LETTER Experimental climate change weakens the insurance effect of biodiversity Johan S. Eklo ¨ f, 1,2,* Christian Alsterberg, 3 Jonathan N. Havenhand, 4 Kristina Sundba ¨ ck, 3 Hannah L. Wood 1 and Lars Gamfeldt 3 Abstract Ecosystems are simultaneously affected by biodiversity loss and climate change, but we know little about how these factors interact. We predicted that climate warming and CO 2 -enrichment should strengthen tro- phic cascades by reducing the relative efficiency of predation-resistant herbivores, if herbivore consumption rate trades off with predation resistance. This weakens the insurance effect of herbivore diversity. We tested this prediction using experimental ocean warming and acidification in seagrass mesocosms. Meta- analyses of published experiments first indicated that consumption rate trades off with predation resistance. The experiment then showed that three common herbivores together controlled macroalgae and facilitated seagrass dominance, regardless of climate change. When the predation-vulnerable herbivore was excluded in normal conditions, the two resistant herbivores maintained top-down control. Under warming, however, increased algal growth outstripped control by herbivores and the system became algal-dominated. Conse- quently, climate change can reduce the relative efficiency of resistant herbivores and weaken the insurance effect of biodiversity. Keywords Food web, Gammarus locusta, Littorina littorea, marine, mesograzer, metabolic theory of ecology, realistic biodiversity loss, response-effect traits, temperature, Zostera marina. Ecology Letters (2012) 15: 864–872 INTRODUCTION Biodiversity can provide insurance against disturbance and environ- mental change by increasing the chance that at least some resistant species can maintain ecosystem functioning when others fail (McNaughton 1977; Naeem & Li 1997; Elmqvist et al. 2003). Even though rarely stated, this ‘insurance hypothesis’ hinges on the assumption that the traits that determine how species respond to factors or changes in their environment (‘response traits’, e.g. the degree of physical and/or behavioural protection against predation) are uncorrelated with the traits that determine how species affect their environment (‘effect traits’, e.g. consumption rate, see Fig. 1a). For the insurance effect to operate, the traits should be distributed in such a way that at least some efficient species are also resistant enough to maintain functioning (Suding et al. 2008). In systems with such trait distributions, loss of vulnerable species (i.e. ‘realistic’ or ‘non-random’ loss) has weak effects, because most species are func- tionally redundant (Chapin et al. 1996; Walker et al. 1999). Theoreti- cally, biodiversity loss can also have negligible, or even positive, effects if response and effect traits are positively correlated (Fig. 1b). However, loss of vulnerable species will have stronger effects than those expected from random loss of species, when response and effects traits are negatively correlated (Fig. 1c) (e.g. Solan et al. 2004; Zavaleta & Hulvey 2004; Larsen et al. 2005; Bracken et al. 2008). In such circumstances high diversity provides weak insurance. It is a well-known fact that negative correlations between response and effect traits can be caused by evolutionary trade-offs, e.g. between competitive ability and resistance (Kneitel & Chase 2004). However, our understanding about which trait correlations dominate in differ- ent systems or contexts is limited (Suding et al. 2008), and this con- strains our ability to predict the consequences of biodiversity loss. In this article we hypothesise that the insurance effect of biodiver- sity can be altered by climate change. We herein define insurance as the ability of herbivore species diversity to buffer food webs against effects of predation on herbivores that cascade to plants. Our hypothesis links predictions based on the insurance hypothesis of biodiversity, with predictions about the roles of environmental warming and CO 2 -enrichment for plant-herbivore interactions. In theory, high herbivore diversity should weaken trophic cascades by increasing the chance that some herbivores are inedible to predators and maintain trophic control (Strong 1992; Duffy et al. 2007). How- ever, even though prey diversity can weaken direct predation effects (Hillebrand & Cardinale 2004; Edwards et al. 2010), evidence for a reduction in the strength of trophic cascades is mixed (Duffy et al. 2005; Kurle & Cardinale 2011), potentially because predation resis- tance often trades off with consumption rate (Gruner et al. 2008; Sieben et al. 2011). Climate change may strengthen or weaken these processes: in many ecosystems environmental warming increases metabolism and growth of both primary producers and ectotherm herbivores (Nemani et al. 2003; Chavez et al. 2011). Since respiration of ectotherm herbivores is more temperature-limited than photosyn- thesis (Allen et al. 2005), sub-lethal warming can strengthen top- down control (solid black line in Fig. 1d) (O’Connor 2009; O’Con- nor et al. 2011). It can also be predicted that even though trophic 1 Department of Biological and Environmental Sciences – Kristineberg, University of Gothenburg, 451 78, Fiskeba ¨ ckskil, Sweden 2 Department of Systems Ecology, Stockholm University, 106 91, Stockholm, Sweden 3 Department of Biological and Environmental Sciences – Gothenburg, University of Gothenburg, 405 30, Go ¨ teborg, Sweden 4 Department of Biological and Environmental Sciences – Tja ¨ rno ¨ , University of Gothenburg, 452 96, Stro ¨ mstad, Sweden *Correspondence: E-mail: johane@ecology.su.se © 2012 Blackwell Publishing Ltd/CNRS Ecology Letters, (2012) 15: 864–872 doi: 10.1111/j.1461-0248.2012.01810.x