USING TEMPORAL COHERENCE TO DETERMINE THE RESPONSETO CLIMATE CHANGE IN BOREAL SHIELD LAKESSHELLEY E. ARNOTT1,∗, BILL KELLER2, PETER J. DILLON3, NORMAN YAN4,MICHAEL PATERSON5and DAVID FINDLAY51Department of Biology, Queen’s University, Kingston, Ontario, K7L 3N6;2CooperativeFreshwater Ecology Unit, Ontario Ministry of the Environment, Sudbury, ON, P3E 2C6,3TrentUniversity, Peterborough, ON, Canada. K9J 7B8;4York University and the Ontario Ministry of theEnvir onment, 4700 Keele St. Toronto, ON, Canada, M3J 1P3;5Freshwater Institute, Department ofFisheries and Oceans, 501 University Crescent, Winnipeg, Manitoba, Canada, R3T 2N6(∗author for corr espondence, e-mail [email protected])Abstract. Climate change is expected to have important impacts on aquatic ecosystems. On theBoreal Shield, mean annual air temperatures are expected to increase 2 to 4◦C over the next 50years. An important challenge is to predict how changes in climate and climate variability willimpact natural systems so that sustainable management policies can be implemented. To predictresponses to complex ecosystem changes associated with climate change, we used long-term bioticdatabases to evaluate how important elements of the biota in Boreal Shield lakes have responded topast fluctuations in climate. Our long-term records span a two decade period where there hav e beenunusually cold years and unusually warm years. We used coherence analyses to test for regionallyoperating controls on climate, water temperature, pH, and plankton richness and abundance in threeregions across Ontario: the Experimental Lakes Area, Sudbury, and Dorset. Inter-annual variationin air temperature was similar among regions, but there was a weak relationship among regions forprecipitation. While air temperature was closely related to lake surface temperatures in each of theregions, there were weak relationships between lake surface temperature and richness or abundanceof the plankton. Ho wever, inter-annual changes in lake chemistry (i.e., pH) were correlated withsome biotic variables. In some lakes in Sudbury and Dorset, pH was dependent on extreme events.For example, El Nino related droughts resulted in acidification pulses in some lakes that influencedphytoplankton and zooplankton richness. These results suggest that there can be strong heterogeneityin lake ecosystem responses within and across regions.Keyw ords: climate change, long-term data, multiple stressors, phytoplankton, regional drivers, syn-chrony, temporal coherence, zooplankton1. IntroductionHuman activ ities are resulting in large-scale changes in aquatic ecosystems (Schin-dler, 2001). One of the most important stressors is climate change (Schindler,1998). There is strong evidence of both increasing global air temperatures and in-creasing temperature variability. Global surface temperatures have increased 0.6◦Csince 1861, with the 1990s being the warmest decade and 1998 being the warmestyear during the instrumental record (IPCC, 2001). Further increases are anticipated.For example, mean air temperature in the Boreal Shield is expected to increase 2 toEnvironmental Monitoring and Assessment 88: 365–388, 2003.© 2003 Kluwer Academic Publishers. Printed in the Netherlands.366 SHELLEY E. ARNOTT ET AL.4◦C over the next 50 years (Hengeveld, 2000). There is also evidence of increasedclimate variability associated with increased frequency of El Nino events (Urbanet al., 2000). An important challenge is to predict how these changes in climate andclimate variability will impact our natural systems so that sustainable managementpolicies can be implemented.The current and anticipated changes in climate will have complex and regionallyheterogeneous impacts on the physical, chemical, and biological characteristics oflake ecosystems (Magnuson et al., 1997). There are numerous anticipated effectsincluding potential changes in water clarity and thermal regimes associated withdeclines in dissolved organic carbon (DOC) inputs (Fee et al., 1996; Snucins andGunn, 2000), alteration of water chemistry including contaminants (Webster et al.,1996; Yan et al., 1996), and changes in the distribution of organisms (De Stasioet al., 1996; Vander Zanden et al., 1999; Leech and Williamson, 2001). Directchanges in temperature and precipitation and indirect changes resulting from com-plex responses of the physicochemical environment are likely to impact freshwaterecosystems through changes in species abundance, distribution, and composition.In addition to potential direct influences, there is evidence that climate change mayexacerbate the impact of other stressors such as invading species, pollution, andhabitat alteration (Schindler, 2001).To undertake the challenge of predicting biotic responses to complex ecosystemchanges associated with climate change, we used long-term databases to evaluateho w several physical, chemical, and biotic v ariables in Boreal Shield lakes hav eresponded to past fluctuations in climate. Our data records span a two decade periodwhere there ha ve been unusually cold years (associated with the Mt. Pinotubo erup-tion in 1992) and unusually warm years (associated with an exceptionally strongEl Nino in 1998). Lakes within a region will experience similar fluctuations inclimate and therefore, limnological v ariables strongly influenced by climate willbe expected to vary in a similar way through time. Individual lake characteristicssuch as water residence time and fish population cycles may create lags and devi-ations from the general regional response. Therefore, the similarity in lake responsewithin a region will depend on the relativ e importance of extrinsic (i.e., regional)characteristics versus intrinsic (i.e., individual lake) characteristics (Rusak et al.,1999). We used coherence analyses (e.g., Magnuson et al., 1990; Kratz et al.,1998) to test for regionally operating controls on climate, water temperature, pH,and plankton richness and biomass. Coherence is calculated as the mean Pear-son correlation coefficient for all lake-pairs and is a measure of the synchrony orsimilarity in variability through time. Plankton richness and biomass were chosenbecause pre vious studies have indicated they should respond to climatic signals(e.g., George and Harris, 1985; Stemberger et al., 1996) and because this inform-ation was av ailable for each of the sites. We hav e included pH in our analysesbecause a relationship between El Nino e vents and pH has been detected