1.1. Significance of Atmospheric Iron Deposition to the Southern Ocean
The availability of iron (Fe) is known to limit the growth rate and biomass of phytoplankton over much of the modern-day Southern Ocean [Martin et al., 1990; Boyd et al., 2000; Coale et al., 2004; de Baar et al., 2005]. This vast circumpolar region has extremely low surface water Fe concentrations and receives the lowest atmospheric Fe input of the world's oceans [Duce and Tindale, 1991; Jickells et al., 2005]. The low concentrations of iron that are available to phytoplankton in surface waters of the present-day Southern Ocean surface are largely supplied by the upwelling of deeper waters [de Baar et al., 1995; Fung et al., 2000; Watson et al., 2000]. However, the transport of iron-bearing dust to the oceans is thought to have been much greater during glacial periods of the late Quaternary, which may have enhanced the supply of biologically available iron to the Southern Ocean during those times [De Angelis et al., 1987; Martin, 1990; Petit et al., 1999; Mahowald et al., 1999; Ridgwell and Watson, 2002; Wolff et al., 2006]. On the basis of the inverse correlation between atmospheric dust loads and atmospheric CO2 levels inferred from ice core records, Martin  proposed the glacial iron hypothesis, which contends that enhanced atmospheric iron deposition during the last glacial age stimulated oceanic export production, primarily in the Southern Ocean, thus affecting an 80 to 90 ppm decrease in atmospheric CO2 concentration relative to interglacial conditions.
Major arguments that have been leveled against this hypothesis concern (1) the relative timing of atmospheric dust and CO2 changes as recorded in ice cores, which indicate significant differences in the timing of decreases in dust deposition versus increases in atmospheric CO2 at glacial terminations [Broecker and Henderson, 1998; Wolff et al., 2006; Gaspari et al., 2006]; (2) the magnitude of atmospheric CO2 drawdown resulting from plausible increases in Southern Ocean new production, which modeling studies have estimated at less than 10 ppm CO2 [Lefèvre and Watson, 1999; Popova et al., 2000]; and (3) the flux of iron-bearing dust to the ocean during glacial times, which may have been considerably less than that required by Martin's iron hypothesis [Maher and Dennis, 2001]. However, more recent modeling studies have simulated a significant drawdown in atmospheric CO2 (>10 ppm) due to elevated Southern Ocean export production forced by increased iron deposition during glacial times, with the relative timing of simulated changes in atmospheric dust and CO2 levels in good agreement with ice core records [Watson et al., 2000; Ridgwell and Watson, 2002; Ridgwell et al., 2002; Bopp et al., 2003]. Furthermore, increased Si input to the surface ocean and/or changes in the Si:N uptake ratio of phytoplankton due to enhanced iron availability afford additional mechanisms for reducing atmospheric CO2 during glacial periods, via a decrease in the rain ratio of calcium carbonate to organic carbon [Brzezinski et al., 2002; Matsumoto et al., 2002; Ridgwell et al., 2002].
An important conclusion of these modeling studies is that both the magnitude and timing of simulated glacial-interglacial atmospheric CO2 changes are highly sensitive to the magnitude and time variation of atmospheric iron deposition to the Southern Ocean. The model simulations of Watson et al.  and Ridgwell and Watson  are informed by an atmospheric dust deposition that follows the Vostok ice core dust record [Petit et al., 1999] between maximum (Last Glacial Maximum) and minimum (Holocene) deposition estimates derived from the general circulation dust-transport model of Mahowald et al. , with the assumption that the dust contains 3.5% iron by mass of which 2% is readily soluble (i.e., biologically available to marine phytoplankton). The resulting estimates of atmospheric iron deposition contain significant uncertainties, particularly in the model-derived Holocene and Last Glacial Maximum (LGM) dust deposition fluxes, which may significantly overestimate deposition in the Antarctic [Mahowald et al., 1999; Ridgwell and Watson, 2002; Bopp et al., 2003], and in the estimated solubility of aerosol iron, which may exceed 10% for dust deposited in and on Antarctic snow [Edwards and Sedwick, 2001]. There is thus a pressing need for accurate estimates of the past and present atmospheric deposition of biologically available iron in the Southern Ocean, to evaluate the role of this flux in regulating oceanic algal production and the ocean-atmosphere CO2 balance.
1.2. Records of Atmospheric Iron Deposition From Antarctic Ice Cores
Model-derived estimates of past and present eolian iron deposition to the Southern Ocean contain large uncertainties [Tegen and Fung, 1994; Mahowald et al., 1999, 2005; Moore et al., 2000; Prospero et al., 2002; Jickells et al., 2005], whereas the records of eolian dust derived from deep-sea sediment cores [e.g., Rea, 1994; Kumar et al., 1995; Hesse and McTainsh, 1999] may be obscured by physical and chemical processes in the water column and sediments. The most reliable estimates of atmospheric iron deposition to the Southern Ocean are likely to be those derived from the analysis of modern and ancient Antarctic snow. Measurements of iron in snow collected from Antarctic pack ice and coastal sites have been used to infer the present-day fluxes and solubility of eolian iron to maritime East Antarctica [Edwards and Sedwick, 2001], whereas records of past atmospheric iron deposition have been derived from analysis of aluminum (Al) [De Angelis et al., 1987], dust particles [Petit et al., 1999] or iron [Edwards et al., 1998; Traversi et al., 2004; Gaspari et al., 2006, Wolff et al., 2006] in the ancient snow of Antarctic glacial ice cores. Direct measurements of iron in snow and ice are expected to provide the most robust estimates of eolian iron deposition, because differences in sources and/or atmospheric transport of mineral aerosols can result in variations in Fe/Al or Fe/particle ratios, relative to average crustal values [Zoller et al., 1974; Arimoto et al., 1995; Holmes and Zoller, 1996; Watson, 1997; Hinkley and Matsumoto, 2001; Jickells and Spokes, 2001; Prospero et al., 2002; Bay et al., 2004; Delmonte et al., 2004b, 2004c; Gaspari et al., 2006].
Gaspari et al.  and Wolff et al.  have recently published estimates of eolian iron fluxes for Antarctica spanning the Last Deglaciation (LD) and the last eight glacial cycles, respectively, based on the analysis of iron in the EPICA ice core, which was drilled at Dome C on the inland plateau of East Antarctica. Gaspari et al.  estimate that atmospheric deposition of acid-dissolvable iron at the LGM was more than 30-fold higher than during the early Holocene. This recent work represents an important advance in our knowledge of past atmospheric iron deposition to Antarctica, and in our ability to evaluate the impact of changes in atmospheric iron deposition of algal production in the surface ocean. However, it is important to recognize that there may be significant meridional and zonal gradients in the deposition of iron-bearing mineral dust to Antarctica and the Southern Ocean [Mahowald et al., 1999, 2006; Lunt and Valdes, 2002; Delmonte et al., 2004a, 2004b]. For example, Edwards and Sedwick  reported order-of-magnitude differences in the estimated seasonal deposition fluxes of acid-dissolvable iron between different sectors of maritime East Antarctica, based on the analysis of iron in modern snow samples. Hence ice core records from multiple coastal locations around the Antarctic continent are required to better constrain the past and present atmospheric deposition of iron to the surrounding Southern Ocean.
In this paper, we present results from the analysis of iron in sections of glacial ice cores from Law Dome, in maritime East Antarctica, which extend the preliminary data set presented by Edwards et al. . By combining our iron concentration data with estimates of snow and ice accumulation rates for these core sections, we calculate atmospheric iron deposition fluxes for this location during recent times, the early and late Holocene, the LD, and the LGM. Our data thus provide information on the temporal variability of eolian iron deposition at Law Dome over timescales ranging from seasons to millennia, albeit at rather coarse resolution. Given that this site is exposed to predominantly maritime atmospheric circulation, these flux estimates are likely to be representative of atmospheric iron deposition over the adjacent polar waters of the Southern Ocean.