(2012) to derive a simplified wind forcing for our model. For this derivation, the RACMO2 data is compared to observations from an automatic weather station (AWS) that was operational from January 2010 to January 2012 on the FIS at the location indicated in Fig. 2(a). Fig. 4(a) shows the time series of the 48-h low-pass filtered zonal wind component obtained from the AWS together with the atmospheric simulations (interpolated to the same location) that were available at

the time when the simulations for our study were set up. RACMO2 convincingly captures MAPK inhibitor the timing and magnitude of the major wind events observed on the FIS, whereas more quiet periods and reversing westerly winds are generally less well reproduced by the simulations. Both time series also show a primarily high-frequency variability of the zonal wind stress, with no clear seasonal cycle in wind strength or frequency of Epigenetics Compound Library concentration storm events (not shown) being apparent during the observational period. We also note that there appears to be no obvious connection between the variability of the winds and the warm pulses seen beneath the FIS apparent in Fig. 4(b), discussed in more detail shortly. Additional uncertainty in the wind forcing is added by sea ice that modulates the momentum transfer

from the atmosphere into the ocean. In the FIS region, only small amounts of land-fast ice, which would entirely block the transfer of momentum onto the ocean surface, are found (Fraser et al., 2012). But also the seasonally varying ice cover, illustrated by the gray line (right axis) in Fig. 4(a) (Spreen et al., 2008), of predominantly drifting ice alters the momentum transfer, possibly introducing seasonal variations to the ASF current strength (Nunez-Riboni Flavopiridol (Alvocidib) and Fahrbach, 2009). This effect is difficult to assess, because ice drift may either increase or decrease the momentum transfer depending on its properties (Lüpkes and Birnbaum, 2005). Thus, the simplest approach for

our process-oriented study is to neglect the effect of sea ice and to compute the climatological mean ocean surface stress (τu,τv)(τu,τv) directly from the RACMO2 “2 m” winds (u,v)(u,v) as τu=ρaCau2+v2u,andτv=ρaCau2+v2vwith the density of air being ρa=1.4ρa=1.4 kg m−3, and with a drag coefficient of Ca=1.3×10-3Ca=1.3×10-3 at the air–ocean interface (Smith, 1988). In addition, the model sensitivity to different surface stress fields will be explored by a set of idealized forcings described in Section 3.4. Essential datasets for evaluating our simulations are provided by Hattermann et al. (2012), who presented sub-ice shelf observations acquired via three hot-water drill holes denoted M1, M2, and M3 in Fig. 2(a) (see supplementary material).