Climate-mediated changes to mixed-layer properties in the Southern Ocean: assessing the phytoplankton response

Abstract. Concurrent changes in ocean chemical and physical properties influence phytoplankton dynamics via alterations in carbonate chemistry, nutrient and trace metal inventories and upper ocean light environment. Using a fully coupled, global carbon-climate model (Climate System Model 1.4-carbon), we quantify anthropogenic climate change relative to the background natural interannual variability for the Southern Ocean over the period 2000 and 2100. Model results are interpreted using our understanding of the environmental control of phytoplankton growth rates – leading to two major findings. Firstly, comparison with results from phytoplankton perturbation experiments, in which environmental properties have been altered for key species (e.g., bloom formers), indicates that the predicted rates of change in oceanic properties over the next few decades are too subtle to be represented experimentally at present. Secondly, the rate of secular climate change will not exceed background natural variability, on seasonal to interannual time-scales, for at least several decades – which may not provide the prevailing conditions of change, i.e. constancy, needed for phytoplankton adaptation. Taken together, the relatively subtle environmental changes, due to climate change, may result in adaptation by resident phytoplankton, but not for several decades due to the confounding effects of climate variability. This presents major challenges for the detection and attribution of climate change effects on Southern Ocean phytoplankton. We advocate the development of multi-faceted tests/metrics that will reflect the relative plasticity of different phytoplankton functional groups and/or species to respond to changing ocean conditions.

increased vertical stratification, shallower mixed-layer depths, reduced sea-ice, and higher oceanic CO 2 concentrations. Recent experiments indicate that S. Ocean westerly winds may increase in strength and shift poleward, increasing upwelling (Russell et al., 2006). Alteration of these upper ocean properties affect phytoplankton dynamics and growth rates directly and indirectly through modifications of vertical nutrient supply, 10 mixed-layer depth, and light climate (Bopp et al., 2001).
Modeling studies have so far focused on simulating the decadal to centennial timescale effects of climate change on phytoplankton processes, including the construction of biomes (Sarmiento et al., 2004), or the incorporation of more biological detail, such as algal functional groups (Iglesias-Rodriguez et al., 2002;Le Quéré et al., 2005;15 Litchman et al., 2006) or foodwebs (Legendre and Rivkin, 2005). Recent experimental approaches have concentrated on perturbation studies (weeks to months) on phytoplankton responses to light climate, CO 2 or nutrient concentrations, in laboratory cultures (Riebesell et al., 2000), shipboard experiments (Tortell et al., 2002), mesocosms (DeLille et al., 2005) or 50-200 km 2 in-situ patches of the ocean (Boyd et al., 2007). 20 Two steps are required to explore the relationship between climate-change mediated alteration of ocean properties and the consequent phytoplankton response on intermediary time-scales, i.e. decades -relevant to present experimental and observing system design and to policy makers (Dilling et al., 2003). First, models must provide estimates of future ocean environmental changes, and separate the effects of climate Introduction In this study, we focus on the S. Ocean -a region reported to have a disproportionately large impact on global climate (Sarmiento et al., 1998) and which is particularly sensitive to anthropogenic climate warming (e.g. Sarmiento et al., 2004). These waters are characterized by both high macronutrient concentrations (except for silicic acid in sub-5 polar waters) and low rates of aerosol iron supply (Duce and Tindale, 1991;Fung et al., 2000). Macronutrient concentrations in polar waters are predicted to decrease by only ca. 10% due to climate change (Bopp et al., 2001) and thus the resident phytoplankton are unlikely to be subjected to macronutrient limitation. Phytoplankton in large regions of the S. Ocean are iron-limited, and climate-change 10 driven alterations of ocean physics will likely impact surface iron concentrations. It is not known whether regional dust deposition rates will increase or decrease due to climate change (Mahowald and Luo, 2003;Moore et al., 2006); likewise predictions of climate-mediated changes in UV irradiance are also inconclusive (Denman et al., 2007). Thus, the impact of changes in aerosol-iron supply or UV irradiance are not 15 considered further here. In these waters, the biogeography of the key algal functional groups is already well defined, for example there are no nitrogen fixers in the S. Ocean (Westberry and Siegel, 2006), and a southwards decrease in coccolithophore abundance is observed (Cubillos et al., 2007).
2.1 Climate model formulation 20 Here, we utilize results from numerical climate simulations generated with the National Center for Atmospheric Research (NCAR) Community Climate System Model (CCSM) (Blackmon et al., 2001). The CSM1.4-carbon model is a fully coupled, global climatecarbon cycle simulation. The model formulation of this version and analyses of its preindustrial equilibrium control solutions are detailed in Doney et al. (2006). The physical Introduction EGU ocean and sea-ice model resolution are 3.6 • in longitude and 0.8 • to 1.8 • latitude, with 25 levels in the vertical. The pre-industrial control solutions display stable surface climate and minimal deep ocean drift without requiring surface heat or freshwater flux adjustments. The water cycle is closed through a river runoff scheme. Biogeochemistry in CSM1.4-carbon is simulated with modified versions of the ter- 5 restrial model CASA and the OCMIP-2 oceanic biogeochemistry model (Doney et al., 2004;Najjar et al., 2007). In the fully-coupled carbon-climate model, atmospheric CO 2 is a prognostic variable and is predicted as the residual after carbon exchanges with the land and ocean. The ocean biogeochemical component includes in simplified form the main processes for the solubility carbon pump, organic and inorganic biological carbon 10 pumps, and air-sea CO 2 flux. The prognostic variables transported in the ocean model are phosphate, total dissolved inorganic, dissolved organic phosphorus, dissolved inorganic carbon, alkalinity, and oxygen. New/export production is computed prognostically as a function of light, temperature, phosphate and iron concentrations. The maximum production as a function of 15 temperature is multiplied by nutrient and light limitation terms. The nutrient term is the minimum of Michaelis-Menten limiting terms for PO 4 and Fe: where κ PO4 is 0.05 µmol/L and κ Fe is 0.03 nmol/L. The light (irradiance) limitation term: 20 uses I, the solar short-wave irradiance, and a light limitation term κ I (20 W/m 2 ). A fully dynamic iron cycle also has been added including atmospheric dust deposition/iron dissolution, biological uptake, vertical particle transport, and scavenging. In the simulations, as a simplification, dust deposited onto sea-ice is transferred directly to the ocean, with no modification to its properties, such as solubility (see Edwards and Sed-25 wick, 2001 EGU radiation calculations see a higher effective CO 2 concentration than seen by the land or ocean biogeochemistry. Specifically, radiative CO 2 is computed by calculating the perturbation in atmospheric CO 2 concentration above pre-industrial levels ∆CO 2 , multiplying by factors of 2 and 4, respectively, and then adding back in the preindustrial concentration.

Climate change analysis
Climate warming influences marine ecology and biogeochemistry directly through temperature changes and indirectly through changes in ocean circulation (Boyd and Doney, 2003). The climate change signals in the S. Ocean of the CSM1.4-carbon simulation are approximately (though not exactly) zonal, and therefore we have partitioned the 10 S. Ocean into polar and subpolar waters based on frontal structure, the boundary being set as the simulated 130 Sv streamfunction that approximates well the boundary between the two water masses. Annual time-series plots (model year 2000-2100) for key physical forcing factors and biogeochemical variables from the control and transient climate change simulations (A2, A2-2x, and A2-4x) are shown in Figs Climate trends are significant only if they differ substantially from the drift of the control simulation, if non-zero, and are larger than the intrinsic interannual variability of the 20 model. For the spatial maps, a Student's t-test is used to measure the significance of the differences between the two decades relative to the simulated interannual variability. Regions where ∆χ (2020-2029 minus 2000-2009) Table 1b, please refer to these publications.

Model simulations
The CSM1.4-carbon results are broadly similar to other COAM simulations that exhibit significant surface warming of S. Ocean waters in response to anthropogenic climate  Table 1a). The A2 subpolar anthropogenic SST signal is significantly greater than the natural interannual variability for the region, even on this time-scale. The warming is less pronounced in polar waters 15 closer to Antarctica in the A2 case; the temperature change does not exceed natural variability in most locations, and there is also a large area of cooling in the Atlantic sector. The anthropogenic temperature changes are even larger in the A2-x2 and A2-x4 climate sensitivity cases, and for the A2-x4 case the polar temperature rise becomes significant relative to natural variability. 20 Surface temperature is one of the few COAM variables for which we can compare the simulated surface ocean trends against the observed historical Southern Ocean record over the last several decades. This can be done due to the relatively large climate warming signal compared to interannual variability, and because of the relatively comprehensive in situ and satellite data coverage for temperature. But even for tem-Introduction EGU perature, the historical observation data set (particularly for subsurface waters) has significant gaps in the Southern Ocean, complicating model evaluation . Further, because the coupled model has its own internally-driven, non-linear climate variability that is often out of phase with the interannual variability of the real system, we are restricted to comparing long-term secular trends in the model and ob-5 servations.
With those caveats in mind, the long-term spatial warming patterns in the model are broadly consistent with historical observations (Fig. 3). Like the model, observations exhibit higher surface warming in the Southern Ocean subtropical and subpolar bands than in polar waters, where there has either been no statistical trend or in some cases 10 weak cooling rather than warming (Smith and Reynolds, 2005;Trenberth et al., 2007). Averaged zonally, the historical  upper ocean warming trends in the Southern Ocean are 0.0 to +0.025 K/decade for polar waters and +0.025 to +0.10 K/decade for subpolar waters (Levitus et al., 2005;Bindoff et al., 2007). Using subsurface float data from the 1990s relative to historical hydrography, Gille (2002) reports mid-depth 15 (700-1000 m) warming rates of 0.04+/−0.01 K/decade averaged across 35 S to 65 S and rates as high as 0.08+/−0.02 K/decade in the core of the subantarctic front. The highest rates of subsurface ocean warming are comparable to the observational atmospheric surface warming rates for the Southern Ocean region of 0.13+/−.06 K/decade. The CSM coupled model warming rates for the last half of 20th century (in model 20 years 2000-2009 minus 1950-1959 adjusting for the atmospheric CO 2 delay) are in good agreement with the observed warming. As predicted by other coupled simulated, the model predicted warming rates tend to accelerate in the 21st century, with the simulated warming rates for the next 20 years of subpolar +0.10 to +0.31 K/decade; polar +0.03 to +0.17) (Figs. 1-3, Table 1). 25 The other clear anthropogenic signal is, as would be expected, the increase in surface water pCO 2 due to the rise in atmospheric CO 2 in the transient simulation. North of ∼55 • S, the ∆pCO 2 (here transient -control, not air -water) tracks the increase in Introduction EGU the surface water pCO 2 increases but not as rapidly as the atmosphere due to ice cover and the upwelling of older waters from below; average levels are only 720-760 ppmv by 2100. For most of the other physical and biogeochemical factors in the CSM1.4 A2 case, however, the climate change signal is smaller than or comparable to natural variability 5 on the decadal time-scale (2020-2029 minus 2000-2009), and there are spatial regions (sub-basin and basin scale) of both positive and negative change over twenty years. The anthropogenic climate change signals become more distinct in the A2-x2 and A2-x4 cases, with increased climate sensitivity, for the following properties: decreasing polar sea surface salinity, increased subpolar and polar stratification, de-10 creased subpolar and polar surface dissolved iron concentrations, poleward shift and increased strength in polar upwelling, and increased polar surface ocean irradiance due to decreased ice fraction.
The climate response for surface phosphate and iron concentrations reflect the balance between changing upwelling and mixing of high nutrient, subsurface waters from 15 below (source), residence time of surface water, and biological export production (sink). The rate of change of surface nutrient concentrations χ is given by: where the physical transport is partitioned into horizontal and vertical advection and mixing terms; the biological surface uptake and subsurface remineralization are 20 grouped into a single biogeochemical right hand side term RHS bio . Increased stratification and shallower mixed layers tend to decrease the mixing K Z , while shifts in wind patterns may tend to increase upwelling u Z while decreasing surface resident time-scales. In CSM1.4, surface nutrient concentrations tend to stay constant or decrease. Phos- 25 phate exhibits a small decline from 40-60 • S on average and is constant from 60 • S to the pole. Phosphate concentrations remain large relative to κ PO4 and do not become limiting to organic matter production and export (Eq. 1). There are more notable 4293 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version Interactive Discussion EGU decreases in surface dissolved iron concentrations, −5 to −10 pmol/l relative to mean levels of 80-160 pmol/l for 40-60 • S and −20 to −30 pmol/l relative to mean levels of 150-240 pmol/l south of 60 • S. For comparison, the model half saturation constant for algal iron limitation for biological organic matter export κ Fe is 30 pmol/l. The trends in surface macro-and micro-nutrients, along with the light/mixed layer and SST in turn 5 impact the downward export flux (Eqs. 1 and 2). Integrated over the entire S. Ocean (including the subtropics), simulated organic matter export remains about the same, though it tends to shift poleward due to the southward migration in the band of deep winter mixing and the expansion of subtropical conditions. In the subpolar region, downward export increases by ∼10-15% in October and July while it tends to decrease 10 further south. The trends predicted for climate change of warmer waters, shallower mixed layers, and thus higher light levels, and lower nutrient and dissolved iron concentrations in the mixed layer (Table 1a) are consistent with the change in seasonal properties from spring to summer conditions ( Table 2). The one exception to this is oceanic CO 2 15 concentrations, which are predicted to rise (Table 1a), akin to the progression from summer to winter conditions (Table 2). Together, the combination of the spring to summer progression (light, nutrients, metals) and summer to winter progression (CO 2 ) is something that phytoplankton will not have encountered previously. The seasonal amplitude in key oceanic properties in both polar and sub-polar waters is considerably 20 greater than the magnitude of the predicted changes in these properties due to climate change (Table 1a c.f. Table 2), however there is insufficient resolution in the simulations to investigate how this seasonal amplitude will be influenced by climate change.
Broadly similar physical climate response patterns have been found in other 21st century coupled ocean-atmosphere climate simulations, though as mentioned above 25 the CSM1.4-carbon model tends to be at the low end of the range in terms of climate sensitivity to atmospheric CO 2 perturbations. Using a series of empirical diagnostic calculations, Sarmiento et al. (2004) examined the potential marine ecological responses to climate warming using physical data from six different coupled climate model simu- EGU lations, including a variant of CSM 1. Averaged over the models, climate warming led to a contraction of the highly productive marginal sea ice biome and expansion of the subpolar gyre biome and low productivity, permanently stratified subtropical gyre biome in the S. Ocean. Vertical stratification tends to increase, which would be expected to decrease nutrient supply everywhere but also increase the growth season in some 5 high latitude regions (Sarmiento et al., 2004). They did not investigate the magnitude of upwelling other than in defining the boundaries and shifts in biomes. Sarmiento et al. (2004) suggest that chlorophyll will increase in the open S. Ocean, due primarily to the retreat of and changes at the northern boundary of the marginal sea ice zone; but chlorophyll may tend to decrease adjacent to the Antarctic continent due primarily to 10 freshening within the marginal sea ice zone. Estimated primary productivity generally will increase mainly as a result of warmer temperatures. The climate-change driven trends in surface water properties in the S. Ocean do not occur independently, and synergistic effects need to be accounted for in phytoplankton responses. Table 3 presents property-property correlations for temperature, mixed 15 layer depth and nutrient anomalies for both interannual variability (annual means from control simulation) and anthropogenic climate change (A2 case, decadal means of 2020-2029 minus 2000-2009; Fig. 3). For natural interannual variability, warmer SSTs are correlated with shallower mixed layers and lower surface nutrient concentrations; surface nutrients increase with mixed layer depth, and surface iron and phosphate 20 are well correlated. The SST-surface nutrient correlations are of the same direction in the anthropogenic climate change case, although the magnitudes tend to differ from interannual variability. The correlations with mixed layer anomalies are weak and not statistically significant. 25 Physiological responses of both laboratory-cultured species and natural assemblages to altered environmental conditions are summarized (Table 1b). The range of conditions used in experiments (Table 1b)  EGU these properties due to climate change (Table 1a). Tilzer et al. (1986) investigated temperature effects on photosynthetic performance in cells from the Scotia Sea and Bransfield Strait. Light-saturated photosynthesis (P b max) increased significantly with increasing temperatures from −1.5 • C to 8 • C, c.f. Table 1a). The greatest enhancement of P b max occurred between −1.5 • C and 2 • C (Q 10 >4), but decreased at temperatures 5 of >2 to 5 • C, (Q 10 of 2.6), and at >5 • C was negligible. This response illustrates the high sensitivity of polar organisms to temperature (Clarke, 2003). Studies on coastal phytoplankton indicate that a 1 • C-2 • C warming around the Antarctic Peninsula might not alter photosynthetic rates, but could channel more photosynthate into DOC (Moran et al., 2006). Also in these waters, Moline et al. (2004) reported shifts from diatoms 10 to chrysophytes, between 1991 and 1996, associated with warmer temperatures and a consequent reduction in salinity. Thus, warming can both directly and indirectly impact phytoplankton processes, and probably influences polar (ice melt, Q 10 ) more than sub-polar waters. Natural gradients in surface nutrient concentrations (<5 to 45 µmol Si L −1 ) such as 15 during a southwards progressing bloom along the 170 • W meridian (i.e. Pacific sector; Nelson et al., 2001) provides estimates of how the silicic acid uptake kinetics of diatoms are altered by changing upper ocean conditions. Such a lateral silicic acid gradient is much greater than the relative change in nutrient concentrations due to climate change (see PO 4 , Table 1a). Nelson et al. (2001) observed no clear trend in affinity for silicic 20 acid by diatoms (i.e. K s ) with latitude or silicic acid concentration, however there was an apparent seasonal progression -with a lower affinity for silicic acid evident over time. Critically, information in Nelson et al. (2001) on which species were present is lacking, and thus it is not known whether temporal/spatial shifts in floristics influenced these trends in uptake kinetics. Thus, their findings have limited value in predicting 25 how diatoms, at the species level, might respond to predicted reductions in nutrient concentrations ( EGU to three-fold range of irradiances on a timescale of hours to days (R. Strzepek, unpublished data) Such photoacclimation has been widely reported for non-polar species (Falkowski and LaRoche, 1991). The maximum growth rate of Phaeocystis antarctica occurs at ∼60 µmol quanta m −2 s −1 (Table 1b), and declines rapidly at lower irradiances, whereas at higher than saturating irradiances decreases in growth rate is much 5 slower (R. Strzepek, unpublished data). This is also observed for cultured polar diatoms (R. Strzepek, unpublished data). Predictions in Table 1a indicate that increased mean irradiances, due to a shoaling of the mixed layer, will occur with climate change. However, this increase is probably too subtle to result in a detectable alteration of growth rates. 10 Beardall and Giordana (2002), reviewed phytoplankton CCM's (Carbon Concentrating Mechanisms), and suggest that rising ocean CO 2 concentrations may impact the performance of groups with (diatoms) and without (chrysophytes) CCM's. Rost et al. (2003) examined representative bloom-formers (diatom, coccolithophorid, and Phaeocystis) acclimated to 36-1800 ppmv CO 2 . They reported major differences in 15 each species ability to both regulate carbon acquisition and in the efficiency of acquisition. In the most comprehensive study of CCM's in S. Ocean biota to date, Tortell et al. (2007) 1 present evidence of the presence of inorganic CCM's (i.e. cells transporting HCO − 3 and utilizing carbonic anhydrase to catalyze HCO 3 dehydration to CO 2 ) for each of ten polar species, including P. antarctica, and Fragilariopsis kerguelensis, they 20 investigated. This widespread presence of CCM's is evident for mixed assemblages in the subpolar Bering Sea (Martin et al., 2006) andNE Pacific (Tortell et al., 2006). Thus, it is difficult to predict whether any particular species would have a selective advantage to utilize CO 2 at the predicted higher concentrations (Table 1a).

Polar phytoplankton responses to environmental perturbations
In contrast to light climate and CO 2 concentrations, changes in dissolved iron con- 25 centrations had a pronounced effect on the growth rate (i.e. K m , see EGU dissolved iron or bioavailable/free Fe (Table 1b). This range of dissolved iron concentrations (that set half the maximum algal growth rate) is considerably greater than predicted changes in dissolved iron concentrations due to climate change (Table 1a). Polar diatoms generally exhibited higher values of K m than for P. antarctica (Table 1b), indicative of a greater sensitivity to future decreases in dissolved iron predicted by the 5 model (Table 1a).
There are few studies of the synergistic effects due to simultaneous limitation (or its alleviation) of algal growth by multiple environmental factors. Takeda (1998) reported increased silicification rates at low, relative to high, dissolved iron concentrations in two cultured polar diatoms: silicification will be altered by both iron and silicic acid 10 supply, two properties that will be influenced by climate change (Table 1a). Beardall and Giordano (2002) concluded that factors which may impact algal CCM's include CO 2 concentrations, temperature and also UV-B radiation. In SW Atlantic polar waters, Feller et al. (2001) reported that both SST and low dissolved iron concentrations limited growth rates of resident phytoplankton. These studies all point to the complex inter- 15 actions between environmental properties that will be altered concurrently by climate change (Table 1a).
Other synergistic effects include iron and irradiance, and studies of iron/light interactions have reported up to threefold reductions in K m for colonial P. antarctica following transfer from low to high light conditions (Sedwick et al., 2007; (Table 1b) observed that, for three cultured Thalassiosira species and P. antarctica (non-colonial), the greatest effect of iron-limitation is observed at higher growth irradiances, i.e. µ/µmax is lower at higher light levels (where µ denotes Fe-limited and µmax represents Fe-replete growth rates). His findings differ from reports that low irradiance exacerbate algal iron requirements (Timmermans et al., 2001;Maldonado et al., 1999); 25 this disparity may result from the use of different cultured species, or different methodologies (acclimated versus non-acclimated, i.e. Timmermans et al., 2001, cultures). Raven et al. (1999) suggest that low light results in both increases and decreases in algal iron requirements depending on species.  Table 1a summarises the concurrent changes in factors that influence phytoplankton processes. The predicted decadal rates of change for all properties, including the 5 largest climate change signal -CO 2 (3-4 ppmv yr −1 ), are unprecedented on geological timescales (Denman et al., 2007); however they are small compared with experimental perturbations used to date (Table 1b). The studies, listed in Table 1b, provide valuable insights into what environmental factors control phytoplankton physiology. However, they will be of limited value for predicting the algal response to climate 10 change, as they do not represent the potential for physiological plasticity and hence adaptation to the more gradual trends in environmental properties that will occur due to climate change. Such representation, for example, would require slow increases in CO 2 of 3-4 ppmv yr −1 for six months to several years. Such an experiment, albeit with a small change in CO 2 , might permit extrapolation of the observed response of differ-15 ent species to perturbation, and would refine the approaches used previously (i.e. large instantaneous perturbations, Riebesell et al., 2000). But artifacts due to bottle or mesocosm containment will likely confound any biological response to the climate change signal. Thus, we must explore alternative means to capture, in some way, the effects of relatively slow rates of climate-mediated change on ocean properties. 20 Although the predicted rate of change in each ocean property is too subtle to be represented experimentally, it is possible that the potential multiplicative impact of synergistic effects (i.e. simultaneous limitation of algal growth by several factors) could be reproduced. Here, we consider iron and light. The model predicts lower dissolved iron concentrations and slightly higher underwater light levels (Table 1a), which may 25 result, for some algal species (Raven et al., 1999), in greater iron and light co-limitation (i.e. the greatest displacement of growth rates from their maximum). It is also possible that cells may be growing faster (relative to the present day) due to the predicted 4299 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version Interactive Discussion EGU higher irradiances. However, any amplification of the biological signal due to a greater phytoplankton response to multiple factors (and their alteration by climate-change) may be countered by increased uncertainty, due to increased complexity, in predicting how phytoplankton dynamics will alter when multiple properties are considered. Therefore, the combined effect of simultaneous alteration of these factors is currently beyond pre-5 diction.

Climate change versus variability -implications for biological adaptation
In addition to climate change, other sources of variability will occur concurrently, including climate variability (e.g. SOI), seasonal gradients, and episodic (weeks) perturbations (e.g. dust storms). Although the effects of climate variability and climate 10 change can be deconvolved in simulations (Fig. 3), these overlying effects will potentially confound the detection and attribution of trends in the biota as they respond to climate change. For the coming decades, the magnitude of change in oceanic properties due to climate variability is comparable to that by climate change, and moreover the direction of the change may differ between climate change and variability (Table   15 1a; see Sarmiento, 1993). Thus, it is possible that secular climate change will only induce significant biological effects when the magnitude of the environmental perturbations (and floristic changes) exceed background natural variability on seasonal to interannual time-scales. Such an inflection point might lead to detectable responses by the biota, as species commence selection for eco-types (Medlin, 1994) more suited 20 to changing conditions where the sign of change is constant over time (i.e. adaptation rather than acclimation, Falkowski and LaRoche et al., 1991); as opposed to changing conditions with no prevailing trend due to fluctuations in the dominant control on the alteration of ocean properties (seasonal gradients, climate variability, climate change increase). 25 The assessment of the relative ability of different phytoplankton species to resist change and/or adapt to climate change may be a valuable tool to be used in conjunction with other experimental approaches such as Collins and Bell (2002 EGU toplankton, i.e. the maintenance of a given state when subject to disturbance (sensu, Carpenter and Cottingham, 1997), may result from an in-built tolerance of a wide range of environmental conditions (Margalef, 1978). Physiological plasticity, defined here as the ability to acclimate (i.e. physiological processes, Falkowski and LaRoche, 1991) to, and therefore gradually adapt (i.e. evolutionary processes, Falkowski and LaRoche, 5 1991) to, changing and/or new conditions, is akin to resilience. As stated earlier, such adaptation -over longer timescales -presupposes the need for a clear and sustained change in conditions. It is now established that oceanic organisms can adapt physiologically, over timescales of years, to pronounced environmental changes. Such adaptability has been observed in corals, in response to warming temperature causing bleaching, which successfully responded by recruiting "new" algal symbionts (Baker et al., 2004). Furthermore, a thousand generations of Clamydomonas at elevated CO 2 concentrations resulted in smaller cell size and broader ranges in rates of photosynthesis and respiration (Collins and Bell, 2004). Thus, the relationship between the physiological plasticity of phytoplankton and the rate of change of environmental 15 drivers is a key factor (and unknown) in determining whether such altered environmental forcing will result in floristic shifts and/or altered physiology. There appears to be several key factors that control the degree of plasticity: a) the underlying survival strategy of organisms -specialist (k) or generalist (r) (Margalef, 1978); b) genetic variability and phytoplankton species and c) the timing of the emer-20 gence of different groups and/or species in the geological past. Generalist species tend to dominate in systems close to resource limitation such as oligotrophic waters (Reynolds, 1984). It is likely that the low iron supply, that characterises much of the S. Ocean, would favour r over k species, and the mandala of Margalef (1978) should be modified to include trace metals. Physiological plasticity will be driven by both geno- 25 and pheno-typic characteristics, for example the success of bloom-forming phytoplankton may be due to their genetic variability, such that within-bloom genetic variability may be greater than that between blooms (Medlin et al., 1996). However little is known about whether genetic diversity between different species will result in a correspond- EGU ing degree of physiological or ecological "variability" (Iglesias-Rodriguez et al., 2006). The 2 year (i.e. 1000 generations) study of Clamydomonas by Collins and Bell (2004) presented evidence of accumulated mutations in genes affecting the CCM, after at elevated CO 2 concentrations, that were translated into physiological changes.
It is now evident that the physiological characteristics of different phytoplankton 5 groups was strongly influenced by the ambient conditions when they emerged over time since the Proterozoic era. The trace metal requirements of different algal groups (high iron for diatoms) may have been set by the redox state of the ocean at the time of their emergence (Saito et al., 2003;Quigg et al., 2003;Falkowski et al., 2004). This may also be the case for coccolithophorids which evolved under different CO 2 condi-10 tions (Langer et al., 2006a). Experiments by Langer et al. (2006a, b) have shown that two species have different calcification responses across a range of CO 2 concentrations; Coccolithus pelagicus (no change with increasing CO 2 concentrations) and Calcidiscus leptoporus (maximum rate at present day CO 2 concentrations). Significantly, these responses differ from those reported by Riebesell et al. (2000) for Emiliania hux- 15 leyi. Thus, the ability to respond to changing trace metal or CO 2 conditions may have been imprinted during their evolutionary history.

Scenarios for phytoplankton responses -implications for detection versus attribution
Consideration of the possible effects of such physiological plasticity leads to three sce-20 narios: 1) ecosystems are very "plastic" (Langer et al., 2006b)-with no or limited changes in community structure as the resident cells can adapt to climate change over years to decades; 2) the climate change signal simply results in a poleward migration of "fixed" biomes (Sarmiento et al., 2004); 3) conditions change sufficiently that a "new" community or ecosystem arises that has no analogue in current ocean (Boyd 25 and Doney, 2003). These scenarios provide conceptual frameworks to examine the detection and attribution of such shifts. Cells within "plastic" ecosystems could adjust their physiological properties rather 4302 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version Interactive Discussion EGU than alter species composition. Such a response is particularly difficult to detect and monitor, and would require a time-series of physiological experiments on natural populations. This approach has been advocated in terrestrial ecosystems, via the integration of observations on natural climate gradients with climate change experiments (Dunne et al., 2004). Climate-mediated shifts in biomes will probably be easier to detect and monitor, for example remote-sensing to monitor shifts in coccolithophore distributions was used successfully in the Bering Sea (Merico et al., 2003). Migration of the boundaries of biomes is less likely in the S. Ocean where the geographical isolation and strong meridional frontal boundaries (Smetacek and Nichol, 2005) minimize the impact of phenotypic adaptation (Medlin, 1994) and results in well-defined biomes. Thus, 10 unanticipated floristic shifts will be conspicuous, and thus should be readily monitored.
In contrast, unanticipated shifts to a new assemblage, may not be readily detected, unless they have a different bio-optical signature that results in fortuitous detection by satellite sensors. 15 Our model predictions point to subtle changes in environmental properties that influence phytoplankton processes over the coming decades in the S. Ocean. How will such changes influence the response of the resident phytoplankton? Based on our limited knowledge, such as from long term time-series observations (Smayda et al., 2005), or case studies of the effects of climate variability (Chavez et al., 1999), adapta-20 tion is the most likely algal response to climate change in these waters (N.B. this does not take into account top-down foodweb effects). The extension of biomes is unlikely due to the strong circumpolar features in the S. Ocean and the magnitude of change in oceanic properties is probably insufficient in the next two decades to result in pronounced floristic shifts. Such dramatic shifts, for example from low picophytoplankton 25 stocks to diatom blooms, are mediated by climate variability in the Equatorial Pacific (Chavez et al., 1999), but are driven by large and rapid changes in environmental conditions (i.e. a sixfold increase in vertical iron supply between years, c.f. EGU such they are generally poor proxies for climate change. However, such adaptation to climate change, and its detection, in the next twenty years will be confounded by other sources of climate variability.

Approaches to investigate phytoplankton responses to climate change
No single approach is sufficient to address this pressing issue. Due to the difficulties in 5 conducting manipulation experiments which represent the predicted rates of change in oceanic properties, multi-faceted tests for assessment of change should be developed (e.g. Peterson and Keister (2003). A nested suite of perturbation experiments including laboratory (mechanistic understanding of physiological pathways; see MacIntyre and Cullen, 2005) and shipboard experiments (physiology of assemblages), mesocosms or mesoscale perturbations (floristic shifts and their underlying mechanisms), and novel ecosystem modeling techniques (Follows et al., 2007) are required. Climate-change perturbation studies in terrestrial systems reveal the major influence of experimental duration (up to 5 years) on the outcome of perturbation (Walther, 2007), but extrapolating such conclusions to the ocean is problematic due to differences in the turnover of 15 plant biomass between land (years) and ocean (days) (Falkowski et al., 1998). Boyd and Doney (2002) advocated the use of perturbation experiments, monitoring and biogeography to investigate the effects of climate change and the subsequent feedbacks. Their approach required both global (incorporation of greater biological complexity) and regional (data interpretation based on a scheme of provinces). Our study 20 provides further insights into what degree of biological complexity is required. The use of evolutionary history in conjunction with assessment of the paleo-environment under which algal groups emerged is a powerful tool to interpret the results from perturbation experiments (Langer et al., 2006a, b). Thus, the three-stranded approach of Boyd and Doney (2002) (1994). Phytoplankton were grown in the artificial seawater medium Aquil (Price et al., 1988) prepared in Milli-Q water (Millipore Corp.). The seawater, containing the major salts, was enriched with 10 µmol L −1 phosphate, 100 µmol L −1 silicate and 300 µmol L −1 ni- 15 trate. Trace metal contaminants were removed from the medium and nutrient enrichment stock solutions using Chelex 100 ion exchange resin (Sigma, St. Louis, MO) according to the procedure of (Price et al., 1988). Media were enriched with filtersterilized (0.2 µm Gelman Acrodisc PF) EDTA-trace metal and ESAW vitamin solutions (Harrison et al., 1980). Free trace metal ion concentrations, in the presence of 20 10 µmol L −1 EDTA as the chelating agent, were as follows (-log free metal ion concentration = pMetal): pCu 13.79, pMn 8.27, pZn 10.88, and pCo 10.88. These concentrations were calculated using the chemical equilibrium computer program MINEQL (Westall et al., 1976) with the thermodynamic constants reported in Ringbom (1963) . Selenite and molybdate were added at 10 −8 and 10 −7 mol L −1 respectively. The salin- 25 ity and initial pH of the medium was 35 psu and 8.17±0.04 (n=13), respectively. All cultureware and plastics that came in contact with cultures were rigorously cleaned and sterilized according to the procedures detailed in Maldonado and Price (1996)  as determined with the sulfoxine method. Cultures were also grown in Aquil medium 15 containing either 10 or 100 µmol L −1 of EDTA containing and no added iron. To induce Fe stress, 2 nmol L −1 of Fe was added as a complex with 4, 40 or 400 nmol L −1 of the terrestrial siderophore desferrioxamine B mesylate (DFB, Sigma) to Aquil medium containing 10 µmol L −1 of EDTA to chelate the other trace metals. Stock DFB solutions (0.1-10 mmol L −1 ) were prepared in Milli-Q water, and were 0.2 µm filter sterilized af-20 ter dissolution. The Fe was premixed with the DFB before addition to the media as previously described (Maldonado and Price, 1999). The media with the added FeDFB complex were allowed to equilibrate overnight. As P. antarctica maintained maximum growth rates in Aquil medium containing no added Fe and either 10 or 100 µmol L −1 of EDTA, we assumed that the 1.8 nmol L −1 Fe contamination in the basal medium 25 was bioavailable. Therefore the FeDFB ratios in these media were 3.8:4, 3.8:10, and 3.8:100 nmol L −1 , respectively. Specific growth rates (d  EGU Moline, M. A., Claustre, H., Frazer, T. K., Schofield, O., and Vernet, M.: Alteration of the food web along the Antarctic Peninsula in response to a regional warming trend, Global Change Biol., 10, 1973-1980, doi:10.1111/j.1365-2486.2004 Table 1. Comparison for S. Ocean sub-polar and polar waters of the relative rate of a) climatechange mediated alteration of annual-mean ocean properties that influence phytoplankton processes expressed as change per decade compared with; b) A summary of the range of conditions for each ocean property under which manipulation experiments have been conducted on either key polar bloom-forming species, or natural polar assemblages. The climate change rates in a) are estimated from a suite of fully coupled CSM 1.4-Carbon simulations by spatially averaging over the subpolar or polar domains (boundary defined by the 130 Sv stream function contour) and averaging over the period 2000 and 2100. The model ranges correspond with the simulations with varying model climate sensitivities to atmospheric CO 2 perturbations. The model estimated rms variability in annual-mean, spatial averaged ocean properties due to natural climate variability is included for reference. nd denotes uncertainty over sign of change (Mahowald et al., 2005). Due to the paucity of data on sub-polar phytoplankton this summary focuses only on polar waters.