Sentinel systems on the razor���s edge: effects of warming on Arctic geothermal stream ecosystems G U Y W O O D WA R D *, J O H N B . D Y B K J AE R w , J O �� N S . O �� L A F S S O N z, G I �� S L I M . G I �� S L A S O N �� , E L I �� S A B E T R . H A N N E S D O �� T T I R �� and N I K O L A I F R I B E R G w } *School of Biological & Chemical Sciences, Queen Mary University of London, London E1 4NS, UK, wDepartment of Freshwater Ecology, National Environmental Research Institute, University of Aarhus, Vejls vej 25, DK-8600 Silkeborg, Denmark, zInstitute of Freshwater Fisheries, Keldnaholt, IS-112 Reykjavik, Iceland, ��Institute of Biology, University of Iceland, Sturlugata 7, IS-101 Reykjavik, Iceland, }Macaulay Land Use Research Institute, Catchment Management Group, Craigiebuckler, Aberdeen AB15 8QH, UK Abstract The Earth is experiencing historically unprecedented rates of warming, with surface tem- peratures projected to increase by 3���5 1C globally, and up to 7.5 1C in high latitudes, within the next century. Knowledge of how this will affect biological systems is still largely restricted to the lower levels of organization (e.g. species range shifts), rather than at the community, food web or ecosystem level, where responses cannot be predicted from studying single species in isolation. Further, many correlational studies are confounded with time and/or space, whereas experiments have been mostly confined to laboratory microcosms that cannot capture the true complexity of natural ecosystems. We used a ���natural experiment��� in an attempt to circumvent these shortcomings, by characterizing community structure and trophic interactions in 15 geothermal Icelandic streams ranging in temperature from 5 1C to 45 1C. Even modest temperature increases had dramatic effects across multiple levels of organization, from changes in the mean body size of the top predators, to unimodal responses of species populations, turnover in community composition, and lengthening of food chains. Our results reveal that the rates of warming predicted for the next century have serious implications for the structure and functioning of these fragile ���sentinel��� ecosystems across multiple levels of organization. Keywords: food webs, freshwater ecology, geothermal streams, Iceland, stable isotope analysis, trophic interactions Received 15 March 2009 and accepted 1 August 2009 Introduction Global surface temperatures have risen by 0.74 1C in the last century and are projected to increase by a further 3���5 1C over the next century, a rate of change that is unprecedented in human history (Ponds et al., 1999 Parmesan & Yohe, 2003 Thomas et al., 2004). The impacts on the planet���s diverse ecosystems will be profound, but most research to date has focused on the impacts at the lower levels of biological organiza- tion (e.g. range shifts in species populations) (Sala et al., 2000 Hickling et al., 2006 Parmesan, 2006 IPCC, 2007 but see McKee et al., 2002a, b and Durance & Ormerod, 2007, 2009). When communities or ecosystems are sub- jected to environmental stress they can behave in ways that cannot be predicted from studying single species in isolation, due to the complex array of species interac- tions within the ecological network (Raffaelli, 2004 Woodward, 2009). Consequently, a new perspective that considers the higher levels of organization is needed to complement existing approaches to predicting climate change impacts in these complex natural systems (Ings et al., 2009). Unfortunately, many studies on the effects of warm- ing on natural systems are hampered further still by confounding latitudinal (or altitudinal) gradients (e.g. Jacobsen et al., 1997), often making it impossible to disentangle the effects of biogeography and tempera- ture. Changes in community composition with latitude occur for many reasons other than differences in tempe- rature: the interplay between additional environmental Correspondence: Nikolai Friberg, Department of Freshwater Ecology, National Environmental Research Institute, University of Aarhus, Vejls vej 25, DK-8600 Silkeborg, Denmark, tel. 1 45 89 20 14 90, fax 1 45 89 20 14 14, e-mail: nfr@dmu.dk Global Change Biology (2010) 16, 1979���1991, doi: 10.1111/j.1365-2486.2009.02052.x r 2009 Blackwell Publishing Ltd 1979

gradients (e.g. pH, nutrient concentrations, etc,), hy- drology, geology, evolutionary history and dispersal constraints will all influence the composition of the local and regional species pools. Similar limitations can also apply when temporal correlations are used to infer biological responses to temperature change (e.g. Ponds et al., 1999 Moatar & Gaillard, 2006 Durance & Ormerod, 2007, 2009). In contrast to large-scale latitu- dinal surveys, most experimental studies of warming have been restricted to small laboratory microcosms (e.g. Petchey et al., 1999) or pond mesocosms (e.g. McKee et al, 2002a, b, 2003 Moss et al., 2003), largely due to the logistic and financial challenges of heating large systems for a protracted period. Although offering the advantage of being able to replicate environmental conditions in a tightly controlled manner (Benton et al., 2007), this approach inevitably suffers from a lack of realism and the validity of extrapolating from simple protist and microbial assemblages to more complex natural communities has been questioned (Ings et al., 2009). Clearly, although neither extreme is ideal, both are useful for painting at least a partial picture and, between these two extremes lie field manipulations (e.g. Hogg & Williams, 1996 McKee et al, 2002a, b, 2003 Moss et al., 2003) and natural experiments. The latter can be found in a few parts of the world and although not as tightly controlled or as highly replicable as mesocosm experiments they compensate for this by being able to capture more of the true complexity of real systems at larger spatial scales (Woodward, 2009). The study of geothermally heated ecosystems pro- vides a promising avenue of research, as these often span a wide thermal gradient within a relatively small area, and are therefore not confounded with biogeogra- phical and dispersal constraints (c.f. Vannote & Sweeny, 1980). Fresh waters are particularly useful in this regard because they are relatively discrete entities with clearer boundaries than those of their terrestrial or marine counterparts and these ���islands in a dry sea��� are also particularly sensitive to perturbations (Woodward & Hildrew, 2002). The challenge, however, lies in identify- ing suitable study sites that avoid the potentially con- founding effects of water chemistry differences associated with geothermal activity rather than simply temperature per se. High-latitude ecosystems possess the added advantage of being ���sentinel systems��� in ecologically sensitive regions where rates of warming will be among the fastest on the planet, with rises of up to 7.5 1C predicted within the next century (IPCC, 2007). We are fortunate to have identified what we believe to be an ideal study system that meets the above criteria: our characterization of 15 streams within a geother- mally heated area in Iceland, on the edge of the Arctic Circle, represents one of the first detailed community- level studies of the effects of warming across a natural thermal gradient. These sites are thus not only well- suited to studying the effects of temperature at a single point in time, but they also provide important reference baseline conditions that can be used to track responses to future warming. We used this system to address a range of hypotheses related to the impact of warming on natural ecosystems. A recent study of ecosystem functioning in a subset of 10 of these streams has revealed strong increases in primary production and decomposition at elevated temperatures (Friberg et al., 2009): in the current study we investigate how multiple levels of organization, from individuals to species po- pulations and ultimately the community food web, respond to the thermal gradient across all 15 streams in the catchment. We can predict, for instance, that if an individual���s thermal performance reflects evolutionary optimization to local thermal regimes (Huey & Stevenson, 1979 Clarke, 1991a, b, 1993 Karl et al., 2008), increasing metabolic costs will impair the ability of species popu- lations to operate effectively as temperatures change, such that cold-adapted populations at high latitudes should be vulnerable to warming (Addo-Bediako et al., 2000, 2002 Thomas et al., 2004 Deutsch et al., 2008). Given the relatively biogeographically isolated position of Iceland, we might therefore expect species richness to decline with rising stream temperatures (e.g. Lamberti & Resh, 1985) especially because of the lack of a pool of warm-adapted species in an otherwise cold and harsh environment: this would mimic the effects of climate- driven warming at the whole-island scale, due to a lack of suitably adapted colonists from other land masses. Further, we could reasonably expect to extrapolate such responses to other biogeographically isolated islands, both in the literal sense (e.g. Great Britain, Ireland, New Zealand) and, to a lesser extent, to other areas where dispersal is restricted, as fresh waters are typically fragmented within an otherwise predominantly terres- trial landscape (Woodward & Hildrew, 2002). There are other specific predicted responses to warm- ing that we sought to test. For example, increased temperatures should stimulate primary production (e.g. Friberg et al., 2009) and, by extension, secondary production. This increased flux of energy to the higher trophic levels could therefore lead to a lengthening of food chains and thus the height of the trophic network as a whole. We tested this by quantifying the individual body mass of the top predators (brown trout), and by using stable isotope analysis to characterize their ���trophic height��� (e.g. after Vander Zanden et al., 2005). Given that oxygen depletion, enzymatic degradation and/or tissue damage become more prevalent above 35 1C, marked shifts in community structure should 1980 G . W O O D WA R D et al. r 2009 Blackwell Publishing Ltd, Global Change Biology, 16, 1979���1991