Regional scale impacts of distinct CO2 additions in the North Sea

Marine Pollution Bulletin 56 (2008) 1461–1468 Contents lists available at ScienceDirect Marine Pollution Bulletin j o u r n a l h o m e p a g e : w w w . e l s e v i e r. c o m / l o c a t e / m a r p o l b u l Regional scale impacts of distinct CO2 additions in the North Sea J.C. Blackford a,*, N. Jones a, R. Proctor b, J. Holt b a b Plymouth Marine Laboratory, Prospect Place, Plymouth PL1 3DH, UK Proudman Oceanographic Laboratory, 6 Brownlow Street, Liverpool L3 5DA, UK a r t i c l e i n f o Keywords: CO2 Carbon capture and storage Sequestration North Sea Leakage Impacts pH a b s t r a c t A marine system model applied to the North West European shelf seas is used to simulate the consequences of distinct CO2 additions such as those that could arise from a failure of geological sequestration schemes. The choice of leak scenario is guided by only a small number of available observations and requires several assumptions; hence the simulations reported on are engineered to be worse case scenarios. The simulations indicate that only the most extreme scenarios are capable of producing perturbations that are likely to have environmental consequences beyond the locality of a leak event. Tidally driven mixing rather than air–sea exchange is identified as the primary mechanism for dispersal of added CO2. We show that, given the available evidence, the environmental impact of a sequestration leak is likely to be insignificant when compared to the expected impact from continued non-mitigated atmospheric CO2 emissions and the subsequent acidification of the marine system. We also conclude that more research, including both leak simulations and assessment of ecological impacts is necessary to fully understand the impact of CO2 additions to the marine system. © 2008 Elsevier Ltd. All rights reserved. 1. Introduction Emission of anthropogenically derived CO2 to the atmosphere and the subsequent uptake by the oceans, leading to climate change and ocean acidification, respectively, are both predicted to cause severe environmental, ecological and resource impacts (IPCC, 2001; Stern, 2006; Raven et al., 2005). Consequently there is much interest in developing methods for reducing carbon emissions, including carbon capture (from power stations) and its subsequent storage in geological formations. An active sequestration programme has been in operation at the Sleipner field in the North Sea since 1996 run by the Norwegian company Statoil. Here carbon dioxide is striped from natural gas by solvents and disposed of in a saline formation with approximately one million tonnes of CO2 sequestered each year. Further projects are planned for the North Sea, exploiting the large volumes of geological reservoirs in the region. Injection of carbon dioxide under high pressure into depleted reservoirs is also of financial interest as it may lead to enhanced oil recovery. The delivery and geological storage of large volumes of highly pressurised CO2 raises the concern of leakage and its potential environmental consequences to the marine system. A number of mechanisms of leakage are possible, fast flow events such as a pipeline failure, faulty injection well casings and transmissive faults or fractures in the cap rock; and slow flow phenomena such as seepage * Corresponding author. Tel.: +44 (0)1752 633468; fax: +44 (0)1752 633101. E-mail address: jcb@pml.ac.uk (J. Blackford). 0025-326X/$ - see front matter © 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.marpolbul.2008.04.048 through porous geological structures. Research is scarce but suggests that in the long-term only a small fraction of sequestered CO2 might escape (DTI, 2003 and references therein). However, given the possibility of leakage, it is prudent to assess the potential for causing environmental impacts and to compare this with the predicted environmental impacts of ocean acidification. 2. Methodology A marine system model (POLCOMS-ERSEM-HALTAFALL), describing the North West European continental shelf, is used to simulate the dynamics of added CO2 and it’s consequences in terms of the resulting perturbation in pH. The model system is as described in Blackford and Gilbert (2007) except for the extension to cover the whole of the North Western Shelf; salient details are briefly reviewed here. The hydrodynamic model POLCOMS is a three-dimensional baroclinic system described by Holt and James (2001) and Proctor and James (1996). It is a primitive equation finite difference model; solving for velocity, surface elevation, potential temperature, salinity and turbulent kinetic energy using spherical polar coordinates in the horizontal and s-coordinates (Song and Haidvogel, 1994) in the vertical. It employs a sophisticated advection scheme (the “Piecewise Parabolic Method”; James, 1996) to minimize numerical diffusion and ensure the preservation of features even on coarse grids under oscillatory flows. Turbulent viscosities and diffusivities are calculated using a Mellor–Yamada level 2.5 turbulence closure, but with an algebraically specified mixing length. The model is applied to the Northwest European shelf on an approximately 7 km grid 1462 J. Blackford et al. / Marine Pollution Bulletin 56 (2008) 1461–1468 with 18 s-levels giving a vertical resolution between 0.5 and 15 m depending on water depth. Holt et al. (2005) describes the detailed evaluation of the model against observations concluding that the model, with some exceptions, generally accurately describes the spatial and temporal variability in dynamic features of the region. ERSEM is a complex functional type ecosystem model describing carbon and nutrient flows through both pelagic and benthic lower trophic ecosystems (Blackford et al., 2004; Baretta et al., 1995). However the ERSEM model dynamics do not impact on the results presented here as there is no feedback between the altered CO2, pH and the ecosystem processes included in the model at this stage (see Section 3). HALTAFALL (Ingri et al., 1967) is an iterative chemical speciation model which, as applied here, uses calculated dissolved inorganic carbon (DIC, the sum of the chemical species resulting when CO2 dissolves in water) and parameterized total alkalinity (TA) to derive pH and the partial pressure of CO2 in the water. The latter is required to drive the air–sea flux calculation of CO2 which uses the parameterization of Nightingale et al. (2000). Sensitivity to air–sea flux parameterizations is discussed below. The model is forced by an assumed invariant atmospheric CO2 concentration of 375 ppm, riverine DIC inputs derived from Pätsch and Lenhart (2004) and Thomas et al. (2005) and an assumption of zero flux divergence for DIC at the lateral boundaries. We choose to investigate three modes of CO2 release, relating to the possible mechanisms of leakage. Parameterising the rate and duration of a leak event is obviously speculative; apart from the stochastic nature of such an event there is little information available to guide us towards realistic scenarios. We use two sources to guide our choice of leak scenario. Klusman (2003a, b) reports preliminary estimates of seepage from a terrestrial EOR – sequestration project in Colorado, USA of <3800 tonnes CO2 a¡1 over an area of 78 km2 with 14C measurements indicating rates of <170 tonnes CO2 a¡1. These estimates equate to 0.14–3.0 mmol m¡2 d¡1 which are the unit relevant to the model system. The Colorado site has accepted 23 £ 106 tonnes of CO2 since 1986. Secondly, we use the typical capacity of the pipelines used to deliver CO2 to well systems, 100–250 mscfd (million standard cubic feet per day). This equates to 1.34–3.15 £ 1011 mmol d¡1 or 1.60 £ 103–3.75 £ 103 tonnes C d¡1. An important consideration, principally relating to fast-rate leak events is the behaviour of the resulting high pressure CO2 plume; it’s rate of travel to the sea surface and the balance between direct gassing to the atmosphere and solution in the water column. There is evidence from natural shallow (<20 m) high pressure gas seeps that the majority of CO2 in bubble plumes can transfer to the water column (Leifer et al., 2006). Hence we assume for simplicity all CO2 from a leak is dissolved. For low pressure seepages we assume all gas is dissolved in the bottom layer, for high pressure leaks we assume an equal distribution of CO2 input through out the water column. Consequently, and after some sensitivity analysis we elected to report on the following scenarios, summarised in Table 1. i Long-term diffuse seepage: We assume a constant low level seepage of CO2, spread homogeneously across the area of one model box (49 km2), representing a movement of CO2 through permeable geological formations. We employ two seepage rates, 3.85 £ 10 0 mmol m¡2 d¡1 similar to the upper end of the Colorado observations (Klusman, 2003a, 2003b) and a £100 treatment of 3.85 £ 102 mmol m¡2 d¡1, giving a total input over one year of 3.02 £ 103 and 3.02 £ 105 tonnes CO2, respectively. ii Short-term leak: Analogous to a fracture in a pipeline that persists for one day. We use two inputs, 6.93 £ 103 and 6.93 £ 104 mmol m¡2 d¡1 giving total inputs of 1.49 £ 104 and 1.49 £ 105 tonnes CO2, respectively, about 5 and 50 times a typical pipeline capacity. iii Long-term leak: Analogous to say, an immitigable fault in the well casing, we assume a catastrophic out-gassing of 6.93 £ 103 mmol m¡2 d¡1 or 5.43 £ 106 tonnes CO2 over one year, five times the input rate at Sleipner, or 5 years worth of sequestered CO2. Our final assumption is that the point source leaks (ii and iii) disperse instantaneously into a single 7 £ 7 km model box. Clearly this is a weakness although the tidally driven horizontal mixing processes in the region are strong (Holt et al., 2001) and would be capable of achieving this mixing within a few days. All modes of release were simulated at two sites, North (57.75N, 1.00E), approximating to the Forties oil field – and South (54N, 1E), representative of the Viking group of oilfields. The former site is characterised by a water column depth of 138 m which is strongly stratified during the summer. The latter site has a depth of 28.5 m and is generally mixed throughout the year. The short-term leaks (ii) were simulated at four times during the seasonal cycle on Julian days 11, 101, 191 and 281, respectively, 11th January, 10th April, 8th July and 8th October. The scenarios used a four year spin-up simulation with annually repeating forcing conditions (weather and boundary forcing and atmospheric CO2 values fixed at 375 ppm approximating the Table 1 Simulated scenarios Scenario Seepagelow Seepagehigh Short-term leak-low Short-term leak-high Long-term leak Site North South North South North South North South North South Input duration days Depth (m) 365 365 365 365 1 1 1 1 365 365 7.7 1.6 7.7 1.6 138.0 28.5 138.0 28.5 138.0 28.5 Input concentration (mmol m¡3 d¡1) Daily input per metre square Daily input to model environment Total input CO2 (mmol m¡2 d¡1) Carbon (g m¡2 d¡1) CO2 (g m¡2 d¡1) Carbon (tonnes box¡1 d¡1) CO2 (tonnes box¡1 d¡1) Carbon (tonnes) CO2 (tonnes) 0.5 £ 10 0 2.42 £ 10 0 5.0 £ 101 2.42 £ 102 5.0 £ 101 2.42 £ 102 5.0 £ 102 2.42 £ 103 5.0 £ 101 2.42 £ 102 3.85 £ 10 0 4.60 £ 10¡2 1.68 £ 10¡1 2.25 £ 10 0 8.23 £ 10 0 8.23 £ 102 3.02 £ 103 3.85 £ 102 4.60 £ 10 0 1.68 £ 101 2.25 £ 102 8.23 £ 102 8.23 £ 104 3.02 £ 105 6.93 £ 103 8.28 £ 101 3.04 £ 102 4.06 £ 103 1.49 £ 104 4.06 £ 103 1.49 £ 104 6.93 £ 104 8.28 £ 102 3.04 £ 103 4.06 £ 103 1.49 £ 105 4.06 £ 104 1.49 £ 105 6.93 £ 103 8.28 £ 101 3.04 £ 102 4.06 £ 102 1.49 £ 104 1.48 £ 106 5.43 £ 106 Columns as follows: (3) the duration of the simulated input; (4) the water column depth receiving the added CO2 (for the seepage simulations the specified depths represent the bottom layer of the model); (5) the input concentration per cubic metre; (6–8) the daily input per metre squared (column 4 multiplied by column 5); (9 and 10) the daily input to the model environment (columns 7 and 8 multiplied by the area of input, 49.0 £ 106 m2); (11 and 12) the total input to the simulation (columns 9 and 10 multiplied by the input duration in column 3). J. Blackford et al. / Marine Pollution Bulletin 56 (2008) 1461–1468 1463 year 2000) to provide settled initial conditions. The only difference in the one year simulation following the spin up period is the addition of CO2 as detailed in Table 1. Forcing and boundary conditions are as described in Holt et al. (2005) and Blackford and Gilbert (2007). In addition a control scenario with no CO2 input was simulated to provide a baseline data set. The assumptions and scenarios chosen have been deliberately done to address, given current knowledge, extreme worst case scenarios of CO2 leakage, and thus set the upper boundaries of potential environmental impacts. The approximate average concentration of DIC in a typical marine system is approximately 2.1 mol m¡3 (Takahashi et al., 1981). In a 100 m water column this amounts to 2.1 £ 102 mol or 2.4 kg carbon; in a 7 km £ 7 km £ 100 m depth model box this equates to 1.2 £ 105 tonnes of carbon as DIC. 3. Results and discussion 3.1. Long-term seepage We restrict the reporting of results to the pH anomaly. This we take as the difference between pH computed for each leak scenario compared with the ‘normal’ pH field resulting from the control model run with no perturbation. Change in pH is not the only effect resulting from increased DIC that has the potential to modify ecosystem processes; however pH does provide an appropriate proxy for the strength of the sum of ecosystem effects. Much research is currently ongoing into the precise nature of ecosystem response to high CO2 but, because of the complexity of the processes involved and some pronounced inter species-differences this is as yet unquantifiable, although many individual process responses have been identified (summarised in Raven et al., 2005). In order to give some guidance we have elected to use the following colour scheme to illustrate the potential detrimental effect on the ecosystem, it should be noted that this is qualitative and subject to debate. The input rate of 3.85 mmol CO2 m¡2 d¡1, equating to the measured loss rate at the Colorado site represents a perturbation of 0.002% in a 100 m water column. This is insignificant and did not cause a detectable signal in the model. The £100 treatment (3.85 £ 102 mmol CO2 m¡2 d¡1) resulted in a small decrease of pH, relative to the control, with a maximum reduction of 0.12 pH unit in the vicinity of the sediment surface, with small perturbations propagating through the water column (Fig. 1). This is significantly less than the natural range of variability and we assume would not have any significant biogeochemical impact. Natural variability of pH in the region is relatively larger than for the general ocean given that it is driven by high rates of biological CO2 exchange and riverine inputs as well as temperature effects. Variability is typically between 0.2 and 0.4 pH units over the annual cycle, although extreme phytoplankton blooms can increase pH by over 0.5 units compared with background. Like all of the simulations reported here the perturbation in the shallower south site is larger than that in the north site because the same amount of CO2 is injected into a smaller volume of water. Also clearly visible, especially in the south site, are the effects of the tidal mixing cycle with summer neap periods significantly restricting mixing resulting in the largest simulated pH anomalies. I White: Perturbation zero or below detection levels, of no ecological significance. I Green: No or minimal effect likely, perturbation less than natural variability. I Yellow: Perturbation of the order of natural variability, potentially small impacts. I Orange: Some species and processes experiencing significant impacts. I Red: More wide ranging and significant to severe effects predicted. Fig. 1. Annual evolution of pH perturbation with depth of (£50) seep simulation. (a) north site and (b) south site. 1464 J. Blackford et al. / Marine Pollution Bulletin 56 (2008) 1461–1468 3.2. Short-term leak Fig. 2 details the perturbation arising from the smaller (£5 pipeline capacity) short-term leak scenario. For the north site (Fig. 2a) the pH perturbation is no more than ¡0.1 units for at most one day tailing off over a period of 3–8 days depending on mixing rates. In the south site the initial perturbation is as much as ¡0.2 pH units (as the same amount of CO2 is injected initially into a smaller volume) and again short-lived, with the signal disappearing over 5–9 days. Given current knowledge it is unlikely that a CO2 perturbation of this magnitude would have an ecosystem effect. In contrast the £50 pipeline capacity leak scenario (Fig. 3) provokes pH perturbations that exceed ¡0.5 pH units for about a day at the north site and up to five days at the south site. The duration of disturbance is less than 10 days for the northern site and as much as 20 days for the southern site. This scenario indicates the approximate scale of leakage required to provoke what may be serious environmental consequences, although there is not yet enough experimental data to confirm or refute this. 3.3. Long-term leak The results show, for both sites, a small area of high perturbation centred over the release (Figs. 4 and 5). In the north this perturbation does not exceed ¡0.5 pH units; in the south a perturbation of ¡1.0 pH units is recorded. The area of maximum disturbance in both cases remains well constrained, although a plume of acidified water is seen to spread from the release driven by the regional circulation. This plume can be extensive, although the majority of the plume area is acidified by significantly less than 0.1 pH units. Examination of the detailed model output (not shown) indicates the possibility of retentive features in circulation and mixing creating discrete small regions of lowered pH that move away from the release point and persist for a week or so. Examination of the perturbation at the leak location (Fig. 5) for both release sites clearly shows the influence of the tidal cycle in determining the instantaneous perturbation strength. Perturbation maxima are associated with neap tides and minima with springs and can differ by 0.4 or 0.8 pH units depending on site and wind strength. This suggests that not only are the timing of leaks an important consideration but experimental efforts to investigate ecosystem effects might need to consider using cyclically varying pH treatments rather than a constant pH value. 3.4. Dispersion of added CO2 Fig. 6 details the increase in sea to air out-gassing of CO2 for each scenario, at the point directly above each release site, along with the control simulation air–sea exchange rate that could be considered ‘normal’. The increased flux due to the seepage scenario is small but clearly apparent. The long-term leak scenarios increase this flux by between two-fold and two-orders of magnitude throughout the simulation. The short-term leak events provoke similarly short-term but large flux increases. The instantaneous air–sea exchange rate is driven by the episodic nature of the wind forcing fields. Examination of the air–sea CO2 flux model output (Table 2) indicates that by the end of each scenario between 54% and 63% of the CO2 addition has been lost to the atmosphere for the South site leaks, whilst at the North site this drops to between 12% and 27%. This site disparity is driven by the difference in the initial CO2 concentration following from the difference in volume of the input box between each site. In comparison with the short-term events, the lower percentage out-gassing in the long-term scenarios the can be attributed the zero time-lag between cessation of input and the air–sea exchange summation. However, in general out-gassing, in the short to medium term, is significantly less than the simulated CO2 inputs. This is indicative of the slowness of the air–sea exchange process in comparison with the speed of input events and mixing processes. Thus a significant amount of carbon dioxide Fig. 2. Short-term leak scenario (£5 pipeline capacity). (a) north site and (b) south site. All perturbations below 0.01 pH unit have been masked for clarity. J. Blackford et al. / Marine Pollution Bulletin 56 (2008) 1461–1468 Fig. 3. Short-term leak scenario (£50 pipeline capacity). (a) north site and (b) south site. All perturbations below 0.01 pH unit have been masked for clarity. Fig. 4. Long-term leak evolution. (a) north site and (b) south site, times as stated. All perturbations below 0.01 pH unit have been masked for clarity. 1465 1466 J. Blackford et al. / Marine Pollution Bulletin 56 (2008) 1461–1468 Fig. 5. Evolution of pH perturbation with depth at the release site. (a) north site and (b) south site. is predicted to remain in the marine system for long periods after a leak event; however mixing and diffusion ensure that the remaining CO2 is highly diluted and does not raise the concentration of CO2 significantly above normal levels. Hydrodynamic processes are therefore the key mediating processes. Given that the CO2 driven perturbations reported are primarily controlled by carbonate chemistry (which is well constrained and uncontroversial, e.g. Zeebe and Wolf-Gladrow, 2001) and the physics of the system, the reliability of the results rests on the ability of the POLCOMS model to represent realistic physics for the region and the rate of out-gassing to the atmosphere. The ability of POLCOMS is discussed in Holt et al. (2005) and papers cited therein, and can be justifiably claimed to be fit for purpose with the proviso issues relating to the model resolution discussed above. There is not yet a complete consensus on the parameterization of air–sea exchange rates and this is a rapidly developing research area, (e.g. Borges and Wanninkhof, 2007). The chosen air–sea parameterization for this study is the Nightingale and Liss function derived from a study of CO2 fluxes in the North Sea, i.e. the appropriate gas in the region of interest. Sensitivity analysis of alternative parameterizations have identified little impact on air–sea flux dynamics in this model (not published) which concurs with other studies (e.g. Merico et al., 2006) who also found no significant sensitivity. This arises from the feedback between exchange and concentration in that a low exchange rate increases the differential in partial pressure between air and sea, increasing the net instantaneous exchange. With sufficiently resolved time-steps (200 s in this model) this feedback is tightly coupled. Given however that we are examining short-term perturbations significantly greater than natural variability we performed a sensitivity analysis on the air–sea flux parameterization. We repeated selected simulations using the Wanninkof 1992 parameterization, with the results differing by less than 1% (not shown). The lack of sensitivity to air–sea exchange also indicates that the primary driver for CO2 dispersion is mixing and dilution within the water mass. 3.5. Acidification predicted from oceanic uptake of atmospheric anthropogenic CO2 Acidification rates for the UK shelf region arising from the uptake of atmospheric emissions (Blackford and Gilbert, 2007) depend on CO2 emission scenarios but are generally consistent with rates calculated for the surface ocean system (Caldeira and Wickett 2003). Table 3 details predicted acidification for given atmospheric CO2 concentration and suggests approximate dates based on continuing non-mitigated emissions. The key finding is that acidification of 0.7 pH units below pre-industrial levels will be the global continental shelf and surface ocean result if known fossil fuel reserves are combusted. The implications of such a reduction in ocean pH are hypothesised to be severely damaging to the marine ecosystem and the biogeochemical cycles it mediates. 3.6. Future work investigating ecological feedbacks The ERSEM model computes the uptake and production of DIC as controlled by biological activity, which contributes to the natural variability of DIC (and pH) in space and time. An analysis to test the sensitivity of the results to the background or control (DIC), pH and air–sea exchange was performed by repeating a selection of simulations with the ecological model processes turned off, giving a contrasting annual cycle. The results were identical demonstrating that the results are not sensitive to the background carbonate system as long as the system is spun up to equilibrium. This indicates that the results presented would be broadly applicable to contrasting marine regions. There is (as yet) no feedback between the model’s biological processes and (DIC) or pH. Inclusion of such sensitivities may serve to exacerbate or moderate leak driven perturbations, depending on whether feedbacks are positive or negative. Thus the justification for using the fully coupled POLCOMS-ERSEM system in this context are the future plans to couple the simulated pH perturbation to ecosystem processes such as the effect on nitrification (Blackford and Gilbert, 2007). There is an emerging body of evidence on 1467 J. Blackford et al. / Marine Pollution Bulletin 56 (2008) 1461–1468 flux increase 0.25 0.15 0.1 0.1 0.05 0.05 0 61 flux increase 10 182 243 304 365 0 15 4 10 2 5 7 6 5 4 3 2 1 0 61 122 182 243 304 365 0 122 182 243 304 365 304 365 304 365 304 365 Short-term leak - South 0 61 15 Long-term leak - North 122 182 243 Long-term leak - South 10 5 0 61 122 1.5 182 243 304 365 0 0 61 122 182 243 0.4 No input - North 1 No input - South 0.3 0.2 0.5 0.1 0 0 -0.5 61 20 6 0 0 25 Short-term leak - North 8 0 flux increase 122 Seep - South 0.2 0.15 0 air-sea flux 0.25 Seep - North 0.2 -0.1 0 61 122 182 243 julian days 304 365 -0.2 0 61 122 182 243 julian days Fig. 6. Sea to air CO2 exchange rate increase for each scenario (g C m¡2 d¡1), top three rows. The data represents the increase in out-gassing for each scenario compared with the control run. (a) North site, (b) South site, scenarios as labelled. The bottom row shows the air–sea flux (g C m¡2 d¡1) for each site for the control scenario. Out-gassing is positive. Table 2 The percentage of input CO2 out-gassed over each scenario Table 3 Predicted marine pH under a business as usual emissions scenario (%) of input lost to atmosphere Seep Large short-term leak Long-term leak North South 12 27 20 54 63 60 ecosystem response to short-term elevations of CO2 (e.g. Ohsumi, 2004 and other papers in this special issue; Widdicombe and Needham, 2007) and this will be a key (modelling) research area for the next several years. However many of the emerging ecological responses are ambiguous in that there is high species specificity and little research has been done on recovery. There is also an intriguing trade off between the effects of high CO2 on physiological processes and short-term physiological stress responses which in some instances have opposite effects. In summary encoding ecosystem effects is currently at best speculative. Therefore we have taken the approach to address only the direct chemical perturbation in this study as the first of many necessary steps to evaluating ecosystem response. Hence in the work submitted there is no feed back between the ecosystem model and the pH perturbation, Approx. datea Atmospheric pCO2 Marine pH 1800 260 8.20 1900 1950 2000 2050 2100 2150 2250 285 315 375 500 700 1000 1650 8.17 8.14 8.08 7.97 7.84 7.70 7.50 pH change 0.03 0.06 0.12 0.23 0.36 0.50 0.70 a The dates follow the IPCC IS92a scenario until 2100 then assuming a logistic function for the burning of remaining fossil fuel reserves as used by Caldeira and Wickett (2003). which could potentially modify the biological component of the DIC dynamic for the leakage simulations. 4. Conclusions The assumptions and the choices of leakage scenario used in this work were chosen to significantly over-estimate the amount of CO2 injection into the marine system that might arise from a failure of sequestration, based on current knowledge. Throughout, the 1468 J. Blackford et al. / Marine Pollution Bulletin 56 (2008) 1461–1468 model suggests that any leak driven perturbation approaching the magnitude predicted from non-mitigated atmospheric acidification will be restricted spatially and temporally and perturbations on large temporal (greater than a few days) or spatial (more than a few kilometres) scales will be significantly less than natural variability. This work does not address the local scale (<1 km) impacts of massive point source CO2 injections, and it is likely that on some localised scale, such events would have a catastrophic effect on the environment. Whilst such an event may be damaging politically or socially; it should be seen in the context of the 7.46 £ 105 km2 area of the North Sea and the multitude of other ongoing anthropogenic disturbances. Whilst recovery in the water column would likely be a function of the strong horizontal mixing processes, we lack even initial research on the recovery rates of the relatively immobile benthic communities and this may be a key area for research. This work also does not address the detailed modification of sediment chemistry profiles in response to a wide area seepage event, which may or may not have implications for the health of the sediment, although the relatively small seepage rates observed would suggest less, rather than more, effect. More work remains to be done to look at local and fine scale responses and more scenarios covering a wider range of locations, input levels, timings and durations. For example, the model indicates that the tidal and wind state can significantly modify the in situ perturbation; it would be valuable to consider the potentially cyclical nature of the perturbation when designing exposure experiments. There is also a need to quantify the response of the ecosystem and its components and ultimately address impacts on bio-resources. However, on the basis of this limited, coarse scale, first look we conclude that: firstly, seepage of CO2 from geological formations would not have a significant impact on the overlying ecosystem and secondly, that even massive injections of CO2 from sequestration delivery systems would have minimal effect on the regional scale and an insignificant effect when compared with that expected to result from surface ocean acidification driven by continued uncontrolled CO2 emissions to the atmosphere. Acknowledgements This work was part funded by a NERC/ESRC Grant (UKCCSC NE/C5165X/1), a Grant from DEFRA/DTI (IMCO2, ME2107) and the NERC funded core program of Plymouth Marine Laboratory. The authors would also like to thank Dr. Mark Wilkinson (University of Edinburgh) for guidance with regard to observations of CO2 seepage and an anonymous referee for comments which enabled us to improve on the initial version of the manuscript. References Baretta, J.W., Ebenhöh, W., Ruardij, P., 1995. The European regional seas ecosystem model, a complex marine ecosystem model. Netherlands Journal of Sea Research 33, 233–246. Blackford, J.C., Allen, J.I., Gilbert, F.G., 2004. Ecosystem dynamics at six contrasting sites: a generic modelling study. Journal of Marine Systems 52, 191–215. Blackford, J.C., Gilbert, F.J., 2007. pH variability and CO2 induced acidification in the North Sea. Journal of Marine Systems 64, 229–241. Borges, A.V., Wanninkhof, R., 2007. Fifth International Symposium on Gas Transfer at Water Surfaces. Journal of Marine Systems 66, 1–3. Caldeira, K., Wickett, M.E., 2003. Anthropogenic carbon and ocean pH. Nature 425, 365. DTI 2003. Review of the Feasibility of Carbon Dioxide Capture and Storage in the UK, DTI/Pub URN 03/1261. Holt, J.T., James, I.D., 2001. An s-coordinate model of the North West European continental shelf: part 1. Model description and density structure. Journal of Geophysical Research 106 (C7), 14015–14034. Holt, J.T., James, I.D., Jones, J.E., 2001. An s-coordinate density evolving model of the northwest European continental shelf 2, seasonal currents and tides. Journal of Geophysical Research 106 (C7), 14035–14053. Holt, J.T., Allen, J.I., Proctor, R., Gilbert, F.G., 2005. Error quantification of a high resolution coupled hydrodynamic ecosystem coastal ocean model: part 1. model overview and hydrodynamics. Journal of Marine Systems 57, 167–188. Ingri, N., Kakolowicz, W., Sillén, L.G., Warnqvist, B., 1967. High-speed computer as a supplement to graphical methods: V. HALTAFALL, a general program for calculating the composition of equilibrium mixtures. Talanta 14, 1261. IPCC, 2001. Climate change 2001: the scientific basis. In: Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., van der Linden, P.J., Dai, X., Maskell, K., Johnson, C.A. (Eds.), Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK, p. 881. James, I.D., 1996. Advection schemes for shelf sea models. Journal of Marine Systems 8, 237–254. Klusman, R.W., 2003a. Rate measurements and detection of gas microseepage to the atmosphere from an enhanced oil recovery/sequestration project, Rangely, Colorado, USA. Applied Geochemistry 18, 1825–1838. Klusman, R.W., 2003b. A geochemical perspective and assessment of leakage potential for a mature carbon dioxide-enhanced oil recovery project and as a prototype for carbon dioxide sequestration; Rangely field, Colorado. AAPG Bulletin 87, 1485–1507. Leifer, I., Luyendyk, B.P., Boles, J., Clark, J.F., 2006. Natural marine seepage blowout: contribution to atmospheric methane. Global Biogeochemical Cycles 20, GB3008. Merico, A., Tyrrell, T., Cokacar, T., 2006. Is there any relationship between phytoplankton seasonal dynamics and the carbonate system? Journal of Marine Systems 59, 120–142. Nightingale, P.D., Malin, G., Law, C.S., Watson, A.J., Liss, P.S., Liddicoat, M.I., Boutin, J., Upstill-Goddard, R.C., 2000. In situ evaluation of air–sea gas exchange parameterizations using novel conservative and volatile tracers. Global Biogeochemical Cycles 14, 373–388. Ohsumi, T., 2004. Introduction: what is the ocean sequestration of carbon dioxide? Journal of Oceanography 60, 693–694. Pätsch, J., Lenhart, H.J., 2004. Daily loads of nutrients, total alkalinity, dissolved inorganic carbon and dissolved organic carbon of the European continental rivers for the years 1977–2002. Berichte aus dem Zentrum für Meeres-und Klimaforschung, Reihe B: Ozeanographie. Proctor, R., James, I.D., 1996. A fine-resolution 3D model of the southern North Sea. Journal of Marine Systems 8, 285–295. Raven, J. et al., 2005. Ocean acidification due to increasing atmospheric carbon dioxide. The Royal Society. ISBN: 0 85403 617 2. Song, Y., Haidvogel, D., 1994. A semi-implicit ocean circulation model using a generalized topography-following coordinate system. Journal of Computational Physics 115, 228–244. Stern, N., 2006. Stern Review on the Economics of Climate Change. Cambridge University Press, ISBN 0-521-70080-9. Takahashi, T., Broecker, W.S., Bainbridge, A.E., 1981. Supplement to the alkalinity and total carbon dioxide concentration in the world oceans. In: Bolin, B. (Ed.), Carbon Cycle Modelling, SCOPE 16. John Wiley, Chichester, pp. 159–201. Thomas, H., Bozec, Y., de Baar, H.J.W., Elkalay, K., Frankignoulle, M., Schiettecatte, L.-S., Kattner, G., Borges, A.V., 2005. The carbon budget of the North Sea. Biogeosciences 2, 87–96. Widdicombe, S., Needham, H.R., 2007. Impact of CO2-induced seawater acidification on the burrowing activity of Nereis virens and sediment nutrient flux. Marine Ecology Progress Series 341, 111–122. Zeebe, R.E., Wolf-Gladrow, D.A., 2001. CO2 in seawater: equilibrium, kinetics and isotopes. Elsevier Oceanography Series 65, 346.