2 Methods
2.1 The Community Earth System Model Version 1
All simulations were performed using the ocean and sea-ice components of the Community Earth System Model version 1 (CESM1) [Danabasoglu et al ., 2012] from the National Center for Atmospheric Research. Ocean and sea ice were forced with surface fluxes from the European Center for Medium-Range Weather Forecast ERA5 reanalysis for the 1951-1980 period [Hersbach et al ., 2020]. A horizontal resolution of ~1o (grid gx1v6) was used for both ocean and sea-ice components. The ocean is divided into 60 unevenly-spaced levels, with thicknesses ranging from 10 m at the surface to 250 m in the deep ocean. All simulations were initialized with temperature and salinity fields from the Polar Hydrographic Climatology [Steele et al ., 2001]. A spin-up approach fromLee et al., [2017, 2019 ] is followed here where each year in the model is forced with randomly chosen historic fields from ERA5.
Previous work showed that most of the widely used Earth System Models inaccurately represent AABW formation. A possible reason for this misrepresentation is that most models’ current horizontal and vertical resolutions are too coarse to reproduce the downslope flow of dense waters off the Antarctic shelf and its mixing along the way, an essential process to form AABW. Coarse-resolution models exhibit excessive mixing of dense shelf waters with lighter surface waters, decreasing the density of shelf water and inhibiting AABW formation on the shelf [Heuz é et al , 2013]. These models create AABW through excessive deep convection in the open ocean, resulting in bottom waters that are often too fresh [Heuzé 2021]. Nevertheless, since the addition of shelf overflow parameterizations in the Southern Ocean [Danabasoglu et al ., 2012], AABW in CESM1 is properly formed on the continental shelf [Heuz é, 2021], making CESM1 an appropriate model to study AABW sensitivity to coastal freshwater fluxes.
2.2 Water mass definition
We define AABW as the waters with a neutral density higher than 28.27 kg m-3 [Orsi et al., 1999 ]. AABW salinity was calculated as the average salinity of AABW waters south of 60oS. The Southern Ocean was divided into five sectors (Fig 1a) according to Parkinson and Cavalieri [2012 ]: the Weddell Sector, Indian Sector, West Pacific, Ross Sector, and Amundsen Sector. The AABW transport in the Southern Ocean was measured as the absolute value of the minimum of the streamfunction at 65oS in density coordinates. The confidence interval for the AABW transport was obtained from the 95% confidence level of the yearly-averaged AABW transport estimates from the last 100 years of each simulation. Finally, although basal melting occurs in both ice shelves and under icebergs, in this study the term basal melting specifically refers to melting under ice shelves.
2.3 Freshwater distribution simulations
We carried out three model simulations to test to what extent AABW formation and salinity are affected by the spatio-temporal distribution of Antarctic meltwater fluxes. The first experiment (UNIF. , as in uniform) was forced with a freshwater flux of 2075 Gt/yr added uniformly in the ocean grid points closest to the Antarctic coast (Fig. 1a, Table S1). The flux magnitude of 2075 Gt/yr is based on total ice mass loss estimates from Rignot et al, [2019], and a full derivation of the value can be found in Supplementary material S2. The freshwater flux field of UNIF represents the surface meltwater fluxes suggested by OMIP1 and does not account for the spatial variation in freshwater fluxes from calving or basal melting.
The second simulation (BM ) was forced with zonally varying fluxes to mimic the spatial variations in AIS basal melting (Fig. 1c). The third simulation (VARI , Fig 1b) was forced with a freshwater flux that mimics the spatial variation of both ice sheet basal melting and calving (Supplementary S1). The spatial distribution of basal melting takes into account the meltwater production of Antarctic ice shelves estimated by Rignot et al [2013], while iceberg melting distribution is based on satellite-tracked iceberg positions from 1979 until 2017 (Supplementary S2). VARI differs from BM in the meridional distribution of freshwater fluxes, i.e., by having part of its freshwater fluxes displaced offshore, as seen by the higher melting fluxes offshore (Fig 1d, northward distances from coast larger than 3o of latitude), and lower melting fluxes along the coast (Fig 1d, distances lower than 3o) in VARIcompared to BM . The spatially varying freshwater flux fields used in VARI and BM simulations were produced according to the method discussed in Hammond and Jones [2016].