Although shelf seas account for only 7% of the oceanic surface area, recent observations demonstrate that they host significant ocean-atmosphere CO2 fluxes. A mechanism implicated in driving a significant CO2 sink in the temperate shelf seas is the Continental Shelf Pump. Here we present an analysis of the impact of sea-level rise, and the consequent flooding of continental shelves, on the growth of the continental shelf CO2 pump over the last deglacial transition. We combine reconstructions of shelf palaeogeography, bathymetry and tides, with contemporary shelf sea – atmosphere CO2 flux estimates, to demonstrate the potential of the expanding shelf seas to have impacted on the global carbon cycle, via the continental shelf CO2 pump, over the past 21,000 years and, by inference, earlier glacial-interglacial cycles.
 Ice core records show that during interglacial stages such as the pre-Industrial Holocene, atmospheric pCO2 was typically 280 ppmv, whilst during peak glacial times, such as the Last Glacial Maximum (LGM), atmospheric pCO2 was 180–200 ppmv, some 80–100 ppmv lower [Monnin et al., 2001] (Figure 1a). The change equates to a rate of increase in atmospheric pCO2 over the last deglaciation of on average 0.01 ppmv a−1 [Monnin et al., 2001]. Whilst the flooding of the shelf seas was long ago discarded as a potential first order mechanism in regulating atmospheric pCO2 [Broecker, 1982], recent measurements which demonstrate the magnitude of the contemporary air-sea CO2 fluxes observed in shelf seas cannot be ignored and merit more detailed attention.
 Over the last 21,000 years glacio-eustatic sea level has risen by 130 m [Lambeck et al., 2002] resulting in continental shelf submergence and a massive expansion of the surface area of the shelf seas (Figure 1a). Although this flooding of the continental shelves has been described as the most important geological event of recent time [Newell, 1961], there has been limited consideration of the impact of this event on changes in atmospheric pCO2, and these have focused on the impact of sea-level rise on the terrestrial carbon reservoir [e.g., Montenegro et al., 2006]. In this paper we analyse the potential impact of the flooding of the continental shelves and the creation of the shelf seas through extension of the oceanic biological CO2 pump.
 Despite the limited extent of the shelf seas compared to the global ocean, significant atmosphere-ocean CO2 fluxes have recently been inferred, via seasonally integrated ΔpCO2 measurements combined with standard exchange parameterisations, for a number of continental shelf seas [Tsunogai et al., 1999; Thomas et al., 2004; Borges et al., 2005; Cai et al., 2006; Bates, 2006]. For example, the temperate seasonally stratified North Sea is found to be a significant sink absorbing 8.5 × 1012 g C a−1, which is postulated to be exported to the deep ocean via a “continental shelf pump” (CSP) [Thomas et al., 2004]. The seasonally stratified northern North Sea is found to be particularly effective at CO2 uptake, whilst the shallow regions to the south, which remain well mixed throughout the year, tend to be weak sources [Thomas et al., 2004].
 Similar mechanisms are implicated in accounting for the observed net annual uptake of CO2 by other temperate and polar shelf seas [Tsunogai et al., 1999; Bates, 2006]. Extrapolating these results to all temperate and polar continental shelf seas globally equates to a net carbon drawdown equivalent to a reduction in atmospheric pCO2 of 0.17 – 0.23 ppmv a−1 [Borges et al., 2005; Cai et al., 2006]. Global circulation model experiments incorporating a crude shelf parameterisation demonstrate that the CSP plays a potentially important role in the global carbon cycle [Yool and Fasham, 2001].
 In this paper we assess the potential impact of the flooding of temperate shelf seas, since the LGM, on the global carbon cycle, via the CSP, by combining reconstructions of shelf paleogeography, bathymetry and tides with contemporary shelf sea-atmosphere CO2 flux estimates, with particular focus on the northwest European shelf seas. We then extrapolate the result globally, concluding with a discussion of uncertainties in the calculations presented.
2. Data and Methods
 In order to assess the impact of the expanding shelf seas on atmospheric pCO2 over the last deglacial transition (0–21 ka BP), we combine palaeotidal model (PTM) predictions parameterized using glacial isostatic adjustment (GIA) model output defining shelf palaeogeography and bathymetry [Peltier, 1994, 2004] with contemporary estimates of seasonally averaged shelf sea – atmosphere CO2 exchange [Borges et al., 2005; Cai et al., 2006; Thomas et al., 2004]. Initially we focus on the northwest European shelf (Figure 1a). We assume that the air-sea CO2 exchange is proportional to the appropriate shelf area, an assumption supported by contemporary observations [Thomas et al., 2004; Frankignoulle and Borges, 2001].
 We reconstruct the shelf with a temporal resolution of 0.5 ka for 0–17 ka BP and 1 ka for 18–21 ka BP using the ICE5-G GIA model output [Peltier, 2004]. These data have been converted from the original 1 degree resolution to degree resolution and combined with the present day bathymetry adopted for the deep ocean to construct a set of palaeobathymetries, and thus shelf areas, for selected timeslices since the LGM [Uehara et al., 2006].
 The CSP only operates in regions subjected to seasonal stratification, with the other (seasonally well mixed) regions acting as a weak source of CO2 to the atmosphere [Thomas et al., 2004]. In order to subdivide the shelf into seasonally stratified and well mixed sectors, the PTM is parameterized to calculate the position of the transition zones, known as tidal mixing fronts. Simpson and Sharples  show that the energetics of vertical exchange is the dominant control on frontal position, so we use the energetics criterion of Simpson and Hunter  to predict the position of tidal mixing fronts. This criterion is based on a balance between buoyancy input, as heat, and tidally driven turbulent mixing. The location of the front is dependent on a number of parameters: ɛαQH/(cDuT3) where α is the thermal expansion coefficient, Q is the rate of heat input, ɛ is the mixing efficiency, H the water depth, cD the bottom drag coefficient and tidal current amplitude uT. All these parameters will vary for any particular location over the glacial cycle.
 Sensitivity tests show that the positions of fronts are relatively insensitive to variations in α and Q over the last 21 ka (variations in Q are based on work by Berger and Loutre  and α with values from Schumacher et al. ). Clearly buoyancy input through ice melting, ice formation and ice cover will vary considerably during deglacial transitions. However, comparison of predictions of frontal positions in the contemporary northwest European shelf seas and the arctic Bering Sea show that whilst the timing of the onset of stratification is influenced by meltwater and ice cover [Schumacher et al., 1979], the capacity of the model to predict frontal positions is unaffected, implying that a constant mixing efficiency may be assumed for the entire deglacial transition. The major factors determining the position of tidal mixing fronts are therefore water depth, H and tidal current amplitude, uT. We use the stratification parameter:
with a value S = 1.5 as defining the frontal position, with values of 0.87 < S < 1.83 representing the transitional zone between mixed and stratified [Pingree and Griffiths, 1978].
 The PTM reconstructs tidal amplitudes and dynamics using a 2D version of the Princeton Ocean Model for the northwest European shelf seas forced at the ocean boundaries using a global tidal model output (see Uehara et al.  for details). The timings of the temporal migration of the predicted tidal mixing frontal positions in the open boundary runs are supported by empirical palaeodata (benthic foraminiferal assemblages, oxygen/carbon foraminiferal stable isotopes) from shelf basins [Austin and Scourse, 1997; Scourse et al., 2002].
 We apply this analysis to temperate (latitude range 30–60°) shelf seas globally. In order to assess the impact of the evolving shelf seas on the continental shelf CO2 pump, average seasonal CO2 fluxes, based on contemporary measurements for the North Sea, are then applied: an uptake rate of 30 gCm−2a−1 in the seasonally stratified zones, and an outgassing rate of 6 gCm−2a−1 in the well-mixed zones (following Thomas et al ).
3.1. Evolution of Tidal Mixing Front Position
 The extent of the northwest European shelf seas, together with the predicted positions of the tidal mixing fronts for a number of time slices is shown in Figure 1. The results show that, at present, approximately 80% of the northwest European shelf area is subject to seasonal stratification, with seasonally well mixed conditions prevailing only in the North Channel and eastern and southern sections of the Irish Sea, in the English Channel and the southern North Sea. The major tidal mixing frontal zones are the Flamborough Head Front in the North Sea, the Celtic Sea and Ushant Fronts to the west of the Irish Sea and English Channel respectively, the Western Irish Sea Front and Malin Front, west of Scotland.
 At the LGM (21 ka BP) all that existed of the northwest European shelf seas was a relatively narrow zone of the present deep outer continental shelf oriented north-south to the west of the United Kingdom. This zone was shallow, and swept by very strong tidal currents and so remained well mixed throughout the year.
 By 14 ka BP, immediately following Meltwater Pulse 1A [Weaver et al., 2003], much of the Celtic Sea and Malin shelf areas of the northwest European shelf, and parts of the northern North Sea, had submerged. However, the PTM output predicts tidal currents to be sufficiently strong to maintain well mixed conditions with both the Malin and the Celtic/Ushant Fronts corresponding approximately to the continental shelf break. Consequently the extent of seasonal stratification remained very limited.
 The prediction for 10 ka BP, towards the end of the period of major sea-level rise, shows that, with the exception of the southern North Sea, much of the current area of the northwest European shelf seas had submerged. However, on account of the shallower water and stronger tidal currents, the proportion of the shelf sea which seasonally stratified was more limited than it is today with, for example, much of the Celtic Sea and southern Irish Sea remaining mixed throughout the year.
 An important result with respect to the growth in the strength of the CSP is the progressive increase in the area of the shelf seas which stratify seasonally. The results presented here illustrate that the growth of the CSP is not simply a function of the increase in the area of the shelf seas, but is determined by the combination of water depth increase and tidal current regime in controlling seasonal stratification.
3.1.1. Growth of the Continental Shelf CO2 Pump
 The hindcasts (Figure 2b) predict that over the last deglacial transition the temperate shelf seas globally act as a net sink for atmospheric CO2, with the uptake rate increasing by 3–4 times between 17 and 8 ka BP. To highlight the periods of greatest change, the difference in CO2 uptake/emission between successive reconstructions has been plotted for each zone (Figure 2c).
 A significant increase in the global shelf sink is predicted at 14.5 ka BP, equivalent (if considered in isolation) to a decrease in atmospheric CO2 of ∼5 ppmv over the 500 year resolution of the model (corresponding to transition II [Monnin et al., 2001]). The most significant increase in the CO2 sink is predicted over the period 11 to 8 ka BP, equivalent to a rate of reduction of atmospheric pCO2 of 0.01 ppmv a−1, coeval with a stabilisation and then the largest decrease in atmospheric pCO2 during the entire deglaciation [Monnin et al., 2001].
 To assess the role of seasonal stratification we compare pump strength predictions with an estimate in which the effects of stratification are ignored, and a uniform CO2 drawdown is applied (i.e. the pump strength is assumed proportional to the area of the shelf sea). This result (Figure 2d) emphasises that during the early period of growth of the shelf seas, between 17 and 12.5 ka BP, the combination of shallower water and enhanced tidal currents resulted in a larger proportion of the temperate shelf seas remaining mixed throughout the year and so the CSP remains weak. As water depths increased the area subject to seasonal stratification grew proportionally faster than the shelf sea area as a whole. The CSP therefore initiated later and grew at a faster rate than would be the case if the effects of stratification were ignored.
4. Discussion and Conclusions
 Our results imply that, as sea level increased following the LGM, the temperate shelf seas expanded and their impact on atmosphere-ocean CO2 exchange increased through an extension of the oceanic biological carbon pump (CO2 sink) via a continental shelf pump. The results also highlight the importance of the inclusion of seasonal stratification in determining the strength of the CSP.
 For the purposes of our analysis we have used contemporary estimates of CSP strength, which are likely influenced by recent anthropogenic activity [Cai and Dai, 2004]. In particular anthropogenic nutrient discharges probably influence the level of carbon fixation through export production by influencing both the ecosystem structure and the availability of (limiting) nutrients in the euphotic zone. Estimates of prehistoric primary production in two coastal regions today heavily influenced by man suggest that present day primary productivity may be between two and five times higher than in prehistoric times [Nixon, 1997; van Beusekom, 2005]. These values therefore suggest that the pre-Industrial CSP estimates given here, although the correct order of magnitude, may be overestimates.
 Further measurements of modern air-sea CO2 fluxes in shelf seas are required to strengthen the regional source-sink associations utilised in this analysis, together with a full assessment of changes in both shelf-ocean exchange (which drive the import of limiting oceanic nutrients and the export of carbon) as well the anthropogenic impact on the strength of the CSP. However, we offer here an analytical approach for assessing sea-level driven pCO2 impacts which can accommodate integrations of new air-sea CO2 flux data as they become available.
 T.P.R. was a NERC Advanced Fellow, J.D.S. was a Royal Society-Leverhulme Trust Senior Research Fellow 2008-2009 and S.M. was in receipt of a NERC studentship for her M.Sc. course. K.U. was supported by a Royal Society-JSPS exchange programme.