Basal melting of ice shelves is causing accelerating mass loss from the Antarctic Ice Sheet, yet the oceanographic processes which drive this are rarely observed. This thesis uses new observations from phase sensitive radar and moored oceanographic instruments to describe the processes which drive rapid basal melting of the north-western Ross Ice Shelf.

Oceanographic conditions at the mooring site are strongly influenced by the neighbouring Ross Sea Polynya. High Salinity Shelf Water fills the lower water column continuously, but during summer a southward flow ventilates the cavity bringing Antarctic Surface Water (AASW) to the site. Tides account for half of the flow speed variance, and low frequency variability is influenced by local winds, and eddies associated with sea ice production in the polynya.

Four years of basal melt rate observations show a mean melt rate of 1.8 m y$\mathrm{<mrow\; data-mjx-texclass="ORD">-1</mrow>}$ at the mooring site and a strong seasonal cycle driven principally by water temperature variations. Radar observations show that melt rates vary rapidly and continuously in response to flow speed variability, and rapid melting occurs only when flow speeds are high.

Radar observations of melt rates from 78 sites on the Ross and McMurdo ice shelves show an area-averaged annual-mean basal melt rate of 1.35 m y$\mathrm{<mrow\; data-mjx-texclass="ORD">-1</mrow>}$, implying a net basal mass loss of 9.6 Gt y$\mathrm{<mrow\; data-mjx-texclass="ORD">-1</mrow>}$ from the region. Melt rates are highest near the ice front where annual-mean and short-term summer rates reached 7.7 m y$\mathrm{<mrow\; data-mjx-texclass="ORD">-1</mrow>}$ and 53 m y$\mathrm{<mrow\; data-mjx-texclass="ORD">-1</mrow>}$, respectively. The seasonal and spatial variations in melt rate are consistent with melting driven by the summer inflow of AASW.

Observations of boundary layer water temperature, flow speed and melt rates indicate that melt rates scale linearly with current speed, but sub-linearly with temperature in the outer boundary layer, possibly due to the stabilising effects of melt water input. Existing melt rate parameterisations which account for flow speed can be tuned to match the observations when thermal driving is low, but overestimate melt rates at higher temperatures, implying the need for further refinements to the models.