ., 2013; 2019;]. This misrepresentation creates surface salinity biases that can propagate into AABW properties [Jongma et al, 2009 ]. In this study, four different freshwater flux experiments were performed to better understand the AABW response to the location (uniform and varying) and to increasing AIS melting. Our results show that AABW salinity is sensitive to where the meltwater flows around the Antarctic continent (i.e., zonal distribution of freshwater fluxes). Varying the AIS melting fluxes zonally according to observations (VARI , Fig 1b,d) produces a 1.9±0.6×10-3PSU increase in AABW salinity compared to a uniform meltwater flux. In VARI , freshwater flux is the largest in the Amundsen sector (Fig 1d) and smaller in the remaining regions along the Antarctic coast. As a result, surface salinity is increased across the Weddell and Ross Seas in VARI compared toUNIF (Fig 2a,d,f). The positive salinity anomalies in the Weddell and Ross Seas propagate to the bottom layer by the shelf overflow. Since no coastal bottom water formation is simulated in the Amundsen sector [Danabasoblu et al, 1999], the freshening signal in this sector does not propagate into the bottom waters (Fig 2f). The higher AABW salinity found in the experiment with spatially varying freshwater fluxes (VARI ) has important implications for the representation of AABW in global ocean models. Specifically, to avoid the low salinity bias in AABW imposed by uniform AIS melting a better practice is to force ocean & sea-ice models with zonally varying fluxes that mimic the observed basal melting and calving distribution as in VARI . This practice is also consistent with the surface meltwater flux protocol suggested in OMIP phase 2 [Tsujino et al 2020].
In contrast, the AABW transport in the Southern Ocean is insensitive to the spatial distribution of AIS melting. Any decrease in the eulerian component of the overturning in the Southern Ocean in response to the freshwater fluxes is compensated by an increase in the eddy-induced component of the overturning (Fig S2). Eddy compensation of the overturning has been reported in experiments with altered Southern Hemisphere Westerlies [Bishop et al., 2016]. Here, we report that eddy compensation can also occur under small surface freshwater perturbations. Note that the CESM1 representation of the ocean overturning and the Southern Ocean circulation is similar in 100 km horizontal resolution (used here) and eddy-rich (10 km) resolution [Hewitt et al., 2020]. Therefore, eddy compensation to freshwater fluxes is unlikely an artifact of oceanic eddy parameterizations. Furthermore, in VARI and UNIF , AABW transport was evaluated for equilibrium conditions. Consequently, the similar AABW transport in these simulations can also be caused by salt flux compensation from sea ice expansion - similar to what happens in the 20% increase experiment.
Finally, observations show that AABW salinity decreased since the 1980s [Menezes et al ., 2017; Purkey et al ., 2012]. To answer if AABW freshening in the Southern Ocean could be driven by increased AIS melting, we explored the sensitivity of AABW transport and salinity to a 20% increase in AIS melting (415Gt/year). The short-term (30 years) AABW freshening in CESM1 in response to the increased freshwater flux is similar to observed trends in the Southern Ocean during the past decades. The average Southern Ocean AABW freshening of 3.0±0.2×10-3 PSU/decade simulated in this study lies within the range of freshening rates estimated from observations (i.e., 1.0×10-3~5.0×10-3PSU/decade) since 1980 [Purkey and Johnson 2012; Anikulmar et al., 2020]. Regional freshening rates also agree with observations. In particular, observations on the Indian and West Pacific sectors show freshening rates of 4.0±1.0×10-3PSU/decade [Menezes et al., 2017] and 3.1×10-3PSU/decade [Shimada et al., 2012], respectively, while simulated freshening rates for these sectors are 4.8±0.4×10-3PSU/decade and 3.4±0.5×10-3PSU/decade, respectively. The agreement between the observed and the simulated AABW freshening rates under a 20% increase in AIS melting suggests that increased AIS melting could be a fundamental driver of the observed AABW freshening.
Observations also show that sea ice in the Southern Ocean has expanded since 1974 at a rate of 33×103km2/yr. In our experiments, a 20% increase in freshwater fluxes from AIS caused sea ice to expand by 22×103km2/yr, roughly in agreement with observations. This similarity in sea-ice expansion rates suggests that enhanced AIS melting could be at least partially responsible for the observed sea-ice expansion trend. Although the mechanisms responsible to expand sea ice in our simulations are not explored in this study, previous studies have suggested that increases in meltwater fluxes from AIS can isolate the Southern Ocean surface from warm deep waters, triggering sea ice expansion [Mackie et al, 2020; Jeong et al, 2020].
It is important to highlight that the simulated impact of increased AIS melting on AABW salinity, transport, and sea ice is not permanent. After 60 years of enhanced AIS melting, both AABW salinity and GMOC transport are restored to their initial values (Fig 3a), suggesting that observed AABW freshening could yet be reversed by enhanced brine rejection from sea ice expansion. Nevertheless, the agreement of the observed sea ice and AABW trends with the simulated ones indicates that both of these trends could be triggered by a 415Gt/yr increase in AIS melting. Indeed, previous studies report an increase in AIS melting between 180Gt/yr and 480Gt/yr since 1978 [Rignot et al, 2019; Adsumilli et al 2020].
It is important to highlight that these results are constrained by the specific CESM1 configuration. For example, shelf overflow parameterizations had a critical role in propagating the surface salinity anomalies to the bottom layer in CESM1. Models without overflow parameterizations on the Antarctic coast form AABW by shelf or open ocean deep convection, which can dampen the surface freshening by mixing the surface freshwater with high-salinity deep waters. Eddy compensation in CESM1 also maintained AABW transport stable in the freshwater forcing experiments. However, the efficiency of eddy compensation can change under eddy parameterization schemes other than the one used in CESM1. Finally, ice shelves can reach up to 1 km down the water column, and as such freshwater fluxes from basal melting are not restricted to the surface as in our experiments. Ice shelf melting at deeper levels can cause coastal freshening rates different from those in our surface freshwater flux experiments [Pauling et al. , 2016]. Further studies are necessary to assess if specific model settings, such as eddy and shelf overflow parameterizations can determine AABW sensitivity to AIS melting.