UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Community context of adaptation to environmental change Kleynhans, Elizabeth J. 2018

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2019_february_kleynhans_elizabeth.pdf [ 13.24MB ]
Metadata
JSON: 24-1.0375772.json
JSON-LD: 24-1.0375772-ld.json
RDF/XML (Pretty): 24-1.0375772-rdf.xml
RDF/JSON: 24-1.0375772-rdf.json
Turtle: 24-1.0375772-turtle.txt
N-Triples: 24-1.0375772-rdf-ntriples.txt
Original Record: 24-1.0375772-source.json
Full Text
24-1.0375772-fulltext.txt
Citation
24-1.0375772.ris

Full Text

Community context of adaptation toenvironmental changebyElizabeth J. KleynhansB.Sc., University of the Witwatersrand, 2000B.Sc. with honours, University of the Witwatersrand, 2001M.Sc., University of Groningen, 2005A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinThe Faculty of Graduate and Postdoctoral Studies(Zoology)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)December 2018© Elizabeth J. Kleynhans 2018The following individuals certify that they have read, and recommend to the Faculty of Graduate and Post-doctoral Studies for acceptance, the dissertation entitled:Community context of adaptation to environmental changesubmitted by Elizabeth J. Kleynhans in partial fulfillment of the requirements for the degree of Doctor ofPhilosophy in ZoologyExamining Committee:Sarah Otto, ZoologyCo-supervisorMark VellendCo-supervisorDolph Schluter, ZoologySupervisory Committee MemberMichelle Tseng, Botany and ZoologyUniversity ExaminerGregory Henry, GeographyUniversity ExaminerAdditional Supervisory Committee Members:Amy Angert, BotanySupervisory Committee MemberiiAbstractHumans are causing rapid changes to the biotic and abiotic conditions on Earth. My thesis investigates howcompetition might shape adaptation to an altered environment. Intraspecific competition creates diversifyingselection, which alters the genetic variation available for adaptation. Using an individual-based model, Ifound that intraspecific variation altered the genetic variance-covariance matrix by pushing standing geneticvariation to more closely resemble available resources. This changed the “genetic line of least resistance”so that standing genetic variation and de novo mutation both provided possibilities for evolutionary rescuein different directions. Competition between species can also influence evolution to abiotic change. Usingan individual-based model I found that differences in population size and competitive ability, between twospecies, could facilitate coexistence and in some cases cause evolution to occur in the opposite direction tothat predicted from environmental change. In an empirical setting, I asked whether species diversity mayalter adaptation to abiotic change by changing population size, increasing genetic diversity, and/or by alteringselection experienced by a focal species. Using a reciprocal transplant experiment on grasses evolved for 14years under ambient and elevated CO2 conditions, in communities of low or high species-richness, I foundthat the biological community altered the nature of selection in elevated CO2, so that adaptation was observedprimarily when species were grown in a community similar to the one in which they were previously selected.Lastly, I tested whether functional traits of species observed today might reflect differences in the nature ofselection experienced in different biotic and abiotic environments. In contrast to expectation I only found themain effects of species diversity and abiotic change to influence plant functional traits. Overall, my researchhighlights an important role for species interactions in altering adaptation to abiotic environmental change,which cannot be overlooked when predicting how species will adapt to climate change.iiLay SummaryHumans are causing rapid changes to the biotic and abiotic conditions on Earth. My thesis highlights theimportant and often surprising effects of species interactions on adaptation to abiotic change. Using models,I found that competition within a species can alter whether populations adapt to environmental change vianew mutations or from variation already present in the population, and that competition between species cansometimes enhance species coexistence and occasionally cause populations to evolve in directions oppositeto that predicted from the direction of abiotic change. I also present field evidence showing that adaptationto abiotic change is observed primarily when species are grown in a community similar to the one in whichthey were previously selected. However, these differences in selection were not reflected in current planttraits. Overall, species interactions strongly influence evolution and cannot be overlooked when predictinghow species will adapt to climate change.iiiPrefaceChapter 2 is in preparation for publication in collaboration with S.P. Otto. I conceived the idea, wrote thesimulations and drafted the manuscript. All stages received considerable assistance from S. P. Otto.A version of Chapter 3 has been published as “Van Den Elzen C. L., Kleynhans E. J., & Otto S. P.(2017) Asymmetric competition impacts evolutionary rescue in a changing environment. Proc. R. Soc. B284: 20170374” with C. L. Van Den Elzen and me listed as co-first authors. S.P. Otto and I conceived theoriginal project, C.L. Van Den Elzen wrote the simulations with assistance from S.P. Otto and all authorsran the simulations. C.L. Van Den Elzen, S.P. Otto and myself jointly wrote the manuscript. My specificcontribution to the writing was to write the introduction and discussion sections of the manuscript and providemajor revisions on the methods and results sections. The completed manuscript was jointly edited by allauthors.A version of Chapter 4 has been published as “Kleynhans, E. J., Otto, S. P., Reich, P. B., & Vellend, M.(2016). Adaptation to elevated CO2 in different biodiversity contexts. Nature Communications, 7:12358”I conceived the project together with S.P. Otto and M. Vellend. P. B. Reich designed, implemented, andmaintained the original BioCON experiment in which this study was performed. I carried out the fieldwork,analysed the data and wrote the first draft of the manuscript. All authors jointly edited and wrote subsequentdrafts.Chapter 5 is in preparation for publication with coauthors M. Vellend and P.B. Reich. M. Vellend cameup with the original idea which I built upon. P. B. Reich designed, implemented, and maintained the originalBioCON experiment in which this study was performed. I carried out the fieldwork, performed the statisticalanalyses and drafted the manuscript. M. Vellend contributed to manuscript edits.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Effect of species interactions on adaptation to abiotic change . . . . . . . . . . . . . . . . . 31.1.1 Do species interactions constrain adaptation to abiotic change? . . . . . . . . . . . 31.1.2 Do species interactions promote adaptation to abiotic change? . . . . . . . . . . . . 51.1.3 Do species interactions alter the selective landscape? . . . . . . . . . . . . . . . . . 52 Evolutionary rescue when resources change. . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.2.1 Ecological Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.2.2 Genetic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.2.3 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 Asymmetric competition impacts evolutionary rescue in a changing environment . . . . . . 213.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28vTable of Contents3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314 Adaptation to elevated CO2 in different biodiversity contexts . . . . . . . . . . . . . . . . . . 354.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374.2.1 Sampling design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374.2.2 Growth in a common greenhouse environment . . . . . . . . . . . . . . . . . . . . 404.2.3 Planting ramets in the assay plots . . . . . . . . . . . . . . . . . . . . . . . . . . . 404.2.4 Measurement of plant performance . . . . . . . . . . . . . . . . . . . . . . . . . . 414.3 Statistical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414.3.1 Analysis of biomass data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414.3.2 Analysis of survival and inflorescence production . . . . . . . . . . . . . . . . . . 434.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.4.1 Results for biomass production . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.4.2 Results for survival and inflorescence production . . . . . . . . . . . . . . . . . . . 444.4.3 Analysis with selection and assay environments . . . . . . . . . . . . . . . . . . . 444.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465 How plant diversity influences intraspecific trait responses to abiotic environmental change . 485.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485.1.1 Response of species from different functional groups to elevated CO2 . . . . . . . . 505.1.2 Response of species from different functional groups to elevated nitrogen . . . . . . 505.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515.2.1 BioCON experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515.2.2 Trait measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525.3 Statistical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 605.5.1 Community context has no detectable impact on how functional traits respond toabiotic change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 605.5.2 Trait responses to functional group richness . . . . . . . . . . . . . . . . . . . . . . 635.5.3 Trait responses to the abiotic environment . . . . . . . . . . . . . . . . . . . . . . 636 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656.1 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656.2 Conclusions and future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656.2.1 Chapter 2: Evolutionary rescue when resources change. . . . . . . . . . . . . . . . 656.2.2 Chapter 3: Asymmetric competition impacts evolutionary rescue in a changing en-vironment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676.2.3 Chapter 4: Adaptation to elevated CO2 in different biodiversity contexts . . . . . . 68viTable of Contents6.2.4 Chapter 5: How plant diversity influences intraspecific trait responses to abiotic en-vironmental change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 696.3 Future directions and general conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 70Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90A Appendix for Chapter 2: Evolutionary rescue when resources change. . . . . . . . . . . . . 91A.1 Supplementary results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91A.2 Supporting Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92B Appendix for Chapter 3: Asymmetric competition impacts evolutionary rescue in a changingenvironment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95B.1 Supplementary Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95B.2 Supporting Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97B.3 Supporting Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99C Appendix for Chapter 4: Adaptation to elevated CO2 in different biodiversity contexts . . . 105C.1 Supplementary Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105C.1.1 Analysis with assay environment instead of change in environment. . . . . . . . . . 105C.1.2 Correlation between aboveground biomass and percent cover of species in the assayplot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106C.1.3 History and construction of experimental assay plots . . . . . . . . . . . . . . . . . 106C.2 Supporting Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107C.3 Supporting Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116D Appendix for Chapter 5: How plant diversity influences intraspecific trait responses to abioticenvironmental change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120D.1 Supplementary Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120D.1.1 Analysis with species richness instead of functional group diversity. . . . . . . . . . 120D.2 Supporting Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120D.3 Supporting Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125viiList of Tables2.1 Default model parameters used in the individual-based simulations. . . . . . . . . . . . . . 133.1 Model parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2 Extinction time of populations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285.1 PERMANOVA results for functional group diversity and abiotic treatment . . . . . . . . . . 575.2 Mixed effect model results for the response of plant traits to CO2, nitrogen and the functionalgroup richness of the plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57B.1 Equilibrium (initial) population configurations . . . . . . . . . . . . . . . . . . . . . . . . . 97B.2 Extinction time of populations, in generations, with higher mutation rate (ν) or mutationaleffect sizes (σν) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98C.1 Model output for analysis of aboveground, belowground, total biomass and survival andinflorescence data, when diversity environment is held constant . . . . . . . . . . . . . . . . 108C.2 Model output for analysis of aboveground, belowground, total biomass and survival andinflorescence data, when average over the diversity assay environment . . . . . . . . . . . . 109C.3 Linear mixed effects model analysis of log-transformed aboveground, belowground and totalbiomass data for ∆CO2 and ∆diversity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110C.4 Linear mixed effects model analysis of log-transformed aboveground, belowground and totalbiomass data for selection and assay environments . . . . . . . . . . . . . . . . . . . . . . . 111C.5 Summary of Aster model comparisons on P. pratensis survival and inflorescence production 112C.6 Summary of Aster model comparisons to test for differences in P. pratensis survival andinflorescence production as a result of the selection and assay CO2 . . . . . . . . . . . . . . 113C.7 GLM of aboveground biomass versus percent cover of species . . . . . . . . . . . . . . . . 114C.8 GLM analysis of aboveground biomass and percent cover by functional group . . . . . . . . 114C.9 Percent cover of species grown in the assay and in the selection (BioCON) plots . . . . . . . 115D.1 Number of plots sampled . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121D.2 Sample sizes per species and trait in ambient and elevated CO2 . . . . . . . . . . . . . . . . 122D.3 Sample sizes per species and trait in ambient and elevated nitrogen . . . . . . . . . . . . . . 123D.4 PERMANOVA results for abiotic treatment and species richness . . . . . . . . . . . . . . . 124viiiList of Figures2.1 The shape of the G-matrix resulting from a balance of selection and mutation . . . . . . . . 92.2 Probability of rescue for the genetic model from standing genetic variation, de novo muta-tion, or both as a function of the new environmental optimum in two dimensions . . . . . . . 162.3 Probability of rescue for the ecological model from standing genetic variation, de novo mu-tation, or both as a function of the new environmental optimum in two dimensions . . . . . . 172.4 Probability of rescue from standing genetic variation, de novo mutation, or both in the geneticand ecological models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183.1 Symmetric and asymmetric resources and their impact on population size and trait value. . . 263.2 The evolution of two-species communities competing for asymmetrical resources in a chang-ing environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.3 Higher adaptability can allow a lagging species to out-evolve a leading species in a rapidlychanging environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314.1 The hypothetical influence of species richness on adaptation of plants to elevated CO2 . . . . 384.2 Steps involved in the transplant experiment. . . . . . . . . . . . . . . . . . . . . . . . . . . 394.3 Response to selection under elevated versus ambient CO2 in species-poor and species-richcommunities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455.1 Trait responses of C3 grasses and C4 grasses to functional group diversity and the abioticenvironment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.1 Trait responses of legumes and forbs to functional group diversity and the abiotic environment 595.2 Response of plant traits to species richness, CO2 enrichment and nitrogen enrichment . . . . 61A.1 Mutation selection balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92A.2 Probability of rescue as a function of the new environmental optimum for the genetic andecological models when the fitness or resource distribution variances are unequal . . . . . . 93A.3 Probability of rescue from standing genetic variation or de novo mutation in the genetic andecological models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94B.1 Symmetric and asymmetric competition coefficients and their impact on population size andtrait value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99B.2 Asymmetrical resource distributions and the evolution of a one-species community in achanging environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100ixList of FiguresB.3 The evolution of two-species communities with asymmetrical competition coefficients in theface of environmental change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101B.4 The evolution of a one-species community with asymmetrical competition coefficients in theface of environmental change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102B.5 Evolutionary response to a changing environment with a higher mutation rate . . . . . . . . 103B.6 Evolutionary response to a changing environment when mutation effect sizes are drawn froma broader distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104C.1 Local adaptation of P. pratensis to elevated CO2 when holding diversity environment constant 116C.2 Adaptation of P. pratensis to elevated CO2 when selected in communities of low and highspecies richness averaged across diversity assay environments . . . . . . . . . . . . . . . . 117C.3 Average aboveground, belowground, and total biomass and number of inflorescences pro-duced by plants originating from each of the selection environments and grown in each ofthe assay environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118C.4 Photographs of the original selection and assay plots . . . . . . . . . . . . . . . . . . . . . 119D.1 Traits responses of C3 and C4 grasses to CO2 and nitrogen when grown in plots with one forfour functional groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126D.2 Traits responses of legumes and forbs to CO2 and nitrogen when grown in plots with one orfour functional groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128D.3 Trait responses of C4 and C3 grasses to the abiotic environment and species richness . . . . 130D.3 Trait responses of legumes and forbs to the abiotic environment and species richness . . . . 131xAcknowledgementsThere are many people I need to thank for this thesis as it could not have been possible or nearly as enjoyablewithout the help of lots of people.First off, I’d like to thank my supervisors, Sally Otto and Mark Vellend. Thank you so much for yourhelp, you were both the best supervisors I could ever have wished for. I am very luck and very grateful foryour help, guidance, and support, and the work I produced here could not have been done without you. Sally,you were my primary supervisor and you were amazing. I appreciate your style of guidance, so much so thatnow whenever I am in the position of supervising someone I always ask myself "what would Sally Otto do".Thank you for being such a fantastic role model.Next I’d like to thank my committee, Dolph Schluter and Amy Angert. You both provided comments onmy project throughout and gave helpful statistical advice on the Nature Communications paper. Thank youfor all your input and support.Much of this project would not have been possible without the help of Peter Reich, the LTER networkand the Cedar Creek Field staff and crew. Thank you Peter for being so supportive, open and enthusiasticabout my project and allowing me to perform the transplant experiment in BioCON plots. I really appreciatedthat you joined me in the field and gave me suggestions and very helpful ideas across the duration of thisproject. The support you provided me was invaluable. Kally Worm, thank you for always being available tohelp and also for organising equipment and field staff when I needed it. The rest of the field staff and internsat Cedar Creek also provided me with lots of support and friendly smiles. Thank you for everything. Mytime at Cedar Creek would not have been as enjoyable or stimulating if it had not been for the other gradstudents and post docs. In particular, I’d like to thank Sasha Wright, Peter Wragg, Adam Clark, Jane Cowles,Meredith Steck, Clare Kazanski, Jane Catford and Chandra Moffat.Various people helped me with data collection and/or sample processing. Most importantly, I’d like tothank Angelica Lillico-Ouachour who was my field assistant for two field seasons. Without your help I couldnot have completed the field part of my project. Thank you for putting up with me at 6am, in 40+C heat,and while being carried away by mosquitos and deer flies. I also need to thank Kelsie Beclklin for help withthe traits project. Numerous other people provided help at various times through out this project includingTom Zochowski, Pearl, Andy An, and a whole host of people that helped me wash roots. I also need to thankCourtney Van Den Elzen for her contribution to Chapter 3. It was really fun working on that project withyou.Thank you to the Otto lab for listening and trying to understand my lab presentations and generallyfor many other fun times. During my time here I overlapped with Aleeza Gerstein, Rich FitzJohn, KateOstevik, Jasmine Ono, Nathaniel Sharp, Leithen M’Gonigle, Matt Osmond, Michael Scott, Carl Rothfels,Flo Debarre, Ailene McPherson, Linnea Sandell, and Shahab Zareyan.xiAcknowledgementsThank you to everyone in the BRC for making grad school a fun experience overall. In particular I’d liketo thank Julie Lee-Yaw, Kim Gilbert, Megan Bontranger, Matt Siegle, Holly Kindsvater, Jeremy, Draghi,Chris Lee, Chris Muir, Emily Gereck, Adam Wilkinson, Bridgette Clarkson and Paolo Segre.I’d like to thank the funding sources that made this project possible: The Four-Year Fellowship fromUBC, the Rosemary Grant Award for Graduate Student Research and the Natural Sciences and EngineeringResearch Council (NSERC) of Canada Discovery grants to Sarah Otto and Mark Vellend.I’d like to thank my friends external to the BRC for distracting me from my thesis. The time spent skiing,running or just hanging out really energised me to work harder.Last and not least I’d like to thank my family. Thank you Helen, Charles and Martin Kleynhans for allyour support through out my studies and for being understanding and encouraging when I chose to moveto Vancouver. I will never forget Martin pulling out the globe to firstly show me where Vancouver was andsecondly to point out that I could not have chosen a more distant location on this planet to do my PhD. ToBill and Freya Harrower thank you for all your support. Thank you Bill for being there for me throughout,I could not have done it without you and to Freya for being the best possible distraction over the past 15months. Both of you made coming home after hours of working the most enjoyable experience.xiiDedicationTo my family Bill, Freya, Charles, Helen and MartinxiiiChapter 1IntroductionHumans are altering biotic and abiotic conditions on Earth at a rapid rate. The last three decades havebeen warmer than any period for the last 1400 years and atmospheric concentrations of CO2, methane andnitrous oxide are higher than they have been for more than 800 000 years (Pachauri and Meyer, 2014). Inresponse to warming, many species are shifting their ranges towards higher latitudes and elevations (Chenet al., 2011; Parmesan and Yohe, 2003). However, species differ in the rate at which they can track climaticchanges due to differences in mobility, habitat specialization and physiological tolerances (Chen et al., 2011).Furthermore, alien species are being introduced to new areas at rates faster than observed at any other timein human history (Early et al., 2016). Together these changes result in contemporary communities beingreshuffled (Walther et al., 2002), and thus for species to avoid extinction species are required to adapt to newabiotic conditions while interacting with novel species. Therefore, understanding how species will adapt tochanges in their abiotic and biotic environment while living in complex ecosystems is one of the most criticalchallenges facing biologists (Alexander et al., 2015; Barraclough, 2015). Although this is an enormous task,my thesis attempts to fill a portion of this knowledge gap by investigating how competition might shapeadaptation to an altered environment.There are three ways in which species can respond to abiotic and biotic change in a local environment;they can migrate or disperse to track abiotic conditions, they can respond via phenotypic plasticity or theycan adapt via genetic change. The main focus of this thesis is on genetic change, but I will briefly describephenotypic plasticity here because it is a factor that likely plays a significant role in influencing how speciesrespond to biotic and abiotic change and is present in my empirical studies.Phenotypic plasticity is the ability of a single genotype to express different phenotypes in different en-vironments and may entail changes in morphology, physiology, and behaviour (Schlichting, 1986; Stearns,1989). Within plants, phenotypic plasticity is known to be important because many traits, such as leaf size orplant height are known to respond rapidly to changes in biotic and abiotic conditions (Nicotra et al., 2010).Phenotypic plasticity can also occur over longer time scales, for example it can be developmentally triggeredso that exposure to an environment early in life alters the trajectory of development to modify performancein that environment later on. Or, it might occur across generations (transgenerational) where environmen-tal experiences of parents influence the performance of offspring (reviewed in Bonduriansky et al., 2012).Theoretically, phenotypic plasticity may play two contrasting roles in adaptation. On the one hand, it mightallow species to persist for longer providing them with more time to adapt genetically to altered conditions(Chevin, 2010). Alternatively, it might retard genetic adaptation by moving the species closer to the fitnesspeak of the trait distribution, reducing the gradient of selection (Sunday et al., 2014).In contrast to phenotypic plasticity, genetic adaptation is a more enduring change where natural selec-1Chapter 1. Introductiontion leads to changes in some characteristic/s of an organism that improve its survival or reproduction in itsenvironment (Futuyma, 2009). Genetic adaptation to recent changes in abiotic conditions has been observedin many different taxa (reviewed in Bradshaw and Holzapfel, 2006; Jump and Penuelas, 2005; Otto, 2018).Examples include shifts in bird egg laying dates in response to earlier springs (Nussey et al., 2005), modi-fication of the diapause cue in mosquitoes in response to longer growing seasons (Bradshaw and Holzapfel,2001), and the evolution of metal-tolerance in plants growing on contaminated soils on a mine boundary(Antonovics, 1971). These examples demonstrate that species have the capacity to adapt to a variety ofdifferent environmental circumstances. Yet, there must also be a limit to how readily species can adapt (re-viewed in Blows and Hoffmann, 2005), because if there were not, every species would occur everywhere.Thus, understanding the limits to adaptation and whether some traits have more potential for evolution thanothers would aid our understanding of how species will adapt to biotic and abiotic change.One common way to investigate adaptive potential in natural populations is to measure the amount ofstanding genetic variation. For traits under selection, traits with more genetic variation are thought to fa-cilitate faster evolution, biasing evolution in that direction. This idea has been met with some support. Forexample, Schluter (1996) found that morphological diversification in recently diverged species was biasedtowards the direction with greatest genetic variation (also see Begin and Roff, 2004; Renaud et al., 2006).However, other studies have found no such relationship (e.g. Berner et al., 2008, 2010; Eroukhmanoff andSvensson, 2011). Standing genetic variation is shaped by various factors including mutation, selection,drift, migration, and the past history of environmental change (reviewed in Arnold et al., 2008; Kopp andMatuszewski, 2014). However, no one has investigated how competition might influence the amount ofstanding genetic variation. I address this gap in Chapter 2 by investigating how intraspecific competitionmight influence evolution in response to an abrupt environmental change using an individual-based, geneti-cally explicit model. In contrast to previous results I find that some directions of evolutionary change can bepredicted from standing genetic variation while new mutations are more likely to contribute to evolution inother directions.Competition can play an important role in influencing the direction of evolutionary change, but typicallythe effects of species interactions are studied in isolation from the effects of abiotic change on evolution. Thisis highlighted by the entire field of coevolution that describes the reciprocal evolution of traits in species in-teracting with one another (Futuyma and Slatkin, 1983), almost exclusively in the absence of abiotic change.Some fascinating examples of evolutionary change in response to species interactions include the alterationof lodgepole pinecone morphology as a result of the presence or absence of red squirrels and crossbills.When crossbills are the only cone predator, coevolution between beak size and cone structure was observed.However, when crossbills and squirrels are both present, squirrels primarily drive cone morphology, andcrossbill beak structure evolves to match the cones surviving squirrel predation (Benkman et al., 2003).In plants, character displacement has been found in response to growing in experimental plots withmultiple species, where competition results in selection favouring genotypes with more distinct niches -relative to plants of the same species growing in monocultures (Zuppinger-Dingley et al., 2014). Thus, boththe biotic and abiotic environments can be strong selective forces and although they often covary, e.g. theintroduction of a species to a new habitat, scientists have still typically studied them independently. My21.1. Effect of species interactions on adaptation to abiotic changethesis aims to fill this gap by investigating whether the biotic environment alters evolutionary adaptation toabiotic change (see Chapters 3, 4, and 5).1.1 Effect of species interactions on adaptation to abiotic changeHere, I will briefly describe the main impacts that species interactions might have on adaptation to abioticchange and describe how the rest of my thesis chapters address some of the gaps in our knowledge.In Chapters 4 and 5 I report field studies conducted to gain empirical evidence of how community con-text might influence adaptation to abiotic change. More specifically I performed a reciprocal transplantexperiment on Poa pratensis (Chapter 4) and measured the traits of seven species (Chapter 5) growing in along-term ecological experiment known as BioCON (Biodiversity Carbon dioxide and Nitrogen experiment)at the Cedar Creek Ecosystem Sciences Reserve in Minnesota. Briefly, BioCON was initiated in 1997 withthe aim of investigating how communities will respond to increasing nitrogen, increasing CO2 and decliningbiodiversity. As such, 296 individual plots (2 × 2 m) distributed among six 20m diameter experimentalrings were constructed. In three of the rings CO2 is added to increase the current CO2 concentration by180 ppm; the other three are maintained at ambient conditions. In 1997 all plots were planted with 1, 4, 9or 16 perennial grassland species and from 1998 half the plots received additional nitrogen (4 g.N.m-2.yr-1)applied at three dates each year. This fully factorial manipulation of biotic and abiotic conditions provides aunique and fantastic opportunity to test how community context might have influenced adaptation to CO2 ina semi-natural setting. Furthermore, at the time of my study this experiment had been running for 14 years,allowing possible evolutionary changes in the species in response to the biotic and abiotic manipulations.Although I focus on adaptation as a result of genetic changes for most of this thesis and attempt to controlfor the effects of phenotypic plasticity through raising individuals in a common environment for at least onegeneration (Chapter 4), I cannot fully discount the importance of transgenerational effects in Chapter 4. InChapter 5 phenotypic plasticity probably plays a considerable role and this needs to be kept in mind whileconsidering my results.With the design of BioCON in mind, I return now to discuss how species interactions might influenceadaptation to abiotic change, with a focus on factors that might constrain, augment, or to alter the directionof adaptation (see also Chapter 4 and Barraclough (2015)).1.1.1 Do species interactions constrain adaptation to abiotic change?Species subject to predation, parasitism, or competition for a finite resource are likely to have smaller pop-ulation sizes than in the absence of these species interactions. Smaller population sizes typically have lessstanding genetic variation (Bocedi et al., 2013; Lanfear et al., 2014; Frankham, 1996) and fewer beneficialmutations. As a consequence species interactions might reduce the capacity of a species to adapt to abioticchange.Empirical studies that have documented population sizes and adaptation to abiotic change all includedpredators. Of these, support for the hypothesis of smaller population sizes leading to reduced adaptationis mixed. Schoener et al. (2001) found that predators increased the risk of extinction in prey populations31.1. Effect of species interactions on adaptation to abiotic changefollowing a severe environmental change, likely as a result of reduced prey population sizes. In contrast,Tseng and O’Connor (2015) found no such response, instead noting that predation selected for higher pop-ulation growth rates when evolved to increased temperature. This unexpected result might be due to anevolutionary "push" from predators that consume maladapted prey thereby increasing the rate of evolution(Jones, 2008; Osmond et al., 2017). Alternatively it could be due to the "evolutionary hydra effect" wheredecreased prey density selects for faster generation times and hence increased rates of evolution (Osmondet al., 2017). Whatever the reason, the impact of reduced population size due to predation does not alwaysconstrain evolution and instead may be context dependent.In Chapter 4, I also do not find evidence that species mixtures constrain adaptation. In this chapter(Chapter 4) I transplanted individuals evolved in communities of low and high species richness in ambientand elevated CO2 into all combinations of diversity and CO2, including the combination of CO2 and diversitythey had been selected in for the past 14 years. This design allowed me to determine whether populationsselected in a specific environment had adapted to those conditions and also allowed me to infer how thisadaptation had occurred. One prediction was that species growing in more diverse communities should havesmaller population sizes due to competition for finite resources and therefore reduced adaptive capacity, incomparison to species growing in less diverse communities. Biomass of individual species in 16-speciesplots is significantly less than that of the same species in monoculture, making the assumption of smallerpopulation size in high diversity communities a reasonable one. However, we did not find evidence to supportthis prediction indicating that population size may play a more minor role in adaptation to abiotic change inthis system.As far as I am aware no empirical studies (besides my Chapter 4) have directly investigated the impactof competition on population size and adaptation to abiotic change. However, various theoretical studieshave explored this question. These theoretical studies all induced a form of ecological sorting, where ina shifting environment, one (or more) species possess traits that pre-adapt it to the changing conditions sothat the pre-adapted species increases in population size, imposing more competition and driving down theabundance of the less adapted competitor species (Bocedi et al., 2013; de Mazancourt et al., 2008; Johansson,2008; Norberg et al., 2012). In these previous studies, the species that possess traits that pre-adapt it to thechanging conditions is always the one more likely to survive the change. However, these models typicallyassume that in a stationary environment competing species have equal population sizes and competitiveabilities. In nature, equivalent population sizes or competitive abilities are rare. Thus, in Chapter 3 weinvestigate the impact on adaptation to abiotic change of differences in population size or competitive abilityprior to the environmental shift. Interestingly, we find that asymmetric resource availability or competitioncan facilitate coexistence and even occasionally cause the "leading" species to go extinct. Furthermore, wealso find cases where traits evolve in the opposite direction to that of the changing environment because asthe lagging species declines in size not all resources the lagging species can access are utilised creating a"vacuum" of competitive release.41.1. Effect of species interactions on adaptation to abiotic change1.1.2 Do species interactions promote adaptation to abiotic change?Conversely, species interactions might enhance adaptation to abiotic change if species richness promotesdiversifying selection allowing different genotypes to be favoured in competition with different species (Vel-lend and Geber, 2005). Although not examined in relation to environmental change, Booth and Grime (2003)found that communities of grassland species planted with higher genotypic diversity retained greater speciesdiversity over time than those with less genotypic diversity. In addition, Zuppinger-Dingley et al. (2014)demonstrated selection for character displacement in diverse communities. Taken together, these studiessuggest that species diversity can promote genetic diversity but whether this leads to enhanced adaptationhas not previously been studied.In Chapter 4 we tested this idea by looking for improved fitness in plants selected and assayed in com-munities of higher diversity relative to those selected and assayed in species poor communities. However,we found no support for this prediction suggesting that species interactions promoting genetic diversity maynot improve adaptation to abiotic change.Another way that species diversity might promote genetic diversity is through gene transfer or hybridiza-tion (reviewed in Barraclough, 2015). For example, in microbes horizontal gene transfer could increasegenetic diversity and facilitate the acquisition of new functions. In more diverse communities the chancethat another species has a compatible donor genome or that a plasmid or virus is present that will transfergenes may increase.In eukaryotes, examples of hybridisation improving adaptation do exist such as warfarinresistance in mice (Song et al., 2011) but how common this is and whether it occurs at high enough rates tofacilitate adaptation to abiotic change remains unknown.1.1.3 Do species interactions alter the selective landscape?Finally, species interactions might alter the nature of selection so that species adapt to abiotic change indifferent ways in different community contexts. For example, in species poor communities elevated CO2has been found to reduce belowground microbial biomass while the opposite is true in species rich com-munities, thereby potentially altering the availability of nutrients to plants. These differences in growingenvironment suggest that the adaptive landscapes may have changed, resulting in a species adapting to abi-otic change along different paths in different community contexts. Evidence to support this hypothesis comesfrom Lawrence et al. (2012) who studied decomposer bacteria communities in the lab and found that speciesevolved more and along different trajectories when adapted to a novel environment in communities of in-teracting species than when evolved to the same novel environment in monoculture (for an example withDaphnia see Van Doorslaer et al., 2010).The BioCON experiment described in Chapter 4 provides evidence for species interactions altering theselective landscape so that species adapt to abiotic change in different ways in different community contexts.If the diversity of the community alters selection pressures for adaptation to abiotic change, then differ-ences in selective pressures might influence the functional traits of plants. In Chapter 5 we tested this ideaby measuring the functional traits of seven different species grown in communities of low and high diversityin ambient and elevated CO2 and nitrogen in BioCON. Interestingly we found no interaction between thediversity of the plot and the abiotic environment. However we did find a strong main effect of the diversity51.1. Effect of species interactions on adaptation to abiotic changeof the plot and a significant but less strong effect of the abiotic environment. Taken together these resultssuggest that the composition of a community likely influences adaptation and may even play a larger rolethan abiotic change, but this needs further investigation.Overall, my thesis highlights the important and often surprising effects of species interactions on adap-tation to abiotic change. In particular my results demonstrate that species interactions can cause species toevolve in unexpected directions. For example, in Chapter 2 we demonstrated that diversifying selection forresources can change the shape of the genetic variance covariance matrix so that standing genetic variation isnot always the best predictor of directions of rapid evolutionary change. In Chapter 3 we demonstrated thatspecies can evolve in directions opposite to that predicted by the direction of environmental change whena competitor species is declining rapidly in abundance leaving a vacuum of competitive release. In Chap-ter 4 we demonstrated that the community can alter the adaptive landscape such that species are adaptingto different selective peaks in different community contexts. Nevertheless, examining functional traits, asdone in Chapter 5, to determine whether past biotic and abiotic selective pressures leads to context-specificdifferences in functional traits did not reveal significant results. Instead, it may be that selective pressuresfrom the biotic environment overwhelm abiotic selective pressures so that trait optima in different commu-nity contexts are different but stable over time, but this remains to be tested. Overall, my results suggestthat species interactions play an important role in altering adaptation to abiotic environmental change andstrongly suggest that we cannot ignore species interactions when trying to understand how species will adaptto altered environmental conditions.6Chapter 2Evolutionary rescue when resources change.2.1 IntroductionWhen a population undergoes an abrupt change in its environment, its persistence depends on its toleranceof the novel condition (which could involve phenotypic plasticity), its genetic adaptation to them, or on ashift in geographic range to a location with conditions it tolerates. If the environmental change is large andtolerance of this change relies on genetic adaptation because phenotypic plasticity is insufficient to match themagnitude of change and dispersal is limited (Schloss et al., 2012), genetic adaptation may be a race betweena population declining in size and the spread of alleles with a fitness greater than one, rescuing the populationfrom extinction. This form of persistence is known as evolutionary rescue (Bell, 2013) and has been exten-sively studied both theoretically and in the laboratory (reviewed in Alexander and Bonhoeffer, 2012; Carlsonet al., 2014; Gomulkiewicz and Shaw, 2012; Gonzalez et al., 2013; Kopp and Matuszewski, 2014). Theoret-ical studies in particular have highlighted the importance of the initial population size and genetic diversity,rate of environmental decay, mutation rate, and the fitness effect of adaptive mutations (Gomulkiewicz andHolt, 1995; Orr and Unckless, 2008; Bell and Gonzalez, 2009; Gomulkiewicz and Houle, 2009; Willi andHoffmann, 2009; Ramsayer et al., 2012; Martin et al., 2012; Osmond and de Mazancourt, 2013). Manyof these models, however, have assumed that the traits that evolve are not related to resource competitionor predator avoidance. Competition for resources and avoidance of predators that learn are important be-cause they both can generate diversifying selection such that common phenotypes are less fit, either becausethey compete more strongly for the resources (Bolnick, 2001) or are targeted more frequently by predators(Vamosi, 2005), in contrast to previous models that have only considered stabilising selection. For ease ofreference, we consider resource competition to be the source of negative frequency dependent selection, butwe emphasize that other interactions could also generate this form of selection. Specifically, we compare theresults of a quantitative genetic model with stabilizing selection versus one with intraspecific competitionfor resources. We shift the resource or selective optimum in discrete steps away from its initial positionand investigate whether the probability of rescue from standing genetic variation or from de novo muta-tion changes. Our investigation uses simulations that incorporate resource competition, density-dependentgrowth, and stochastic fluctuations.One concept that is central to understanding how populations might evolve in response to a shift intheir environment is the genetic variance-covariance matrix (G-matrix) (Lande, 1979; Arnold, 1992). TheG-matrix describes the inheritance of multiple traits that are each affected by multiple genes and portrays therelationship between traits in matrix form by representing the additive genetic variance within traits on thediagonal and additive genetic covariation between traits on the off diagonal (Falconer and Mackay, 1996).72.1. IntroductionGraphically, the shape of the G-matrix can be represented in two (or more dimensions) by a cloud of geneticvalues with eigenvectors describing directions of most and least genetic variation (Arnold et al., 2008).Traits with a large amount of genetic variation facilitate rapid evolution along that trait axis while traitswith very little genetic variation evolve more slowly. Theoretical (reviewed in Arnold et al., 2008; Koppand Matuszewski, 2014) and empirical (e.g. Begin and Roff, 2004; Renaud et al., 2006; Schluter, 1996)work show that the multivariate direction with the largest amount of genetic variation (gmax) can bias thetrajectory of evolution. For example, Schluter (1996) found that evolution in recently diverged species andpopulations was more biased towards the direction of gmax than more distantly diverged species, naming thisthe “genetic line of least resistance” and demonstrating that the G-matrix can be used to predict the trajectoryof evolution, assuming the G-matrix is stable over time. Consequently, understanding what factors influencegenetic variances within traits and covariances between traits is imperative for understanding and predictingthe trajectory of evolution. Factors such as mutation and selection (Arnold et al., 2008; Björklund et al.,2013; Jones et al., 2003; Shaw et al., 1995; Wilkinson et al., 1990), drift (Phillips et al. 2001), migration(Guillaume and Whitlock, 2007) and the history of past environmental change (Draghi and Whitlock, 2012)have all been found to influence the shape of the G-matrix. However, no one has explored the impact ofintraspecific competition on the shape of the G-matrix and how this might alter evolutionary rescue.In theoretical models, the G-matrix is most easily represented in two dimensions and is a balance be-tween mutational variance (represented by the mutational matrix) and the adaptive landscape. For example,if mutations generate more variation along the vertical trait axis and the adaptive landscape is symmetricaland bell-shaped the resulting G-matrix will be oval in shape (Figure 2.1 A) (Arnold et al., 2008). Undersuch a quantitative genetic framework, Gomulkiewicz and Houle (2009) examined adaptation to an abruptenvironmental change that shifts the trait optimum along the dimension with least genetic variation andidentified evolutionary constraints with respect to the rate of environmental change, population size, andheritability. However, this model assumed that fitness was density independent and that populations couldgrow infinitely in size after rescue. Furthermore, the stochastic dynamics of finite populations were not con-sidered. We test the consequences of these assumptions by explicitly incorporating intraspecific competitionin finite populations. Competition will have an important impact on the G-matrix because fitness depends onboth the availability of resources and the trait distribution among competitors. Because rare individuals ableto consume resources left over by others are favoured, we predict that intraspecific competition will causethe G-matrix to look more similar to the resource landscape (Figure 2.1 B). Consequently, the diversifyingselection caused by competition might have important implications for predicting the trajectory of evolution.Large effect mutations are more likely to occur in the direction of greatest mutational variation and if thisis different than gmax then adaptation may be facilitated in different directions; depending on whether mu-tations or standing variation allow rescue. Indeed, support for an evolutionary bias towards gmax is mixed(e.g. Begin and Roff, 2004; Berner et al., 2008, 2010; Eroukhmanoff and Svensson, 2011).The importance of standing genetic variation and new mutations in rescuing a population from extinctionvaries depending on whether density dependent competition is included in the model. Models that do notinclude competition have found that rescue from standing genetic variation may be more important when theextent of environmental change is not very severe and when the rescue alleles were previously neutral or only82.1. Introductionmarginally deleterious (Orr and Unckless, 2008). Such variation is immediately available for selection to actupon (Barrett and Schluter, 2008; Carlson et al., 2014; Hermisson and Pennings, 2005; Orr and Unckless,2008). However, if the environmental change is severe, the rescue alleles were strongly deleterious in theoriginal environment, or the rate of population decline to extinction is slow, de novo mutation may be criticalfor rescue (Bradshaw, 1991; Orr and Unckless, 2008). In contrast when population structure and density-dependent competition is included in one-locus, two allele models it was found that competition exertedby the initial predominant allele causes counterintuitive effects on evolutionary rescue (Uecker et al., 2014;Alexander and Bonhoeffer, 2012). For example, while rescue via de novo mutations is always more likely tooccur when the initial population declines slowly, rescue via standing genetic variation is more likely whenthe initial population crashes rapidly, eliminating predominant alleles that are competing with rescue alleles.As a consequence, the total probability of rescue may be related in a non-monotonic fashion to the rate ofpopulation decline, with the probability of rescue being high for slow and fast population decline rates butlower at intermediate decline rates (Uecker et al., 2014). Rescue from standing genetic variation with rapiddeclines in the population size has some support from empirical studies, where, for example strong selectionresults in the competitive release of drug-resistant bacteria from drug-susceptible competitors (Read et al.,2011; Huijben et al., 2013; Pena-Miller et al., 2013). Overall, these results highlight that incorporatingintraspecific competition can change the importance of standing genetic variation and de novo mutation inrescuing a population but whether these dynamics hold when the rescue traits and resources are coupledremains unknown.Stabilizing	SelectionMutationResource	SelectionMutationDiversifying	selection ?Mutation Selection G-matrixABFigure 2.1: The shape of the G-matrix resulting from a balance of selection and mutation. (A) Impact ofstabilising selection and mutation on the shape of the G-matrix (from Arnold et al. (2008)). (B) impactof resource and mutation is predicted to be more similar to the shape of the resource distribution due todiversifying selection.92.2. ModelHere, we develop a quantitative model for two traits that are adapting to an abruptly shifted resourcewhile explicitly incorporating intraspecific competition for that resource in a manner that depends on thetrait(s). Henceforth we refer to this model as the “ecological model". For example, stickleback may evolvelong and numerous or short and few gill rakers in response to a shift in prey size (Berner et al., 2010) orfinches may evolve longer and wider beaks in response to a shift in seed size (Grant and Grant, 2006). Weinvestigate how rescue depends on the direction of the resource peak shift with respect to the trait axes. Weuse individual-based simulations to track evolutionary change at ten unlinked loci, and we determine theprobability of rescue given only de novo mutation, only standing genetic variation, or both. To highlightthe impact of competition we compare our results to a similar model that does not incorporate competition(referred to as the “genetic model"). We find that traits affecting resource competition exhibit standinggenetic variation that is less coupled to the major axis of mutational variation. As a result, for traits impactingresource competition; de novo mutations allow rescue in different directions from those rescues allowed bystanding genetic variation. We conclude that, when it comes to evolutionary rescue, there is no longer asingle “genetic line of least resistance” but rather standing genetic variation and de novo mutation providepossibilities for rescue in different directions.2.2 Model2.2.1 Ecological ModelWe examine how a population with intraspecific competition adapts to an abrupt change in the environment.Our model depicting individuals of the same species competing for resources is loosely based on Johansson(2008) and is an extension of the Lotka-Volterra competition model incorporating multiple traits and resourcedimensions. In the ecological model we assume that each individual in the population is characterised byone or more phenotypic traits that influence their ability to obtain resources in a uni- or multi-dimensionalenvironment, where d specifies the number of trait and environmental dimensions. For example, in a twodimensional trait and resource environment, one trait dimension might be beak width and another beak lengthwhere both traits together influence the size and shape of seeds a bird can consume. The resources availableto an individual of phenotype ~zi are modelled as a multivariate normal distribution in d dimensions anddescribed by the following equation:K(~zi) = Kme−12(~zi−~µK)TΣ−1(~zi−~µK)√(2pi)d|Σ| (2.1)HereKm allows the height of the resource distribution to be altered, ~µK is a d-dimensional column vector ofthe trait values that would be best matched to the resource maximum, the superscript T represents the trans-position of the matrix, and |Σ| is the determinant of the matrix describing the curvature and orientation ofthe resource landscape. For a two-dimensional resource distribution, Σi is the following variance-covariancematrix:102.2. ModelΣi =(σ2i1 ρiσi1σi2ρiσi1σi2 σ2i2)Here, i is set to k to represent the covariance matrix for the resource distribution. Alternatively i can be setto w for the fitness distribution or to ν for the mutational distribution (see below). In the case of the resourcedistribution, σ2k1 and σ2k2 represent the variance of the bivariate normal distribution and describe the width ofthe distribution in each dimension, with smaller values indicating a narrower distribution inducing strongerselection. ρk is the covariance and determines the degree to which the availability of the two resources iscorrelated and consequently the orientation of the resource distribution.The growth of a specific phenotype i is influenced by the resources that it can access, given its own traitvalues, and by the extent of competition for those resources exerted by each of the other individuals j in thepopulation. The strength of competition αij is influenced by the amount of overlap in their resource useaccording to a d dimensional normal distribution:αij = e− 12σ2a(~zi−~zj)2 (2.2)where σa represents the niche width and where ~zi and ~zj are the trait values of the focal individual andcompetitor, respectively. The maximum of the competition kernel is one when individuals of the samephenotype compete.Finally, the fitness of an individual is determined by the number of its offspring that survive, assumingthat the species is hermaphroditic. To reproduce, each hermaphrodite produces a fixed number of eggs, F,which are all fertilized by a single randomly chosen individual from the population acting as a sire (includingpossibly itself). Thus, variation in reproductive success is incorporated through differential siring success(random) as well as differential survival of the offspring, which is assumed to be related to their ability toaccess resources (considered next). After reproduction, all parents die (non-overlapping generations). Theprobability of survival of offspring i with phenotype ~zi is determined byW (~zi)/F whereW (~zi) is the fitnessfunction:W (~zi) = er(1−∑j αijnjK(~zi)) (2.3)where r is the intrinsic growth rate and nj is the number of individuals of phenotype j.To implement our model we conduct individual-based simulations based on the model by Johansson(2008) and Van Den Elzen et al. (2017) implemented in Mathematica 8.0.4. The key model parametersare summarised in Table 2.1. Each simulation tracks the evolutionary dynamics of twenty alleles at tenunlinked, diploid genetic loci subject to mutation. The life cycle of our population is similar to an annualplant, where the population undergoes mating, mutation, recombination during gamete production and theproduction of offspring. The offspring that survive competition then grow up, and the cycle starts overagain. Each offspring acquires one allele at each locus randomly from the mother and the other from thefather. Offspring then acquire a mutation in a single randomly chosen allele with probability ν. Mutationsare drawn from a multivariate normal distribution with mean equal to the parental allelic value and variance112.2. Modelequal to the vector of mutational effect sizes ~σ2ν . For more than one dimension, mutations have pleiotropiceffects and may or may not cause covariation between traits. In two dimensions the mutational correlationis ρν , and the mutational matrix M is given by Σν . The number of dimensions and other parameters canalter the initial population size, so we adjust the height of the resource distribution Km so that the initialpopulation size at mutation-selection balance equilibrates at 400 individuals (Table 2.1).We ran ten independent burn-ins by initiating a population of 400 wild-type individuals with phenotype~zi = {0,0} at the resource optimum ~µK = {0,0} at t = 0. At this resource optimum, the population was allowedto undergo mutation for 5000 generations, facilitating the establishment of standing genetic variation. Boththe population size and genetic variation had stabilised by the end of the burn-in (Supplementary Figure A.1).Following the burn-ins we determine the probability of rescue for environmental shifts (~µK) at incrementallylarger distances from {0,0}. For each burn-in, a single simulation was run for each new position of theoptimum (~µK) and whether the population was rescued or not was noted. The population was consideredrescued in either of two cases. Firstly, if a population survived for at least five generations and reached asize of 100 or more individuals, we considered it to have been rescued (based on estimates of the minimumviable population size from MacArthur and Wilson (1967); Lande (1993); Gomulkiewicz and Holt (1995)).A few populations survived for longer than five generations with population sizes of less than 100 anddid eventually go extinct due to stochastic fluctuations. These populations were not considered rescued.Secondly, if the population survived for 10 000 generations, we also consider the population to have beenrescued. The probability of rescue was calculated for each particular environmental position by determiningthe proportion out of ten runs (each started from a different burn-in).As discussed in the introduction, competition for resources could influence the genetic basis underlyingrescue of a population from an abrupt environmental change. Previous work has shown that a large, abruptchange in the environment might favour rescue from standing genetic variation because adaptive alleles maybe present in the initial population and quickly increase in frequency. In contrast, with a smaller changein the environment, de novo mutation might play a larger role because population size should decline moreslowly, allowing new mutations more time to arise (Uecker et al., 2014; Alexander and Bonhoeffer, 2012).We however predict the opposite, that rescue after a small change in the environment should arise fromstanding genetic variation because diversifying selection has pushed individuals to utilise resources awayfrom the optimum, while rescue after a large change in the environment arise from de novo mutation becauseresource selection may prevent individuals from occurring in regions where resources are not present. Thus,we investigated the importance of each type of rescue by running simulations that included standing geneticvariation only, de novo mutation only, and both. In order to test the impact of standing genetic variation onlywe turned off mutation after the burn-in (i.e. we set ~σν = {0,0}). To test the impact of de novo mutation onlywe generated a population of 400 individuals with the phenotype of ~zi = {0,0} and then shifted the resourceoptimum away from ~µK = {0,0} allowing mutation to occur until the population was either rescued or wentextinct.122.2. ModelTable 2.1: Default model parameters used in the individual-based simulations.Model Parameter Value DefinitionEcological Km 10 304 Maximum resource abundance set to obtain an ini-tial population size of approximately 400 (unlessspecified)Ecological ~σK 2 Standard deviation of the resource abundance distri-butionEcological ρk 0 Covariance of the resource abundance distributionEcological σα 1 Standard deviation of the competition functionBoth F 4 FecundityBoth r 1 Intrinsic growth rateBoth ν 0.02 Genome-wide mutation rateBoth ~σν (1, 4) Standard deviation of mutational effect size (unlessspecified)Both ρν 0 Covariance of the mutational effect sizesBoth ~µK ~0 Mean of the resource abundance distribution, ini-tially at the origin and then moved 1 - 25 units awayin each dimensionGenetic Wm 1.05 Height of selection surfaceGenetic ~σw 20 Standard deviation of the selection surfaceGenetic ρw 0 Covariance of the selection surface2.2.2 Genetic ModelOur genetic model is analogous to Gomulkiewicz and Houle (2009) multivariate trait model. In contrast tothe previous model the fitness of each individual does not depend on how many other individuals are present.Rather a focal individual’s fitness only depends on its trait value. We assume that each individual in thepopulation is characterised by one or more phenotypic traits that influence their survival in a uni- or multi-dimensional environment. To reproduce, each individual mates with a randomly chosen individual (includingpossibly itself) and a constant number of diploid offspring, F, are produced, again with free recombinationbetween each of ten loci. Then all parents die (non-overlapping generations). The probability of survival ofoffspring i with phenotype ~zi is determined by W (~zi)/F where W (~zi) is the fitness function given byW (~zi) = Wme− 12(~zi−~θ)TΣ−1(~zi−~θ) (2.4)Here, Wm is the fitness of an individual with an optimum trait value, ~zi is a vector of phenotypic values,(~θ) is a vector of the optimum trait values, and Σw is the variance-covariance matrix of the bivariate Gaussianfitness function distribution with σ2w1 and σ2w2 representing the variance and ρw the covariance of the fitnessfunction.Due to the form of the fitness function in this model being density independent while it is density depen-dent in the competition model the two models are not directly comparable. However, to make them as similaras possible we scaled the width of the genetic fitness function (σ2w1 and σ2w1) so that the probability of rescue132.3. Resultsfrom de novo mutations was similar as the optimum was shifted along each trait dimension. Furthermore,although rescue was determined in the same way as in the ecological model, due to greater variability inthe population size of the genetic model (some populations went extinct even after reaching a populationsize of 100 individuals), we set the rescue population size to 300. This number was determined through theexamination of many simulation runs.Simulation code is available on request.2.2.3 StatisticsWe calculated the average trajectory of evolutionary rescue in trait space from standing genetic variation,de novo mutation and both for both the genetic and ecological models (Figures 2.2 and 2.3). In Figures 2.2and 2.3 points with black dots indicate the position where 100 individuals (our definition of rescue for theecological model) can persist without having to undergo evolution to survive. Because these points requireno rescue, i.e. the population size at these points never drops below 100 if the resource or fitness optimum ismoved to that location, we exclude them from all statistical analyses. On the rest of the points we performed aleast squares regression to determine the slope of the line. If the variance of the resource or fitness distributionis equal in all directions the least squares slope is not altered by excluding these points however if the resourceor fitness distribution has unequal variances in each direction then the slope is a better representation of theaverage trajectory of evolution (e.g. Supplementary Results A.1 and Supplementary Figure A.2). To create95% confidence intervals around the least squares regression line we performed 1000 bootstrap resampleswith replacement on the points without black dots.To calculate the probability of rescue as a function of distance from the burn-in optimum (Figure 2.4) wefitted logistic regressions to the proportion of ten independent runs rescued at each distance. Likelihood ratiotests were performed to test whether the probability of rescue as a function of distance from the optimumdiffered as a result of rescue coming from standing genetic variation, de novo mutation or both for the numberof independent runs performed.2.3 ResultsAfter the first 5000 generations with no environmental change the population in both the genetic and eco-logical models were at mutation-selection balance (Supplementary Figure A.1). In the genetic model thedistribution of traits at the end of the burn-in (G-matrix) closely resembled the shape of the mutation matrix(Figure 2.2 A and D inset panels). By contrast, the distribution of traits (G-matrix) in the ecological modelmore closely resembled the resource selection surface (Figure 2.3 A and D inset panels).As expected, the probability of rescue declines with the distance that the resource optimum is shiftedfor both the genetic and ecological models. However, differences in the shape of the G-matrix influencedthe probability of rescue from standing genetic variation, de novo mutation or both when the optimum wasshifted away from {0,0}. More specifically in the genetic model, rescue via standing genetic variation, denovo mutation or both were all biased in the direction with the greatest mutational variance. This can be seenfrom comparing the slopes of the lines in Figure 2.2 panels D, E, and F, which are similar. For an example142.4. Discussionwith unequal variances of the selection surface see (e.g. Supplementary Results A.1 and SupplementaryFigure A.2 A, B, and C). In contrast, in the ecological model, standing genetic variation was more stronglyshaped by the resource distribution and thus evolutionary rescue from standing genetic variation was influ-enced by the shape of the resource landscape (Figure 2.3 A and D). By contrast, evolutionary rescue from denovo mutation was biased towards the direction with the greatest mutational variance (Figure 2.3 B and E).As a consequence when both standing genetic variation and de novo mutation contributed to evolutionaryrescue, rescue occurred across a broader range of directions with standing genetic variation allowing morerescue in the horizontal direction (slope of 1.57 in panel D) and de novo mutation in the vertical direction(slope of 3.20 in panel E). For an example with unequal variances of the resource distribution see (e.g.Supplementary Results A.1 and Supplementary Figure A.2 D, E, and F).In Figure 2.4, we show how the probability of rescue falls off as the optimum shifts in the horizontal(first column) or vertical (second column) directions, in the case of an asymmetrical mutation matrix. Inthe genetic model, the pattern of rescue is similar whether the optimum is shifted horizontally or vertically(Figure 2.4 A and B) with standing genetic variation contributing to rescue at distances further out than denovo mutation under most circumstances (but see Supplementary Figure A.3). By contrast, in the ecologicalmodel, the probability of rescue changes according to whether alleles are coming from standing genetic vari-ation or de novo mutation. If the mutational effect size is less than the width of the resource kernel (σν<σk)rescue from standing genetic variation allows populations to adapt to larger changes in the environment thanif adaptation were to come from de novo mutation alone (Figure 2.4 C (P<0.0001)). In contrast, if σν>σkthen de novo mutation plays a larger role allowing rescue to occur further out than if it came from standinggenetic variation (compare blue and red curves Figure 2.4 D (P<0.0001)). Indeed, in this case the probabilityof rescue is lower with standing genetic variation only than with de novo mutation alone. This is becausethere is exceedingly low resource availability at distances far from the optimum in the vertical direction andtherefore mutations that arise during the burn-in at these far off locations do not establish.2.4 DiscussionCompetition for resources can either enhance (Osmond and de Mazancourt, 2013) or hinder adaptation(de Mazancourt et al., 2008; Johansson, 2008; Jones, 2008; Osmond and de Mazancourt, 2013; Van Den Elzenet al., 2017) to rapid environmental change, depending on whether selection from competition works withor against selection from environmental change. But resource competition and other forms of negative fre-quency dependent selection are rarely included in models of evolutionary rescue. Furthermore, most modelsexamining an abrupt environmental change only consider one trait dimension and do not extend results tomultidimensional environments (e.g. Gomulkiewicz and Holt, 1995). Here we develop individual-based sim-ulations that incorporate intraspecific competition for resources in a manner that depends on the trait(s) toinvestigate the probability of rescue when the resource optimum is shifted abruptly away from the origin. Wealso use these simulations to investigate the effect of the source of genetic variation (standing genetic vari-ation versus de novo mutation). We find that competition has profound effects on the importance of rescuefrom standing genetic variation versus de novo mutation. In particular, competition causes populations to152.4. Discussion!"1	=	4!"2 =	4!w1 =	20!w2 =	20!"1	=	1!"2 =	4!w1 =	20!w2 =	20Standing	genetic	variation De	novomutation BothA B CSlope	=	1.03 Slope	=	0.94 Slope	=	1.04Distance	to	new	environmental	resource	optimum	in	! " 2directionDistance	to	new	environmental	resource	optimum	in	!"1 directionD E FSlope	=	2.42 Slope	=	1.83 Slope	=	2.480 10 20010200 10 2001020-10 -5 5 10-10-55100 10 20010200 10 20010200 10 2001020-10 -5 5 10-10-55100 10 2001020Printed by Mathematica for Students1.000.20.40.60.8Proportion	rescued-10 -5 5 10-10-5510Trait	value	1Trait	value	2-10 -5 5 10-10-5510Trait	value	1Trait	value	2Figure 2.2: Probability of rescue for the genetic model from standing genetic variation, de novo mutation, orboth as a function of the new environmental optimum in two dimensions. Grey scale indicates the number ofrescue events out of ten independent simulations. Columns show the results from standing genetic variationonly (A and D), de novo mutation only (B and E), and both (C and F). The squares with dots indicate theareas where fitness is greater than one even for individuals of phenotype {0,0} and thus persistence does notdepend on evolutionary rescue. The inset figures in panels A and D are an example of the results obtainedfrom one burn-in after 5000 generations. The blue ellipse represents the area within which 95% of the popu-lation is present (illustrating the G-matrix) with each blue dot showing the trait combination of an individual;the red ellipse indicates the area within which 95% of mutations occur (M-matrix); the black ellipse indi-cates where the fitness of an individual in the population of phenotype {0,0} equals one, representing theadaptive landscape. The standard deviation of the mutational effect size (~σν) and the standard deviation ofthe selection surface (~σw) are indicated in the left most column of the figure, all other parameters were set totheir default values (Table 2.1).162.4. DiscussionStanding	genetic	variation De	novomutationA BD EBothCF!"1	=	4!"2 =	4!k1 =	2!k2 =	2!"1	=	1!"2 =	4!k1 =	2!k2 =	2Slope	=	1.04Slope	=	1.57Slope	=	1.01 Slope	=	0.99Slope	=	3.20 Slope	=	1.94Distance	to	new	environmental	resource	optimum	in	! " 2directionDistance	to	new	environmental	resource	optimum	in	!"1 direction0 10 2001020-10 -5 5 10-10-55100 10 20010200 10 20010200 10 20010200 10 20010200 10 2001020-10 -5 5 10-10-5510-10 -5 5 10-10-5510Trait	value	1Trait	value	2-10 -5 5 10-10-5510Trait	value	1Trait	value	2Printed by Mathematica for Students1.000.20.40.60.8Proportion	rescuedFigure 2.3: Probability of rescue for the ecological model from standing genetic variation, de novo mutation,or both as a function of the new environmental optimum in two dimensions. Grey scale indicates the numberof rescue events out of ten independent simulations. Columns show the results from standing genetic varia-tion only (A and D), de novo mutation only (B and E) and both (C and F). The squares with dots indicate theareas where per capitata fitness is one for an individual in a population with 100 individuals of phenotype{0,0}. Other details same as in figure 2.2 ).172.4. Discussion!"1 =	1 !"2 =	4Probability	of	rescueDistance	of	fitness	peak	away	from	burnin optimumA BC DDe	novo	mutationStanding	genetic	variationGenetic	modelEcological	model5 10 15 20 250.00.20.40.60.81.05 10 15 20 250.00.20.40.60.81.05 10 15 20 250.00.20.40.60.81.00 5 10 15 20 250.20.40.60.81.0Figure 2.4: Probability of rescue from standing genetic variation and de novo mutation in the genetic andecological models. Points represent the proportion of populations rescued over 10 independent replicatesimulation runs, solid lines are the logistic model fits, and shaded areas are 95% confidence intervals for thelogistic model. All parameters were set to their default values (Table 3.1).evolve to exploit the full distribution of resources available to them. As a consequence if the mutational ef-fect size is small (smaller than the width of the resource distribution), competition diversifies the populationand rescue is more likely to originate from standing genetic variation. In contrast, if the mutational effectsize is larger than the width of the resource distribution, competition becomes more of a stabilizing selectiveforce, and rescue is more likely to result from de novo mutation. Overall, we find that evolution is sometimesshaped by standing genetic variation and sometimes by de novo mutation, the importance of which dependson the distribution of resources, on the composition of the initial population, and mutational effect size.Genetic sources of rescue have been examined with and without density dependent growth to abruptlyshifted and gradually moving environmental optima (e.g. Hermisson and Pennings, 2005; Matuszewski et al.,2015; Orr and Unckless, 2008, 2014; Uecker et al., 2014). In general, models that did not include densitydependence found that adaptation from standing genetic variation was more likely when mutations wereneutral or only slightly deleterious in the previous environment (Orr and Unckless, 2008) and/or the allelescontributing to survival were of small effect. Although not studied with respect to the magnitude of shiftin the resource optimum, it seems likely that a small shift in the optimum should result in rescue by allelesof small effect and therefore by standing genetic variation, while large shifts in the optimum should requirealleles of large effect for rescue and therefore should result from de novo mutation. In contrast, with densitydependence, adaptation from standing genetic variation was found to be more likely when the populationdeclined rapidly (as when there is a large shift in the optimum) while rescue from de novo mutation was morelikely with slower rates of population decline (analogous to a smaller shift in the optimum). Contrary to theseprevious studies, in our model with resource competition we found that the rate of resource deterioration hadlittle impact on the source of genetic variation for rescue, rather the relative width of the resource distributionand mutation matrix were more important. We found standing genetic variation to rescue the population but182.4. Discussiononly when allelic effect sizes were small. If allelic effect sizes were large, rescue was more likely to comefrom de novo mutation than from standing genetic variation because competition for resources limited thewidth of the G-matrix.Empirically, the importance of new mutation in rapid adaptation to abiotic change has been noted inseveral systems. For example, Mwangi et al. (2007) observed the sequential appearance of point mutations inStaphylococcus aureus during drug therapy that increased drug resistance and ultimately led to the failure ofthe treatment. However, other empirical examples highlight the importance of both standing genetic variationand de novo mutation. As an example, the rapid response of highly inbred maize populations selected forearly and late flowering was attributed to the combined effect of standing genetic variation and de novomutations (Durand et al., 2010). Other studies have noted the importance of standing genetic variation only.For example, Gatenby (2009) noted the presence of drug resistant tumor cells prior to treatment. Thus,empirical examples indicate that in some circumstances rescue comes from standing genetic variation whilein other situations its comes from de novo mutation. Our simulation results provide an explanation for whythe source of genetic variation underlying evolutionary rescue may vary from case to case. In particularl ifthe key traits influence the acquisition of resources then the relative widths of the mutational matrix versusthe resource distributions will influence whether standing genetic variation or de novo mutation is moreimportant. Importantly though, if selection is not acting on traits related to resource acquisition then resultsmay mimic those of previous studies more closely (Alexander and Bonhoeffer, 2012; Matuszewski et al.,2015; Uecker et al., 2014, e.g.).The shape of the G-matrix is often thought to influence the trajectory of evolution in natural populations.Some empirical studies support this showing that evolution may be biased in the direction of the major axisof the G-matrix, as given by the eigenvector associated with the leading eigenvalue (gmax) (Schluter, 1996;Begin and Roff, 2004). However, various other empirical studies do not support that prediction that gmaxbiases the direction of evolutionary change (Berner et al., 2008, 2010; Eroukhmanoff and Svensson, 2011).For example, Berner et al. (2010) found that foraging traits of threespine stickleback inhabiting lakes andstreams were not related to the pattern of ancestral trait variance-covariance (considered to be the marinestickleback). Rather the trait covariance of lake and stream stickleback reflected selective pressures imposedby differences in prey resources, and population differences were driven by the strength of resource-mediateddisruptive selection. Our simulation results provide two different potential explanations for why the marinecovariance matrix and the lake and stream covariance matrices do not align. First, if the lacustrine sticklebackare still adapting and the resource distribution in the marine environment does not align with the mutationmatrix then the G-matrix of the lacustrine stickleback may reflect the recent accumulation of mutations ratherthan the resource environment. Alternatively, if the resource landscapes of the marine and lake and streamenvironments are very different then differences in diversifying selection may have reshaped the G-matricesof these populations.In conclusion, we found that diversifying selection from competition can play a major role in shapingthe G-matrix, which evolves to reflect the availability of resources rather than the mutation-selection balanceexpected under stabilising selection. This in turn means that standing genetic variation is not always a goodpredictor of whether evolutionary rescue is possible. Rather, de novo mutation better explains rescue in192.4. Discussiondirections that mutations explore but where few resources previously existed. Overall we conclude that,when it comes to evolutionary rescue, the “genetic lines of least resistance” are sometimes shaped by thestanding genetic variation but sometimes by the distribution of de novo mutations in populations competingfor limiting resources.20Chapter 3Asymmetric competition impactsevolutionary rescue in a changingenvironment 13.1 IntroductionFor species with limited dispersal abilities, phenotypic plasticity may allow survival in a changing environ-ment over the short term but persistence ultimately depends on evolutionary adaptation. Most studies haveinvestigated adaptation to abiotic change in isolation of other species (Huey and Kingsolver, 1993; Lenskiand Bennett, 1993; Partridge et al., 1995; van Doorslaer et al., 2009). Yet the evolutionary potential andtrajectory that a species follows can be strongly shaped by other species within the community. For example,interspecific competition can drive changes in population size (Gause, 1934) and the strength and directionof selection (Grant and Grant, 2006; Kleynhans et al., 2016; Lawrence et al., 2012). Consequently, we ex-pect interspecific competition to be an important determinant of whether and how species adapt to abioticchange. In this paper, we explore how competition for resources – particularly symmetric versus asymmet-ric competition or resources – determines the persistence and evolution of species competing in a changingenvironment.Previous work on the subject has found that resource competition can either increase or decrease thedegree of adaptation to abiotic change (de Mazancourt et al., 2008; Johansson, 2008; Jones, 2008; Price andKirkpatrick, 2009; Norberg et al., 2012; Osmond and de Mazancourt, 2013). On one hand, competition canprevent adaptation if it causes competitive exclusion (de Mazancourt et al., 2008; Johansson, 2008; Jones,2008; Price and Kirkpatrick, 2009). Theoretical work has shown that higher species diversity increases thechance that one or more species may be pre-adapted to a new environment, reducing the ecological oppor-tunity for an evolutionary response in the remainder of the community (de Mazancourt et al., 2008; Norberget al., 2012). Furthermore, pre-adapted species maintain larger population sizes for longer, facilitating theiradaptation, while maladapted species may fail to track a temporally shifting resource distribution because ofdeclining population sizes (Johansson, 2008). On the other hand, competition can exert an additional selec-tive pressure, facilitating adaptation when the abiotic and biotic selection pressures are concordant (Jones,2008). However, this will only assist adaptation when the increased selection pressure is enough to over-come the negative effect of a reduced population size due to competition (Osmond and de Mazancourt,1A version of this chapter has been published as Van Den Elzen C. L., Kleynhans E. J., & Otto S. P. (2017) Asymmetriccompetition impacts evolutionary rescue in a changing environment. Proc. R. Soc. B 284: 20170374. A supporting Mathematic fileis available at Dryad (http://dx.doi. org/10.5061/dryad.72k67)213.1. Introduction2013). Furthermore, the nature of competition may itself evolve, and the impact of a changing environmentcan be muted if species that initially become competitively inferior in a new environment then evolve tobe fiercer competitors (Northfield and Ives, 2013). More likely, however, declines in population size of thecompetitively inferior species may cause it to evolve to be a weaker competitor, exacerbating the impact ofenvironmental change on this species (Northfield and Ives, 2013).In this paper, we explore a model of competition where two species utilize a common resource whoseabundance distribution shifts through time. Previous theoretical work on the effect of competition on adapta-tion to environmental change had assumed that resource availability (Johansson, 2008; Jones, 2008; Osmondand de Mazancourt, 2013) and strength of competition (i.e. competition coefficients) (Johansson, 2008) fol-low symmetrical distributions. These symmetry assumptions cause the initial equilibrium population sizes tobe equal, with two competing species possessing trait values equidistant but on opposite sides of the resourcepeak. As the resource distribution shifts (environment changes) in time, the species whose trait value lies inthe direction of the changing resource (hereafter referred to as the leading species) is more likely to persistbecause its trait is pre-adapted to the changing environment and consequently its population size remainshigher for longer. The species whose trait value lies behind the direction of the changing resource (hereafterreferred to as the lagging species) must persist in the tail of the resource distribution and thus often declinesin abundance. Ultimately, lagging species are more likely to go extinct, even with environments that arechanging slowly enough to allow one species to persist (Johansson, 2008).Asymmetries are rife in nature. Resource distributions commonly exhibit skew (Ashmole, 1968; Boagand Grant, 1984; Brown and Lieberman, 1973; Fenchel and Kofoed, 1976; Kleynhans et al., 2011). Suchasymmetries can reflect age or stage structure of prey or plant species or can emerge from differential con-tributions from a variety of resource species (e.g., producing a variety of seed sizes from different plants).The competitive impact of individuals on one another is also often asymmetric in nature (Bonin et al., 2015;Connell, 1983; Kaplan and Denno, 2007; Lawton and Hassell, 1981; Weiner, 1990) and can arise as a resultof size differences between species or due to “prior-residence effects”. For example, in plant communities,taller individuals shade smaller individuals gaining a disproportionate share of light (Weiner, 1990; Falsterand Westoby, 2003; Schwinning and Weiner, 1998), and in animals larger males often win more contestsgaining access to a larger proportion of females (Clutton-Brock et al., 1979), or earlier arrivers have a com-petitive advantage over resources (Kokko et al., 2006). Asymmetries in either resources or competitioncoefficients generally alter the population sizes of species competing for resources so that the species withthe trait value closest to the resource peak or the species that is more competitively dominant has a largerpopulation size. With different population sizes, species will have different levels of standing genetic varia-tion, will incur different numbers of novel mutation, and will be subject to different degrees of demographicstochasticity, all of which could alter the relative probability of extinction of each species when facing achanging environment (Bell and Gonzalez, 2009; Gomulkiewicz and Holt, 1995). Allowing for differentinitial population sizes is also more realistic, as species competing for resources typically do not have equiv-alent population sizes (e.g., mammalian carnivores - Jones and Barmuta, 1998; mammalian herbivores -Kleynhans et al., 2011; or fish - Persson, 1983). Thus, we explore whether asymmetric forms of competitionfundamentally alter how communities of species respond to a changing environment. In theoretical studies,223.2. Modelasymmetries in resources and in competition coefficients have previously been incorporated into models oftaxon cycles (Rummel and Roughgarden, 1983; Matsuda and Abrams, 1994) and disruptive selection (Kisdi,1999; Doebeli and Dieckmann, 2000; Baptestini et al., 2009). However, they have not, to our knowledge,been used in the context of adaptation to a changing environment. Building on the methods of Johansson(2008), we investigate how asymmetries in resource availability or competition coefficients, by generatingdifferences in population size and in traits related to resource consumption, impact the evolutionary dynamicsof each species and how they alter which species persist and which go extinct in a changing environment.In brief, we develop an individual-based model with a single continuous trait that governs competitiveinteractions and is subject to selection. The strength of competition is assumed to depend only on thesimilarity of individuals’ trait values, whether those individuals are from the same or different species. Thetrait may represent a resource preference (e.g., preferred prey size), a habitat preference (e.g., preferredtemperature), a developmental trait (e.g., flowering time), etc. Following a burn-in period, we allow theenvironment to change in a manner that shifts the distribution of available resources. For example, oceanacidification may alter the developmental rate and/or degree of calcification for marine prey (e.g., (Doneyet al., 2009)), rising global temperatures may shift the availability of sites at a given mean temperature(Burrows et al., 2014), or delays in killing frosts at the end of the growing season may shift flowering time(Dunnell and Travers, 2011). As a consequence of this shift in available resources, the trait values of eachspecies experience selection to track the changing environment. We model the evolutionary response bytracking alleles at ten unlinked, diploid genetic loci subject to mutation following the method of Johansson(2008). We assume that individuals move freely across their range, at least when seeking a mate, so thatmating is random within a species but absent between species.We find that asymmetrical resource distributions or competition coefficients can facilitate coexistencewhen species’ lagging behind the changing resource distribution have a larger population size and can evencause extinction of the pre-adapted leading species before the lagging species. Surprisingly, we also findcases where traits evolve in the direction opposite to the values favoured in the changing environment. Thisoccurs because a shifting environment relaxes competition exerted by the lagging species on the leadingspecies, particularly when the leading species has a smaller initial population size. This creates a “vacuumof competitive release” that causes traits to evolve in the opposite direction of environmental change. Inter-estingly, this vacuum means that species that appear to be failing to adapt in a changing environment mightactually be competitive leaders that are simply taking advantage of steeper declines in lagging competitors.As a consequence, the species exhibiting the slowest rate of trait evolution is not always the species mostlikely to go extinct in a changing environment.3.2 ModelOur model is an extension of Johansson (2008) that explored a Lotka-Volterra based model of competition forresources in a changing environment. We follow the methods of Johansson (2008) closely with the exceptionof adding asymmetry to the resource distribution and competition coefficients. We find that results are similarfor both forms of asymmetry, and thus we present the findings for asymmetric resources in the body of the233.2. Modelpaper and results for asymmetric competition coefficients in the supplementary information. Simulationswere written in Mathematica 9.0, and the key model parameters are summarized in Table 3.1. The modeltracked diploid hermaphroditic individuals, with non-overlapping generations. During reproduction, eachindividual was chosen as a mother, a mate of the same species identified (including potentially the motherherself), and a constant number of offspring, F , was produced with that mate. The probability that anoffspring of genotype i survives was set to W (ui)/F , where:W (ui) = 1 + rk (1−∑k∑j αijnjkK(ui)), (3.1)which depends on the offspring phenotype (ui), the intrinsic growth rate of species k (rk), the resourceabundance at ui (K(ui)), the competition exerted by each of the j types of the k species on individual i (αij ,including itself), and the number of individuals of phenotype uj of species k (njk). If a mother’s offspringall had the same genotype i (temporarily ignoring recombination and mutation), W (ui) would represent themother’s expected number of surviving offspring, and the number of individuals of phenotype ui in speciesk would be expected to change over time according to:nik(t+ 1) = W (ui) nik(t), (3.2)The need to calculateW (ui) over the entire distribution of individuals each generation greatly slows the sim-ulations. For speed, Johansson (2008) linearly interpolatedW (ui) using 100 values of ui in each generation.We instead round individual trait values to 4 significant figures, and calculate W (ui) from these roundedvalues. Therefore, when there is more trait diversity, the fitness function is calculated based on more points.We believe this helps capture the nonlinearities in the fitness function when many phenotypes are present.Assuming rare mutations and a population initially fixed for phenotype ui, the expected rate of evolutionof species k would be:duidt= ν σ2ν nˆk∂W (u′)∂u′∣∣∣∣u′=ui, (3.3)(Dieckmann and Law, 1996), where ν is the mutation rate, σ2ν is the variance of mutation effect sizes,nˆk is the equilibrium population size of species k, and∂W (u′)∂u′∣∣∣u′=uiis the fitness gradient evaluated at thephenotype ui. Equations (3.2) and (3.3) were used to determine equilibrium configurations that were bothecologically and evolutionarily stable (see supplementary Mathematica package).Following Johansson (2008), an individual’s phenotype was based on the additive effects of twenty allelesat ten diploid loci. Offspring acquired a mutation with probability ν per diploid genome, with the mutationaleffect drawn from a normal distribution with a mean equal to the parental allelic value and variance σ2ν .When calculating fitness, we rounded trait values to the nearest one-thousandth.The resource distribution, K(ui), represents the resources available to individuals of phenotype ui. Weassumed that K(ui) has a single peak (Doebeli, 2011), and we introduced asymmetry into this resource243.2. ModelTable 3.1: Model parameters. Default values are given, along with alternate values considered as described in the text(square brackets).Parameter Value DefinitionF 4 Fecundityrk 1 Intrinsic growth rateν 0.02 [0.2] Genome-wide mutation rateσν 0.0015 [0.015] Standard deviation of mutational effect sizeKm Variable Maximum resource abundanceσK 1 Standard deviation of the resource abundance distributionµK 0 [+Vmt] Mean of the resource abundance distributionκ 0 [±4] Degree of asymmetry in the resource abundance distributionσα 0.85 Breadth of the competition functionβ 0 [±0.6] Degree of asymmetry in competition functionVm 0.00005 (slow) Speed of environmental change0.0005 (fast)**Additional intermediate values (Vm = 0.0001 and 0.00015) were used in search of leading species extinction with increasedmutation rate.distribution by using a skew normal distribution (Lovric, 2011):K(ui) = Kme−(ui−ξ)22ω2√2piω2erfc(−κ (ui − ξ)√2ω2). (3.4)Here, κmodifies the skew (asymmetry) of the distribution (Figure 3.1A); with κ = 0, we regain the Gaussiandistribution of Johansson (2008) and others (Rummel and Roughgarden, 1983; Roughgarden, 1979; Drosseland McKane, 1999; Dieckmann and Doebeli, 1999). Km scales the height of the resource abundance, ξdetermines the position of the distribution along the resource axis (corresponding to Um in (Johansson,2008)), ω controls the breadth of the distribution, and erfc() represents the complimentary error function.In order to hold the mean and variance of the resource distribution constant while varying the skew, weset the first and second moments of the skew-normal distribution equal to those for a Gaussian distributionwith mean µK and variance σ2K , yielding:ξ = µK − κ σK√(1 + κ2)(pi/2− 1) + 1 (3.5)for the location of the resource distribution andω = σK1√1− 2κ2pi(1+κ2)(3.6)for the breadth of the distribution.253.2. Model–2 –1A100020003000400050001000200030004000500002 –1 1 20B100020003000400050000C0 1 2Phenotype, ui Phenotype, uiK(ui )K(ui )Figure 3.1: Symmetric and asymmetric resources and their impact on population size and trait value. (A)The resource distribution (Equation (3.4)) available to individuals of phenotype ui is shown for κ = 0 (solid:symmetrical), κ = 4 (dashed: right-skewed), and κ = −4 (dotted: left-skewed). Right-hand panels show theinitial ESS for each species (thin red and blue vertical lines), the post-burn-in species distributions (in 0.01bins, narrow red and blue histograms, doubled in height for clarity), and the competition exerted by thosetwo populations on individuals of phenotype ui (red and blue curves, respectively) for (B) the symmetricalresource distribution case (κ = β = 0), (C) right-skewed resource distribution case (κ = 4, β = 0). Note in(C) that the species with a negative phenotype has a larger population size (red histogram) than the specieswith a positive phenotype (blue histogram).Alternatively, we introduced asymmetrical competition coefficients using the function:αij =( 11 +(ui−uj+β)22σ2α)(1 +β22σ2α)(3.7)The parameter β describes the asymmetry of competition coefficients, and the width of the competition ker-nel is controlled by σα. When β = 0, we regain the symmetrical case considered by Johansson (Johansson,2008), with αij = αji. When β > 0, the distribution is right-skewed, implying that individuals exert thestrongest competitive effects on individuals of a slightly smaller phenotype (Figure B.1A). The last paren-thetical term in Equation (3.7) scales competition so that the effect of individuals on others of the samephenotype is, by definition, one (i.e., αii = 1).The evolutionary stability of communities with one or more than one species depends on the ratio of thecurvature at the modes of the competition and resource functions (Doebeli, 2011; Dieckmann and Doebeli,1999). For symmetric distributions, there is only one evolutionarily stable species when σασK > 1, but multi-species communities can persist for σασK < 1 (Johansson, 2008; Doebeli and Dieckmann, 2000). For thisreason, Johansson (2008) chose to use different values of σασK for his simulations with one species (σασK=1.5) versus two species (typically using 0.85). He also showed, however, that increasing σασK reduced the263.2. Modelmaximum velocity that could be sustained in the two species case (see Table B2 in (Johansson, 2008)).To avoid confounding the potential effects of σασK and the number of species, we chose to fixσασKat 0.85.Because neither assortative mating nor asexuality were allowed to evolve within species, branching was notpossible (Doebeli and Dieckmann, 2000; Doebeli, 2011), and consequently this choice allowed both one-species and two-species communities to persist at equilibrium. Interestingly, for the asymmetry parametersconsidered (κ = ±4 or β = ±0.6, Figs. 3.1, B.1), two-species and one-species communities are not onlyboth stable equilibria, but they also both represent evolutionary stable strategies, at least locally. That is,branching by small mutational steps would not be expected from either equilibrium, even if the populationwere asexual (see supplementary Mathematica package). The parameter values governing asymmetry werechosen to generate similar ESS trait values and population sizes, regardless of whether asymmetries wereintroduced into the resource distribution (κ = ±4, Figure 3.1) or the competition coefficients (β = ±0.6,Figure B.1). Specifically, these parameter choices led to trait values near 0.5 for the leading species and -0.5for the lagging species, where the larger of the two populations was 6-7 times the size of the smaller (seesupplementary Mathematica package).Simulations were initiated from communities of one or two species, each of which was initially fixedphenotypically according to the evolutionary stable strategy (ESS) obtained numerically (or in the symmetriccase with one species at the singular point; Table B.1). With one species, the ESS phenotype was centeredat the peak – not the mean – of the resource distribution when the resource distribution was asymmetric(κ 6= 0). Asymmetrical competition coefficients (β 6= 0), however, caused a shift in the ESS away fromthe peak of the resource distribution and towards phenotypes experiencing weaker competition (i.e., towardthe right with right-skewed competition, β > 0). When two species were present and the distributions wereperfectly symmetrical, the ESS consisted of two species of equal population size, lying equidistant but onopposite sides of the resource peak. With asymmetrical distributions (either κ 6= 0 or β 6= 0), however,the phenotypes of the two distributions were different distances away from the mode, and the equilibriumpopulation sizes were unequal (Table B.1, and see supplementary Mathematica package).In all cases, the height of the resource distribution, Km, was chosen so that the average number of in-dividuals per species was initially expected to equal 2500 at equilibrium, regardless of the parameter values(Table B.1). While initial population sizes were equal in the symmetrical case, the two species varied sub-stantially in population size with asymmetrical parameter choices (Table B.1): 7.0 fold with asymmetricalresources (κ± 4) and 6.6 fold with asymmetrical competition coefficients (β ± 0.6).Starting from an initial resource mean of µK = 0, we then allowed the communities to acquire geneticvariation and stabilize for 5000 generations (“burn-in” from t = −5000 to 0). At time t = 0, we simulateda changing environment by shifting the resource distribution linearly upwards at speed Vm over time, sothat µ = Vmt. The speed was chosen to be either slow enough that a single species could persist on itsown for 15000 generations (Vm = 0.00005) or fast enough that a single species went extinct by that time(Vm = 0.0005). We then examined the persistence of communities consisting of two species, relative to theone species case.273.3. ResultsTable 3.2: Extinction time in generations. Populations were censused every 100 generations for extinction orpersistence. If no replicate went extinct by 15000 generations, an “∗” is reported. In all other cases, the meanextinction time is recorded, with SEM in parentheses. All data has been deposited in Dryad.Slow environmental change Fast environmental changeOne species Two species One species Two speciesκ β Leading Lagging Leading Lagging0 0 ∗ ∗ 7880 (404) 7480 (97) 8720 (37) 820 (20)4 0 ∗ ∗ 14840 (160)† 4100 (55) 6480 (37) 1380 (20)-4 0 ∗ ∗ 7740 (385) 10200 (95) 11100 (45) 520 (20)0 0.6 ∗ ∗ 1640 (51) 9440 (68) 10260 (81) 160 (24)0 -0.6 ∗ ∗ 13640 (316) 6060 (75) 8660 (40) 1400 (0)† One replicate went extinct at 14200 generations and the remaining four persisted to 15000 generations; these simula-tions were extended until all replicates went extinct, which occurred on average after 15300 (329) generations.3.3 ResultsFigure 3.2 illustrates the effect of an asymmetric resource distribution on the persistence of a two-speciescommunity. When the environment changed slowly (left panels), the leading species (higher data points)persisted the full 15000 generations, as observed in the single species communities (Figure B.2; Table 3.2).Persistence was observed whether the leading species was initially larger (red indicates species with thelarger initial population size; panel C) or smaller (blue; panel E) than the lagging species. By contrast,the lagging species (lower data points) typically went extinct before 15000 generations, unless resourceasymmetries allowed the lagging species to have access to substantially more resources (panel E); in thiscase, with a larger initial population size, the lagging species better tracked the changing environment, andmost replicates (4 out of 5) persisted throughout the 15000 generations of environmental change.When the environment changed rapidly (right-hand side of Figure 3.2), however, all species went extinctby generation 15000. Nevertheless, species with initially larger population sizes persisted for longer. Fur-thermore, for these parameter values, the leading species always persisted longer than the lagging species,regardless of whether it was smaller or larger initially. The leading species also persisted longer than a singlespecies community (see asterisks on x-axis and Figure B.2).The leading species from the two-species communities had two separate advantages that allowed it topersist longer than single species communities when the environment changed rapidly: a “head-start” in thetrait mean and a larger population size. The head-start was due to the “competitive push” by the laggingspecies (c.f., (Jones, 2008)), causing the leading species to have a higher initial trait mean and hence to bebetter pre-adapted to the environmental change. The larger population size was due to our assumption that,initially, the average population size for the two-species communities was equal to that of the single speciescommunity (2500 individuals). Consequently, once the lagging species went extinct, the leading species hadaccess to more resources, causing its population size to rise above 2500 at least transiently (inset panels inFigure 3.2). In the Supplemental Methods B.1, we describe additional simulations to disentangle the head-start effect from the population size effect. These simulations indicate that both play a role; the head start283.3. ResultsTrait ValueGeneration-0.50.00.51.00 5000 10000 15000-5000-0.50.00.51.00 5000 10000 15000-5000**(b)(a)(d)(c)(f)(e)-0.50.00.51.0*Lagging (smaller N)Leading (equal N)Lagging (equal N)Leading (larger N)Lagging (larger N)Leading (smaller N)Resource peakResource peakResource peakPop sizePop sizePop sizeSlow Environmental Change Fast Environmental ChangePop sizePop sizePop sizeFigure 3.2: The evolution of two-species communities competing for asymmetrical resources in a changingenvironment. The resource distribution was either normally distributed (panels A,B), left skewed (κ = −4,panels C,D), or right skewed (κ = 4, panels E,F), as illustrated by the grey distribution along the y-axis.In all panels, the mean resource was initially zero for 5000 generations, after which the distribution shiftedupwards slowly (Vm = 0.00005, left panels) or rapidly (Vm = 0.0005, right panels); the dashed line tracksthe peak (mode) of this resource distribution and also corresponds to the trajectory of the mean in panelsA,D, where the distribution is symmetric. The initial ESS trait dynamics are given by the arrows on theleft of panels A,C,D and the corresponding simulation trait dynamics are illustrated in the main plots (mean±1 SEM shown while more than one replicate persisted). The asterisks show the average time that a singlespecies community went extinct (only observed in a rapidly changing environment). Inset plots show thepopulation size dynamics for each species (inset plot x-axis: -5000 to 15000 generations, y-axis: 0 to 5000individuals), where red is the initially larger population and blue the initially smaller population. Remainingparameters were set to their default values (Table 1). All data has been deposited in Dryad.293.3. Resultswas most important in Figure 3.2 F, where the leading species was rare initially and the trait far ahead of thepeak resource, while the population size advantage was most important in Figure 3.2 D, where the leadingspecies initially had a large population size but a trait value close to the resource peak.Very similar patterns were also observed when competition coefficients, not resources, were asymmetric(Figure B.3). In particular, asymmetries helped maintain both species in a community over longer periods oftime when the lagging species had a larger population size (Figure B.3 E, F). The leading species in the two-species community also persisted longer than either the lagging species or the single species communities(Figure B.4). The fact that the source of the asymmetry did not greatly impact the results is not unexpected,given that we chose the parameters κ and β so that the initial abundances and trait values for the lagging andleading species were similar in the two cases.One of the more intriguing patterns seen with either form of asymmetry was evolutionary responsesin the opposite direction to the changing environment. This was repeatedly observed when the leadingspecies had the smaller initial population size (Figure 3.2 and B.3, panels E, F). As the resource peak movedaway from the lagging species, its population size began to decline, increasing the resources available at thelower end of the resource distribution for use by the leading species. The leading species then evolved tomonopolize the resource left behind by the declining lagging species, rather than evolving in the direction ofthe environmental change. In other words, the decline of the lagging species caused a vacuum of decreasedcompetition into which the leading species evolved. In all cases, we only observed this pattern while thepeak of the resource distribution remained between the mean phenotypes of the two species. Once the peakcrossed over both species’ mean trait, the direction of evolution for the leading species shifted to track thechange in the environment. Trait evolution in the opposite direction of the environmental change persistedfor the entire 15000 generations in the case of a slowly changing environment but was transient with a rapidlychanging environment, reversing soon after the lagging species went extinct. On the basis of these results,we emphasize that species appearing to fail in tracking a changing environment may, in fact, be gaining anecological foothold, evolving into the niches of species declining within the community.In none of the cases explored above, however, did the lagging species ever advance ahead of the leadingspecies. We speculated that this might have been due to the low amount of genetic variation for the traitunder the default parameter set based on Johansson (Johansson, 2008). We thus ran simulations with eitheran increased mutation rate (ν = 0.2 rather than 0.02) or mutational effect size (σν = 0.015 rather than0.0015) and an asymmetric resource distribution (κ = ±4). In these cases, genetic variation was sufficientto ensure that one of the two species always survived, even in a rapidly changing environment (Table B.2).The surviving species was always the leading species when it had the larger initial population size. Wheninstead the lagging species had the larger population size, however, the lagging species had access to moregenetic variation (both standing and de novo) and was capable of evolving faster than the leading species.Thus evolutionary advantage allowed the lagging species to persist, in some cases, while the leading specieswent extinct. The competitive exclusion of the leading species was observed in one of five original replicateswith a larger mutational effect size (σν = 0.015, see Figures 3.3, B.6). To better estimate the frequency ofleading species extinction, we ran fifteen additional replicate simulations (three from each original burn-in)and observed two more cases where the leading species went extinct while the lagging species persisted (for303.4. DiscussionTrait ValueGeneration0246A B0 5000 10000 15000-5000 0 5000 10000 15000-5000Figure 3.3: Higher adaptability can allow a lagging species to out-evolve a leading species in a rapidly chang-ing environment when mutational effects are larger. The parameters for these simulations were identical toFigure 3.2F except for σν = 0.015. Four out of five original replicates (see results section for additional sim-ulations) showed patterns similar to panel A, but in one replicate, the lagging species survived, driving theleading population to extinction (panel B). Observe that the two species are very similar in mean trait (mainplots) and population size (insets) at around 3000 generations, which is when the fate of the two speciesis determined stochastically by which species has the largest positive mutation. See Figure 3.2 for furtherdetails and Table 3.1 for parameter values. All data has been deposited in Dryad.a total of 3/20 replicates, unrelated to which of the five burn-in populations was used). For these parameters,the faster evolution of the lagging species brought its trait value towards that of the leading species; at thatpoint, population sizes of the two species converged, and which species ultimately prevailed was stochastic(recall that there are no other ecological differences between these species). By contrast, competitive exclu-sion of the leading species by the lagging species was not initially observed when the mutation rates werehigher (ν = 0.2 rather than 0.02; Figure B.5). To explore this case further, we ran 10 additional replicatesimulations (two per burn in), for each of two intermediate environmental change speeds (Vm = 0.0001 andVm = 0.00015) that were chosen to allow for greater trait convergence before extinction. While the traitvalues did become closer, the leading species always stayed ahead of the lagging species, despite its smallerinitial population size, and never went extinct before the lagging species. Finally, we ran five additional sim-ulations at the slow rate of environmental change (Vm = 0.00005) until one of the two species went extinct(after 38934 generations, on average). With such slow environmental change, the traits converged even moreclosely, and in one of five replicates, the leading species went extinct first (Table B.2). We thus concludethat asymmetries in resource competition make it possible for lagging species with a larger population sizeto out-evolve a pre-adapted species along a resource gradient, as long as there is sufficient genetic varianceto allow the trait values to converge given the rate of environmental change.3.4 DiscussionPredicting how species will adapt to abiotic change when in competition with other species remains a chal-lenge for biologists, given that competition can alter the population sizes of the component species, as well asthe strength and direction of selection (Grant and Grant, 2006; Lawrence et al., 2012). Our study contributes313.4. Discussionto understanding how competitive interactions between species might influence adaptation to abiotic changein several distinct ways.First, by exploring skewed resource distributions, we were able to manipulate the population sizes ofcompeting species. As expected, species with initially larger populations persisted for longer, likely becauselarger populations take longer to decline deterministically to zero (even when holding trait values constant),are better protected from demographic stochasticity, maintain more standing genetic variation, and incurmore beneficial mutations (Holt and Gomulkiewicz, 1997; Hufbauer et al., 2015). The advantage of a largerpopulation size was most pronounced when the species with the larger population size was also the leadingspecies (Figure 3.2 C, D). In comparison to previous models (Johansson, 2008; Jones, 2008), our workadds additional support to the finding that population size is an important determinant of persistence andadaptation to abiotic change in communities of competitors.An additional consequence of interspecific competition is that it drives character displacement. Withsymmetrically distributed resources in a constant environment, two symmetrically competing species evolvethe same distance away from, but on opposite sides of, the resource peak. In contrast, resource skew alterstrait values of coexisting species by displacing them from their symmetrical equilibrium position. For exam-ple, Slatkin (Slatkin, 1980) noted that with an asymmetrical resource distribution one species will occupy aposition near the centre of the resource distribution and the other will be displaced to the tail of the distri-bution. The species in the tail is prevented from adapting to utilize more abundant resources by competitionfrom the species occupying the centre of the distribution. We found very similar results and in additionfound that the competitive push exerted by the abundant species on the less abundant species can influenceadaptation to abiotic change. With a right skewed resource distribution (Figure 3.2 E, F), the leading species(blue) is pushed far in front of the peak of the resource distribution. This “head-start” allows it to persistfor longer than it would have in the absence of competition (see asterisk), even though its population sizewas initially lower. By restarting single-species simulations with the same head-start but the same popu-lation size as in a single species community, we confirmed that this head-start could greatly lengthen thepersistence time of the leading species (Supplementary Materials B.1). This preadaptation is a second wayin which competition can aid persistence to a changing environment, beyond the selective push that occurswhile the environment is changing if competition selects in the same direction as the environmental change(Osmond and de Mazancourt, 2013). Mellard et al. (2015) also highlighted the importance of a head-start infacilitating adaptation in a plant-herbivore system, suggesting that similar mechanisms may be at play acrosstrophic levels.We also repeatedly observed adaptation in the opposite direction, relative to the environmental change,when the leading species initially had a lower population size. This phenomenon, which we call the “vac-uum of competitive release”, occurred during the decline in population size of the lagging species in the faceof environmental change, which released resources at the trailing end of the resource distribution. Conse-quently, the leading species evolved to utilize available resources (in the opposite direction to the shiftingenvironment) and only later tracked the environment after the lagging species went extinct.Another contribution of our study is that we found that it is not universally true that the lagging specieswill go extinct before the leading species, in contrast to previous results (Johansson, 2008; Jones, 2008).323.4. DiscussionIn particular, when the lagging species initially had the larger population size and had substantial geneticvariation (Figure 3.3), the trait value of the lagging species could evolve rapidly towards that of the leadingspecies as the environment changed, occasionally allowing the leading species to go extinct and the laggingspecies to be rescued. The competitive exclusion of the leading species by the lagging species was observedboth when mutational effects were larger (raising σν from 0.0015 to 0.015; Figure B.6, D with Vm = 0.0005)and when the mutation rate was higher (Table B.2 with Vm = 0.00005). Given that the two species initiallydiffered by about one phenotypic unit in the trait, a mean mutational effect size (i.e., mean absolute deviation)of 0.0012 (√2σ2/pi where σ2 is the variance) implies that the species differ in this trait by about 836mutations of average effect, which is orders of magnitude larger than the average number of QTL estimatedfrom studies of recent species (Orr, 2001). Thus, increasing the mutational effect size to 0.012, an expecteddifference of 84 genes, is not unreasonable. Importantly, even if a species has little standing genetic variation(assessed at marker sites), access to sufficiently large effect mutations may still allow it to be rescued fromextinction.In general, we conclude that competition for resources makes it difficult for multi-species communitiesto persist in a changing environment. That said, competition also sets up the conditions whereby speciation(or spread of an invasive competitor) could recur (Johansson, 2008; Doebeli and Dieckmann, 2000), as seenin the branching-extinction cycles of (Kisdi et al., 2002). Given that one of the two species always wentextinct under rapid environmental change, we infer that it would be difficult for new or invading species topersist until the environment stabilizes.Although a few empirical examples exist that show adaptation in the opposite direction to that predictedby the changing environment (Lehikoinen et al., 2009, 2013; Kimball et al., 2010), finding clear empiricalexamples for the vacuum of competitive release is challenging, requiring that we track changes in popula-tion size and mean trait values as well as infer shifts in the trait optimum with environmental change. Onepromising approach is to examine phenological shifts due to climate change for evidence of a vacuum ofcompetitive release (assuming that population sizes do differ, because of asymmetric resources and/or asym-metric competition coefficients). For example, common buzzards (Buteo buteo) breeding at the northerndistribution limit in Finland have advanced their timing of breeding by 11 days over the past 30 years. Thisshift in breeding is correlated with a warming climate and is thought to benefit this species through facil-itating range expansion. Nevertheless, populations of common buzzards in northern Finland are declining(Lehikoinen et al., 2009). Goshawks (Accipiter gentilis), a possible competitor for nest sites, have insteadbeen marginally increasing in abundance, even though they have not advanced their timing of breeding to thesame extent (Lehikoinen et al., 2013). Although not the conclusion of these studies, it would be interesting toinvestigate whether goshawks are under less selective pressure to alter their breeding date partly because ofrelease from competition by buzzards. A second system in which this could be occurring is among tits. Greattits (Parus major), as well as other tit species such as blue tits (Cyanistes caeruleus) and willow tits (Poecilemontanus), rely on insect larvae to feed their young during the breeding season (Dhondt, 1977; Dhondt andAdriaensen, 1999; Pakanen et al., 2016). With warming spring temperatures, the peak larval abundance hasshifted earlier in the season. Great tits, which breed later in the season, are not, however, keeping up withthis resource shift, creating an increasing lag between breeding and food supply (Visser et al., 1998). Our333.4. Discussionwork suggests that blue tits and willow tits, which initiate breeding earlier, may be selected to shift to laterbreeding dates because of the release of resources from declining great tit populations, despite the advancein peak food supply with advancing spring temperatures.Laboratory experiments may be particularly suitable for demonstrating the vacuum of competitive re-lease. The classic Asterionella-Cyclotella system of algae competing for phosphorus and silicate (studied byTilman (1977)) could be used, for example, starting with a stable community where Asterionella is initiallyrare. Environmental change could then be induced by decreasing the phosphorus to silicate ratio. Given thatAsterionella is the better competitor under phosphate-limited conditions, this species would be predicted torise numerically over time. In terms of traits, however, its ability to use the limiting phosphorus resourcemay initially decline before improving, as the initially more common Cyclotella drop in numbers and releasecompetitive pressure on phosphorus. Similarly, shifts in the distribution of food size availability (algal size)could be used to explore the evolution of artificial communities of competing Daphnia clones. By increasingthe average size of supplied algae over time, the prey preference of larger clones (leading) could be trackedas smaller clones (lagging) decline in numbers, again looking for evidence of counter-selection between theenvironmental change and release of resources from competition.Overall, we find that accounting for asymmetries in resource distributions and competition coefficientshas substantial impacts on the fate of competing species in a changing environment, even leading to thepersistence of species whose traits are initially maladapted over those that are pre-adapted to a changingenvironment. Counterintuitively, when the population size was smaller for the leading species, we typicallyobserved a period of adaptation in the opposite direction to the environmental change. In these cases, thespecies that appears to be failing to track a changing environment is not necessarily the one most at risk ofextinction.34Chapter 4Adaptation to elevated CO2 in differentbiodiversity contexts 14.1 IntroductionSpecies do not evolve in isolation but within a community of interacting species. While some evidenceexists for the impact of predator-prey (Gould et al., 1991; Van Doorslaer et al., 2010; Liu et al., 2014) orhost-parasite interactions (Wolinska and King, 2009) on adaptive evolution, we lack experimental data onthe impact of competition on adaptation in natural systems. Laboratory and mesocosm studies have foundcontrasting results for how competition influences adaptation. Competition can inhibit adaptation, as foundin algal cultures (Chlamydomonas reinhardtii) evolving to elevated CO2 (Collins, 2011). Similarly, adaptivediversification to habitat heterogeneity in Pseudomonas fluorescens was prevented in the presence of inter-specific competitors, as these competitors exclude intraspecific variants (Bailey et al., 2013). Competitioncan also alter the nature of selection (Van Doorslaer et al., 2010; Lawrence et al., 2012; Lau et al., 2014). Forinstance, increased water temperature caused the zooplankton Daphnia magna and D. pulex to evolve fastergrowth in the absence of competition but to evolve a larger size at maturity in the presence of competitors andpredators (Van Doorslaer et al., 2010). Similarly, bacterial species adapted differently to a novel environmentwhen grown alone or with other bacteria species (Lawrence et al., 2012). In plants, the presence, composi-tion, and diversity of competing species shows tremendous spatial variation (Rosenzweig, 1995), can have amajor impact on individual performance (Turkington and Harper, 1979; Wedin and Tilman, 1993), and thusmight have an important influence on species’ adaptation to environmental change. Yet, how biotic com-munity context alters how species adapt to environmental change in natural field settings remains entirelyunknown.To address this knowledge gap, herein we report on an investigation of the impact of prairie grasslandcommunities on the evolutionary response to elevated CO2 over 14 years in the Biodiversity Carbon dioxideand Nitrogen experiment (BioCON) at the Cedar Creek Ecosystem Sciences Reserve (Minnesota, USA) (Re-ich et al., 2001a). To determine how the surrounding biological community influences a species’ evolutionin response to abiotic change we focused on very different community structures: monoculture versus highdiversity. By focusing on the presence or absence of interspecific competitors, we increased our power todetect the impacts of the surrounding species diversity on evolution. We tested four possible scenarios bywhich species diversity might affect the evolutionary responses of a focal species to abiotic environmental1A version of this chapter has been published as Kleynhans, E. J., Otto, S. P., Reich, P. B., & Vellend, M. (2016). Adaptation toelevated CO2 in different biodiversity contexts. Nature Communications, 7:12358354.1. Introductionchange:The first scenario is that species diversity has no effect on adaptation to abiotic change (Figure 4.1 a, b)If selective pressures exerted by changing abiotic conditions overwhelm those from the biotic community,species diversity should have no impact on local adaptation to abiotic change. Statistically, the responseto selection (fitness of plants evolved under elevated CO2 (eCO2) minus fitness of plants evolved underambient CO2 (aCO2) should be predicted only by the change in CO2 environment (∆CO2), regardless ofspecies diversity (Figure 4.1 a-b).The second scenario is that species diversity constrains adaptation to abiotic environmental change (Fig-ure 4.1 c-d) When grown with more species, local adaptation might be reduced because competition forspace and resources results in smaller effective population sizes of each competing species, reducing stand-ing genetic variation (Bocedi et al., 2013; Lanfear et al., 2014) and the rate at which new mutations arise(Johansson, 2008). In addition, with more species in a community one species may, by chance, possesstraits that pre-adapt it to the new niche(s) created by the changing environment (Johansson, 2008). Thispre-adapted species will increase in abundance, exerting more competition and resulting in a further declinein abundance, and therefore ability to adapt, in the other species (Bocedi et al., 2013; de Mazancourt et al.,2008; Norberg et al., 2012). Under this scenario, local adaptation to elevated CO2 should be more evident forplants that experienced selection in a species-poor community (Figure 4.1c) than in a species-rich commu-nity (Figure 4.1d), regardless of the diversity of the community into which the plants were transplanted (the“assay” community). Statistically, the response to selection should be predicted by a three-way interactionbetween the CO2 selection environment (CO2sel), the change in CO2 environment (∆CO2), and diversity ofthe selection environment (divsel).The third scenario is that species diversity promotes adaptation to abiotic environmental change (Figure4.1e, f) A more homogeneous environment caused by low species richness might select for and maintainfewer genotypes, reducing genetic diversity. Likewise, high species richness might increase environmentalheterogeneity, thereby maintaining greater genetic variation and adaptive potential (Turkington and Harper,1979; Vellend and Geber, 2005). This scenario predicts that a focal species may adapt faster to abiotic change(e.g., eCO2) when it is subject to selection in a species-rich community. As with scenario two, a significantCO2sel × ∆CO2 × divsel interaction should support this scenario, except with the opposite relationship todiversity (Figure 4.1e, f).The fourth scenario is that species diversity changes the fitness landscape (Figure 4.1g, h) The biologicalcommunity may modify the selection environment due to eCO2 thereby changing the shape of the fitnesslandscape. That is, the surrounding community may act like a prism, transforming an applied selective pres-sure into the selective pressure a focal species actually undergoes (Van Doorslaer et al., 2010; Lawrenceet al., 2012; Osmond and de Mazancourt, 2013). For example, increased CO2 might select for faster growth,resulting in selection for more efficient nitrogen use in a species-poor community but not in a species-richcommunity that includes nitrogen-fixing plants (Lee et al., 2003). As another example, belowground micro-bial biomass has been found to decline with eCO2 in species-poor communities but to rise in species-richcommunities (Eisenhauer et al., 2012), thereby potentially altering the supply and abundance of various nu-trients to plants (Mikola et al., 2002) in different biotic and abiotic contexts imposing selection on plants.364.2. Materials and MethodsWhatever the mechanism, this scenario predicts local adaptation to eCO2 when fitness is assessed in a com-munity with similar diversity as the community in which selection occurred (Figure 4.1g-h). Statistically,this scenario should be supported by a significant interaction between CO2 selection environment (CO2sel),change in CO2 environment (∆CO2), and change in diversity (∆div).We tested these predictions using a reciprocal transplant experiment (Figure 4.2) involving Poa pratensis(Kentucky bluegrass), a species widespread and abundant across several continents, and one of the morecommon species in BioCON (Reich et al., 2001b). We collected seeds from plots that had been exposedfor 14 years to ambient or elevated (ambient + 180 parts per million) concentrations of CO2 in species-poor(monoculture) or species-rich (16 species) grassland communities (Reich et al., 2001a). We transplantedindividuals with all four “histories” into all four of these treatment combinations. Consistent with the fourthscenario, we find that the biological community alters the fitness landscape in elevated CO2, so that localadaptation is observed primarily when species are grown in a community similar to the one in which theywere previously selected.4.2 Materials and Methods4.2.1 Sampling designBioCON was initiated in 1997 and consists of six, 20 m-diameter circular rings, each with ∼ 66 2x2 m plantcommunities (i.e., plots) (Reich et al., 2001a). In three randomly selected rings, atmospheric CO2 is ele-vated by ∼ 180 ppm above ambient, using free-air carbon dioxide enrichment technology (FACE), the otherthree rings are maintained at ambient conditions. In each plot, 1, 4, 9 or 16 grassland species were initiallyseeded (12 g m-2 of seed partitioned equally among all species planted in a plot). The 16 species planted intoBioCON are four C4 grasses (Andropogon gerardii, Bouteloua gracilis, Schizachyrium scoparium, Sorghas-trum nutans), four C3 grasses (Agropyron repens, Bromus inermis, Koeleria cristata, Poa pratensis), fournitrogen-fixing legumes (Amorpha canescens, Lespedeza capitata, Lupinus perennis, Petalostemum villo-sum), and four non-nitrogen-fixing forbs (Achillea millefolium, Anemone cylindrica, Asclepias tuberosa,Solidago rigida). The plots are maintained through regular weeding. Although BioCON also manipulatednitrogen (ambient and elevated), only plots exposed to ambient nitrogen were included in the current study.Poa pratensis (Poaceae), the focal species in our study, is a perennial, facultatively apomictic grass.It reproduces largely via asexually produced seeds (Mazzucato et al., 1996; Akerberg, 1939) or via tillers(ramets). Although native to Europe, P. pratensis is extensively naturalized in North America due to its useas a fodder and turf grass (USDA and NRCS, 2015).To assess the impact of community diversity on adaptation to CO2, we conducted a reciprocal trans-plant experiment, with an initial six-month period of plant growth and vegetative reproduction in a commongreenhouse environment to reduce maternal effects. In June 2011 we collected P. pratensis seeds fromspecies-poor (monoculture) and species-rich (16-species) plots in aCO2 and eCO2 conditions (Figure 4.2a).Four of the six BioCON rings contain a single P. pratensis monoculture plot (two rings with aCO2 and twowith eCO2). Within these four rings, we sampled seeds from both the monoculture plot and 16-species plots.For each P. pratensis monoculture plot, we sampled eight evenly spaced individuals (i.e., 2 rings × 1 plot374.2. Materials and MethodsAssay	CO2	environment	aCO2	 eCO2	 aCO2	 eCO2	Selec4on	diversity	environment		Response	to	selec4on	(eCO2sel	 –	aCO2sel)	 Species	diversity	constrains	adapta4on	CO2sel	x	ΔCO2	x	divsel	Species	diversity	promotes	adapta4on	CO2sel	x	ΔCO2	x	divsel	Species	diversity	alters	the	fitness	landscape	CO2sel	x	ΔCO2	x	Δdiv		Species-poor	 Species-rich	c	 d	e	 f	g	 h	Scenarios	Rich	Poor	Assay	diversity	environment	0	0	0	a	 b	0	Species	diversity	has	no	effect	on	adapta4on		ΔCO2	Figure 4.1: The hypothetical influence of species richness on adaptation of plants to elevated CO2. Theresponse of a focal species to selection in a given context can be quantified as the difference in performance(for example, biomass or fitness) between plants originating from elevated CO2 (eCO2) plots and those fromambient CO2 (aCO2) plots (y axis). (a,b) If species diversity has no effect on adaptation, local adaptation toeCO2 should be similar in species-poor and species-rich communities regardless of the species richness of theassay environment. (c,d) If species diversity constrains adaptation, an evolutionary response to eCO2 shouldbe more evident for plants that experienced selection in a species-poor community than in a species-richcommunity, regardless of the species richness of the assay environment. (e,f) If species diversity promotesadaptation to eCO2, then plants that experienced selection within a species-rich community may show greaterfitness in eCO2 regardless of assay species richness. (g,h) If species diversity alters the fitness landscape inresponse to CO2, then plants may only show improved fitness in eCO2 when planted back into a communityof similar richness to the one in which they experienced selection. Each scenario can be represented by aparticular statistical term, shown on the right.384.2. Materials and MethodsaCO2	&	1	sp.	8	samples	x	2	plots	=	16	mothers	aCO2	&	16	spp.	2	samples	x	8	plots	=	16	mothers	eCO2	&	1	sp.	8	samples	x	2	plots	=	16	mothers	eCO2	&	16	spp.	2	samples	x	8	plots	=	16	mothers	aCO2	&	1	sp.	(n	=	63)	aCO2	&	16	spp.	(n=	64)	eCO2	&	1	sp.	(n=	63)	eCO2	&	16	spp.	(n=	68)	aCO2	&	1	sp.	(4	x	63	=	252)	aCO2	&	16	spp.	(4	x	64	=	256)	eCO2	&	1	sp.	(4	x	62	=	252)	eCO2	&	16	spp.	(4	x	62	=	276)	aCO2	&	species-poor	eCO2	&	species-poor	eCO2	&	species-rich	(a)	Seed	collec8on:		Seeds	were	collected	from	each	BioCON	selecCon	environment	(June	2011)	(b)	Common	garden	treatment:	Four	seeds	(offspring)	from	each	of	the	16	mothers	were	germinated	per	selecCon	environment	&	grown	in	a	common	environment		(Dec	2011-May	2012)		(c)	Common	garden	treatment:		Four	ramets	per	offspring	were	collected	&	transported	to	BioCON		(May	2012)	(d)	Reciprocal	Transplant	experiment:		One	ramet	per	offspring	was	planted	into	each	assay	environment		(June	2012)	aCO2	&	species-rich	Figure 4.2: Steps involved in the transplant experiment. (a) P. pratensis seeds were collected from BioCONplots from different CO2 and diversity treatments. (b) Seeds were germinated in a common garden green-house environment to reduce maternal environmental effects. Values in brackets indicate the total numberof germinated seeds per BioCON selection environment (departures from 64 due to failed germination oradditional sampling are specified in the data table). (c) Four daughter ramets were sampled per germinatedseed (offspring) in the greenhouse and transported back to BioCON for the transplant experiment. (d) Onedaughter ramet per offspring was placed into each of the assay diversity and CO2 environments (see textfor additional details). 1 sp., P. pratensis monoculture BioCON plots; 16 spp., 16 species BioCON plots;species-poor, P. pratensis dominated species-poor plots; species-rich, species-rich plots (note that the lattertwo plots types describe the assay plots; Supplementary Methods C.1.3).394.2. Materials and Methods× 8 mothers = 16 mothers sampled per CO2 treatment). Each ring also contains four 16-species plots, fromwhich we sampled two widely spaced individuals per plot (i.e. 2 rings × 4 plots × 2 mothers = 16 motherssampled per CO2 treatment) (Figure 4.2a).4.2.2 Growth in a common greenhouse environmentIn order to reduce maternal environmental effects, all plants were subjected to a period of growth and veg-etative reproduction in a common greenhouse environment. We broke seed dormancy by storing the seedsin a refrigerator (∼ 4◦ C) in airtight containers with Drierite desiccant (to absorb moisture) for 5 months. InDecember 2011 we weighed all seeds and planted five per 656ml pot (6.4 × 25.4 cm Deepots, model D40H; Stuewe & Sons, Tangent, OR, USA), filled with potting soil at the University of British Columbia green-houses. Twenty seeds collected from the same mother were planted. Seed weight (mean = 9.6 mg per seed)was not statistically different between mothers or between CO2 or diversity treatments, suggesting that therewere not substantial differences in maternal provisioning among treatments. After planting, the pots wererandomized and placed on an unlit greenhouse bench and misted with water every 20 minutes. When the firstseed within a given pot germinated, the pot was moved to 16 hours of full spectrum light. 95% of pots hadat least one germinated seed within 3 weeks of planting and within one week of one another. Only two potscompletely failed (seeds never germinated and/or all seedlings died). The most robust seedling per pot waskept and the rest were removed, yielding a total of 258 plants (4 plants × 16 mothers × 2 CO2 treatments ×2 diversity treatments - 2 that failed + 4 that were mistakenly planted) (Figure 4.2b). All plants were handwatered every second day with water enriched with fertilizer and all pots were randomized monthly.The greenhouse plants never produced inflorescences, so we used vegetatively propagated daughter ram-ets (i.e., ramets not from the central clump) for reciprocal transplantation back into BioCON. From 1- 4 May2012 we took five to six young ramets of approximately equal size from each P. pratensis plant raised in thegreenhouse. These ramets were then planted into 164ml (3.8 × 21cm) cone-tainers (Ray-Leach, model SC10 Super; Stuewe & Sons, Tangent, OR, USA) filled with potting soil and maintained under standard green-house conditions (Figure 4.2c). On 27 May 2012 we removed all ramets from their cone-tainers and washedthe roots free of soil. The roots were then wrapped in moist paper towel, and 1364 ramets were transportedback to BioCON. The healthiest four ramets per plant were weighed and then randomly assigned to one ofthe four treatment groups (aCO2 species-poor; aCO2 species-rich; eCO2 species-poor; eCO2 species-rich)(Figure 4.2d). No significant difference in biomass between ramets assigned to each treatment group wasfound.4.2.3 Planting ramets in the assay plotsDue to the disturbance that planting and watering a substantial number of ramets would have imposed on theon-going BioCON experimental plots, we used supplementary plots (assay plots) that were established inthe BioCON rings in 1999 but that were not in use at the time of our study. These plots (1.5m × 2m, six perBioCON ring) were created within the FACE rings but on the edge of the main BioCON plots. Within eachof the six BioCON rings, we chose three species-poor plots, consisting predominantly of P. pratensis (andfrom which we regularly weed out other species), and three species-rich plots that best matched the original404.3. Statistical analysismonoculture and 16-species plots (Figure C.4; Supplementary methods C.1.3). On 6-8 June 2012 rametswere planted into these 36 plots.Ramets from all four CO2 × diversity source combinations were planted into each plot. As P. pratensisis rhizomatous, each plot was divided into quadrants (1m x 0.75m) by sinking sheet metal to a depth of30cm to prevent individuals from different selection environments interfering with one another. Within eachquadrant, six to eight plants from the same selection environment, randomly drawn from among the mothersof the appropriate treatment type, were planted. Planted ramets were individually marked by loosely placinga piece of coloured wire around the base of each plant. To reduce transplant shock, the planted ramets werewatered every day for the first two weeks, every second day for the third week, and every third day for thefourth week, after which watering was discontinued. Survival was high: 98.9% of individuals survived thefirst month. Ramets that died were not replaced.4.2.4 Measurement of plant performanceTransplanted ramets were grown in the field for two growing seasons. We assessed the survival of eachramet once per month between June and August in 2012 and 2013. In June 2013 we counted the numberof inflorescences produced by each plant, and at the end of August 2013 all aboveground biomass washarvested, dried, and weighed. To estimate belowground biomass we measured root mass in a standardizedvolume of soil. Roots were cored by placing a 5cm diameter PVC tube around the originally planted rametand hammering it to a depth of 30cm; all roots and soil contained in this core were extracted. This methodwas used instead of attempting to extract the entire root mass, due both to logistical constraints and becausepreliminary observations suggested little clonal expansion such that the majority of root growth would becaptured within the cored sample. Consequently, we consider this to be a surrogate for root density (biomassper soil volume), rather than total root biomass, but for ease of reference we refer to this as “ belowgroundbiomass”. After the roots were extracted, they were washed free of soil and any roots obviously from adifferent species were removed. The roots were then dried and weighed.Previous work has shown that total vegetative weight (stems, leaves and roots) and reproductive weight(fruits, surrounding glumes and rachis) are highly correlated in P. pratensis (Wagner, 1989). Similarly,amongst the plants that flowered, we find that the combined weight of fruits, glumes and rachis was highlycorrelated with both aboveground biomass (P«<0.0001, adjusted r2 = 0.481) and total biomass (P«<0.0001,adjusted r2 =0.428). We thus consider biomass to be a rough proxy for fitness.4.3 Statistical analysis4.3.1 Analysis of biomass dataLocal adaptation to CO2 environment while holding diversity environment constant - We first tested forlocal adaptation to CO2 by holding the diversity environment constant; i.e., by only including plants thatwere selected and assayed in the same diversity environment. Separate linear mixed effects models wereperformed on the logarithm (to meet assumptions of normality) of aboveground, belowground and total(aboveground + belowground) biomass. As fixed factors, we included all single and two-way interactions414.3. Statistical analysisbetween the previous CO2 selection environment (CO2sel), the change in CO2 environment (∆CO2) relativeto the CO2 selection environment, and the general diversity environment (i.e. either species-rich to species-rich or species-poor to species-poor). All models also included the following random effects: selection plot,family (mother′s ID) nested within selection plot, assay ring and assay plot nested within assay ring. Therewas no statistical difference between the rings in which the plants were selected and much of this variance islikely absorbed by the term selection plot; thus we excluded selection ring from all models. We implementedall final models using the restricted maximum-likelihood method (REML) to estimate variance components(Zuur et al., 2009). The significance of all fixed effects was evaluated using Type III estimable functions,and denominator degrees of freedom were determined using Satterthwaite’s approximation (Satterthwaite,1946).Testing scenario one: Local adaptation to CO2 environment while averaging over diversity environment- To determine whether local adaptation occurred regardless of diversity environment we performed separatelinear mixed effects models on the logarithm of aboveground, belowground and total biomass. Analyses wereperformed as described above except that the performance of plants was averaged over the assay diversityenvironment (divass), while selection diversity environment (divsel) was included in the model instead of thegeneral diversity environment.Testing scenarios two, three and four - We performed separate linear mixed effects models on the log-arithm of aboveground, belowground and total biomass. As fixed factors, we included the previous CO2selection environment (CO2sel) and diversity selection environment (divsel), which together define the Bio-CON plot from which each individual’s mother was sampled. We also included fixed-factor terms indicatingwhether the plants experienced a change in CO2 (∆CO2) or in diversity (∆div), relative to the plots of theirmothers. As an alternative statistical approach, we also analysed the data by treating the selective and assay(current) environments (CO2ass and divass) (not the change in environments) as fixed factors. All modelsalso included the following random effects: selection plot, family (mother’s ID) nested within selection plot,assay ring and assay plot nested within assay ring into which ramets were transplanted. For the reasonsoutline above selection ring was excluded from all models.To test for our hypothesised interactions we ran each full model and eliminated non-significant termsusing likelihood ratio tests until the model contained the CO2sel × ∆CO2 × divsel (scenario two and three)and CO2sel × ∆CO2 × ∆div (scenario four) interactions and all lower terms. If a different 3-way or the4-way interaction was significant then this was retained in the model and lower terms were not eliminated.The fit of each final model was assessed through visual inspection of the fitted and residual values, all modelswere found to meet assumptions of normality and homogeneity (Zuur et al., 2009). Data analysis was carriedout in R version 3.0.2 (RCoreTeam, 2013) using the lme4 (Bates et al., 2014) and lmerTest (Kuznetsova et al.,2014) packages. The analysis of biomass was carried out on all individuals that survived to the end of theexperiment. However, data from 49 individuals that were growing in two plots that burned in a fire in May2013 (three quarters of the individuals in one species-poor plot and all individuals in one species-rich plotin eCO2) and 34 plants that could not be relocated in spring 2013 were excluded. Thus in total, data for765 individuals were included for the biomass analyses. In addition, 11 belowground biomass samples wereexcluded from the analysis as these cores were either hammered in at an angle that missed most of the root424.4. Resultsmass during the core extraction process or half the soil fell out of the core during extraction. We implementedall final models using the restricted maximum-likelihood method (REML) to estimate variance components(Zuur et al., 2009). The significance of all fixed effects was evaluated using Type III estimable functions,and denominator degrees of freedom were determined using Satterthwaite’s approximation (Satterthwaite,1946).4.3.2 Analysis of survival and inflorescence productionTo assess the importance of CO2sel, ∆CO2, divsel and ∆div on survival and inflorescence production weused aster models with random effects (Geyer et al., 2013) implemented in R (RCoreTeam, 2013). Astermodels facilitate the analysis of multiple life history stages as they can analyse survival and reproductionjointly. Furthermore, different life history traits can be modelled with different probability distributionsand account for the fact that later components of fitness (e.g. flowering) depend on earlier components offitness (e.g. survival) (Shaw et al., 2008). In our Aster model we included 1) survival until the time ofinflorescence production in June 2013 (Bernoulli), 2) whether a surviving plant produced inflorescences(Bernoulli) and 3) the number of inflorescences produced (zero-truncated Poisson). The influence of eachfixed factor was tested by comparing models with a given factor to a model containing only the randomeffects (null model) using log-likelihood ratio tests. Similarly the significance of each two-way, three-wayand four-way interaction was tested individually against a reduced model that contained all lower terms.Selection plot, maternal family and assay plot were included as random effects. Analyses were carried outin the same order as described above for biomass: i.e., we firstly tested for local adaptation holding diversityenvironment constant, then we tested for local adaptation averaging across the diversity assay environments,and lastly, we tested scenarios two, three and four. We also confirmed our results by re-running all modelsusing CO2cur and divass instead of ∆CO2 and ∆div (Supplementary Note C.1.1, and Table C.6). Fromthe Aster analysis, we excluded 17 plants that died within the first month after transplantation (June 2012)because death was more likely to be due to transplant shock than any other factor of interest, as well as the 8individuals that could not be relocated in the spring of 2013 and the individuals that were in the two burnedplots.4.4 Results4.4.1 Results for biomass productionWe found that P. pratensis locally adapted to the CO2 environment, but only when the diversity of the com-munity was the same in the past “selection” and current “assay” environment. That is, the term ∆CO2 wasstatistically significant for aboveground (F1, 306.6 =7.8, P =0.006) and total biomass (F1, 302.9 =4.1, P =0.04)(Supplementary Table C.1) when plants were selected and assayed in the same diversity treatment (species-poor to species-poor or species-rich to species-rich) (Supplementary Figure C.1). However, if adaptation toCO2 was assessed by averaging the results from both species-rich and species-poor assay plots, local adap-tation was not statistically detectable for any measure (Supplementary Table C.2 and C.3; SupplementaryFigure C.2). We thus find that the local adaptation does not depend solely on the community context during434.4. Resultspast selection but also depends on the current assay environment, leading us to reject scenarios one to three(Figure 4.1).Local adaptation was only seen when we took into account both the diversity of the selection environmentand the assay environment. This indicates support for scenario in which species diversity alters the fitnesslandscape (Figure 4.1g-h): plants exhibited adaptation to eCO2, but only when assayed in the communitycontext in which selection occurred (Figure 4.3a-c). For example, the response of aboveground biomass toselection in eCO2 (Figure 4.3a) was positive in species-poor environments (left) but chiefly when assayed inspecies-poor environments (dark triangles). Similarly, plants from species-rich environments showed adap-tation to eCO2 but only when assayed in species-rich environments (circles on the right). As a result, whenperforming a full statistical analysis including community diversity in both selection and assay environ-ments, there were significant 3-way CO2sel ×∆CO2 ×∆div interactions for aboveground (F1, 689.1 = 5.8, P= 0.016), belowground (F1,690.3 = 4.2, P = 0.041) and total biomass (F1, 684.3 = 4.2, P = 0.039) (Supplemen-tary Table C.3), with increased adaptation to eCO2 when the community context remained the same betweenselection and assay environments (Figure 4.3a-c). These significant interactions were consistent with plantsdemonstrating a “home” plot advantage when assayed in a biotic and abiotic environment similar to the onein which they experienced selection (Supplementary Figure C.3a-c).4.4.2 Results for survival and inflorescence productionThe high survival of individuals across treatments (>80% survival) and the lack of flowering within species-rich plots (only 9% of individuals produced inflorescences) reduced our power to analyse these fitness com-ponents (see Supplementary Figure 4.3d). Although not significant, the direction of the results from anAster analysis (Shaw et al., 2008) of survival and inflorescence production were consistent with the fitnesslandscape scenario (Figure 4.3d, Supplementary Table C.5).4.4.3 Analysis with selection and assay environmentsAs an alternative statistical approach, we also analysed the data by treating the selection and assay environ-ments (CO2ass and divass) (not the change in environments) as fixed factors (Supplementary Table C.4, C.6).This approach has reduced statistical power because the fourth scenario, the fitness landscape scenario, mustbe tested via a 4-way interaction (CO2sel × CO2ass × divsel× divass) while the second and third scenarios aretested via three-way interactions (CO2sel × CO2ass × divsel). Nevertheless, these results were also consistentwith the fitness landscape scenario (Supplementary Table C.4 and C.6). Analysing the data in this alternativeway aided in teasing apart immediate responses to the environment (i.e., “plastic” responses) from evolu-tionary changes. Indeed, we also detected strong plastic responses, with greater biomass for plants assayedin species-poor plots and for plants assayed in eCO2 (Supplementary Table C.4 and C.6 and SupplementaryNote C.1.1).444.4. ResultsResponse	to	selec+on			(eCO2sel	 - 	aCO2sel)	Selec+on	diversity	environment		Species-poor	 Species-poor	Assay	CO2	environment	aCO2	 eCO2	 aCO2	 eCO2	 aCO2	 eCO2	 aCO2	 eCO2	a	 b	c	 d	-0.4-0.20.00.20.4 Aboveground biomass (g)      -0.4-0.20.00.20.4 Belowground biomass (g)      -0.6-0.20.20.6 Total biomass (g)     -1.5-0.50.51.5 Number of inflorescences     -0.4-0.20.00.20.4 Aboveground biomass (g)      -0.4-0.20.00.20.4 Belowground biomass (g)      -0.6-0.20.20.6 Total biomass (g)     -1.5-0.50.51.5 Number of inflorescences     Rich	Poor	Assay	diversity	environment	Species-rich	Species-rich	Figure 4.3: Response to selection under elevated versus ambient CO2 in species-poor and species-rich com-munities. For each assay CO2 environment we calculated the difference in (a) aboveground biomass, (b)belowground biomass, (c) total biomass in grams or (d) number of inflorescences (±1 s.e.m.) produced byplants that had previously experienced selection in eCO2 and in aCO2 from the raw data. Results are mostconsistent with the scenario that species diversity alters the fitness landscape Figure 4.1(g, h).454.5. Discussion4.5 DiscussionOur study contrasted four scenarios for how the diversity of the surrounding neighbourhood communitycould impact evolutionary change in a focal species. Overall our results were most consistent with the sce-nario in which species diversity alters the fitness-landscape (Figure 4.1g-h) and indicate that while adaptationto eCO2 confers a performance advantage in both community contexts, the advantage does not transfer di-rectly across community contexts. Studies examining the effect of eCO2 on plants and communities havefound that eCO2 typically accelerates plant growth, alters plant tissue chemistry (Knops et al., 2007), in-creases plant biomass and reduces evapotranspiration (Poorter and Navas, 2003). These changes result inincreased soil water content (Adair et al., 2011), altered belowground microbial diversity (Eisenhauer et al.,2012) and modified nutrient cycling. The nature of these changes, however, is likely to depend on thesurrounding community. Indeed, increasing plant diversity also reduces soil moisture (Adair et al., 2011),increases microbial community diversity (Eisenhauer et al., 2013) and increases nutrient cycling (Reich et al.,2012). These changes are expected to interact (Eisenhauer et al., 2013) to shape the selective environment inwhich a plant grows. Since competition exerts strong selective pressures, we suggest that it is the combinedeffect of altered abiotic environment and changed competitive interactions shaping the fitness landscape thatdrive our result. A similar perspective is that, as mutations arise, their pleiotropic effects generate differentselective trade-offs depending on the surrounding community, altering which mutations can spread (whichis another way of saying that the community shapes the fitness landscape experienced by new mutations).Our finding that adaptation to eCO2 does not transfer across community contexts may have importantimplications for understanding how species will respond at both small and large scales to rising global CO2levels caused by fossil-fuel emissions. At small scales the composition of communities can vary tremen-dously due to aspect, slope, soil, etc., and the evolutionary response of a metapopulation to a commonselective pressure (e.g., changing CO2) might be experienced in different ways at different sites, with thelocal community altering the selection experienced. With high gene flow between populations, adaptation torising CO2 levels may be hampered as a consequence of maladaptation to the biotic environment. At largerscales, species are likely to shift their range boundaries in response to climate warming and therefore toencounter novel community contexts (Parmesan and Yohe, 2003; Walther et al., 2002). If range shifts alterthe biotic community, previous adaptations to the abiotic environment (e.g. elevated CO2 or temperature)may no longer improve fitness when in a different community context.To our knowledge, our study is the first experiment conducted in a natural field setting to test whetheradaptation to an abiotic change in a macro-organism is impacted by community context. In plants, severalstudies have investigated the interaction between abiotic and biotic conditions on adaptation, but they eitherwere not conducted in a natural field setting (i.e., a pot experiment with a single competitor species (Lauet al., 2014)) or they focused on immediate phenotypic responses (Bazzaz and Garbutt, 1988; Klanderud,2005; Tomiolo et al., 2015; Alexander et al., 2015). By performing a reciprocal transplant experiment ofplants propagated under different biotic and abiotic conditions for 14 years, we can distinguish plastic fromevolutionary responses in different community contexts.Another contribution of this experiment is that it provides an additional compelling example of localadaptation at both a small spatial and temporal scale. Adaptation was observed in a perennial species in just464.5. Discussion14 years to an important global change driver (eCO2) in a manner that depends on the community contextat a microgeographic scale. This work thus contributes to the growing number of examples from fieldpopulations of local adaptation over short temporal (Hendry and Kinnison, 1999; Reznick and Ghalambor,2001) or microgeographic (Richardson et al., 2014; Linhart and Grant, 1996) scales.There are some limitations to our study. First, the response to eCO2 was relatively modest, and re-sponses to stronger agents of selection could be less or more contingent on biotic context. Second, therewas limited scope for replication given the design of the original BioCON experiment; in particular, onlytwo monoculture plots existed for P. pratensis under each CO2 treatment. Another consideration is thatalthough P. pratensis reproduces predominantly asexually through rhizomatous growth or apomixis (Maz-zucato et al., 1996), making selection within and among clones a likely mechanism of adaptation, on-goinggene flow, through seed or pollen dispersal, from other BioCON plots or the surrounding prairie plant com-munity (including P. pratensis) cannot be discounted (although this would reduce the likelihood of observingthe selection responses that we did). We should also emphasize that we have compared only two types ofcommunities, species-rich and species-poor. We do not know the extent to which our results are drivenby diversity, per se, versus simply the presence of particular other species. However, an analysis of our P.pratensis aboveground biomass data versus percent cover of other species in the species-rich plots did not in-dicate that one species was driving our results (Supplementary Note C.1.2 and Supplementary Table C.7 andC.8). Moreover, earlier studies in BioCON of effects of species-poor vs. species-rich neighborhoods on fo-cal species show that resource competition and environmental stress amelioration (both via higher biomassrelated effects) are both enhanced by higher diversity (Wright et al., 2013, 2014). Such effects may wellhave been at work in the current study. Finally, as with most studies of this nature, we cannot be certainthat maternal effects were completely eliminated by clonal growth for six months in a common greenhouseenvironment. Thus, it remains possible that some of the evolutionary responses we have documented arein fact transgenerational maternal effects: including effects from microbial symbionts . Importantly, theselimitations point largely to reasons one might expect not to find adaptation to CO2 in these plots over the past14 years, but we did indeed find evidence for adaptation that fell clearly in line with one a priori prediction(and not the others).Overall, our results indicate that the evolutionary response of a plant to elevated CO2 is manifestedprimarily when grown in the same type of biological community in which it evolved (species-rich or species-poor). This pattern supports the view that the selection imposed by a shift in the abiotic environment isexperienced through the prism of surrounding species, altering the form of selection actually experiencedby a focal species, consistent with the fitness landscape scenario. How often the biological community actsto change the selection experienced in altered environments can only be determined by additional empiricaltests of these scenarios.47Chapter 5How plant diversity influences intraspecifictrait responses to abiotic environmentalchange5.1 IntroductionClimate change is affecting all species, by causing changes in abundance, distributions and/or species inter-actions (Urban, 2015; Vellend et al., 2017). At the same time, concentrations of atmospheric CO2 and soilnitrogen are increasing at an accelerating pace (Canadell et al., 2007; Galloway et al., 2008; Vitousek et al.,1997). Changes in community composition may alter how plant species respond to these abiotic driversbecause plant-plant interactions are known to mediate the effects of abiotic change (reviewed in Brooker,2006). For example, in a previous study we found that Poa pratensis adapted in response to elevated CO2(eCO2) (via changes in biomass and seed production) in a manner that depended on the surrounding species,so that plants adapted to eCO2 in different ways when living in communities of low and high diversity (Kleyn-hans et al., 2016). Consequently, we also might expect plant functional traits to experience different selectivepressures and hence to vary across biotic and abiotic environments. Understanding how the functional traitsof species will respond to abiotic change in different community contexts is imperative because functionaltraits influence species coexistence and the functioning of ecosystems (Lavorel and Garnier, 2002). Althoughchanges in species composition (Brooker, 2006), biomass and reproduction (e.g. Arp et al., 1993; Kleynhanset al., 2016; Lau et al., 2014; Wedin and Tilman, 1993) and short-term responses in functional traits (Carterand Peterson, 1983; Lau et al., 2010; Patterson et al., 1984; Wray and Strain, 1987) have been observed inresponse to differences in the competitive community under different abiotic treatments, how long-term ex-posure to different abiotic treatments in communities of low and high functional and species diversity mightalter plant functional trait-responses has not to our knowledge been explored. Here, we study the functionaltraits expressed by seven different species from four functional groups. Each species has been growing for16 years in experimental plots that contain species from one or four different functional groups exposed toambient and elevated levels of both CO2 and nitrogen in an experiment known as BioCON (Reich et al.,2001a).Functional traits, such as plant height and leaf area, have been shown to respond to community diversityindependently of abiotic change. Several studies examining the responses of forbs (Lipowsky et al., 2015),grasses (Gubsch et al., 2011) and legumes (Roscher et al., 2011) grown in plots of increasing diversity haveall found that plant height, leaf length and specific leaf area tend to rise. This is likely due to increased485.1. Introductioncompetition for light (Weisser et al., 2017), where taller plants intercept light, thereby prompting phenotypicresponses (e.g., leaf structure or physiology) of shorter plants in response to low light (Navas and Violle,2009). Furthermore, evidence suggests that there may be strong selection for niche differentiation drivingcharacter displacement of plant traits in diverse mixtures of species (Schöb et al., 2018; Zuppinger-Dingleyet al., 2014). If high diversity forces species to partition niche space to reduce competition then the potentialof most species to adapt to abiotic change in diverse communities may be reduced relative to that in speciespoor communities. Furthermore, high diversity might increase the probability that at least some species arepresent that are "pre-adapted" to the new conditions, whose increased abundance might reduce the abilityof other species in the community to respond to the new conditions (Bocedi et al., 2013; de Mazancourtet al., 2008; Johansson, 2008; Norberg et al., 2012). In contrast, under low diversity species might be lessconstrained in niche space and therefore more likely to respond positively to abiotic change (de Mazancourtet al., 2008). Thus, we predict that in general, functional trait changes of species will be influenced by aninteraction between species richness and abiotic change.In pot experiments in which plants are grown with an con- and heterospecific competitor, the withingeneration responses of functional traits, such as leaf area or plant height, to elevated CO2 (eCO2) havegenerally been insignificant (e.g. Carter and Peterson, 1983; Lau et al., 2010; Patterson et al., 1984; Wrayand Strain, 1987), with the exception of Patterson et al. (1984). Patterson et al. (1984) found that the C4 grassSorghum halpense did not respond in leaf area or height to eCO2 in monoculture, but leaf area decreased andheight increased when grown together with a C3 nitrogen fixing competitor in eCO2. In contrast to traits,several studies examining changes in biomass and/or reproduction have also found significant interactionsbetween the abiotic environment and community diversity (e.g. Arp et al., 1993; Bazzaz and Garbutt, 1988;Kleynhans et al., 2016; Lau et al., 2010, 2014; Navas et al., 1999; Steinger et al., 2007; Wedin and Tilman,1993). Interestingly, the functional group identity of the competitor/s and focal species seemed to influencethis response.Species are often found to cluster in multidimensional trait space (Hooper et al., 2002; Petchey and Gas-ton, 2002) and these species usually have "functionally" similar impacts on ecosystem processes (Díaz andCabido, 2001). These groups of species are called functional groups and species within the same functionalgroup are expected to respond in similar ways to abiotic change (Díaz and Cabido, 2001). Furthermore,like species richness, communities with greater functional diversity tend to be more productive (Reich et al.,2004; Tilman et al., 1997, 2001). For example a change in functional group diversity from one to four hasbeen observed to increase total biomass by 28-30% when species richness was held constant at four species(Reich et al., 2004). Consequently, species grown in communities made up of a single functional groupmay respond more similarly to if they were grown in monoculture while species grown in communities offour functional groups may respond more similarly to if they were grown in highly diverse communities.Therefore, changes in functional group diversity are likely to have a large impact on how species respond toabiotic change and may act in a similar way to increasing species richness (Reich et al., 2004; Tilman et al.,1997, 2001). Accordingly, we test whether functional trait responses to abiotic conditions are modified byfunctional group richness. We focus on functional group richness as our index of diversity because plotsof one and four functional groups are better represented within BioCON than monoculture plots. We also495.1. Introductioninvestigate the impact of species richness directly but because results are similar we present these in thesupplementary materials.Overall, assuming that changes in biomass reflect selective pressures that also shape functional traits,we outline some predictions for the response of traits to elevated CO2 or nitrogen in communities consistingof species from the same or from four different functional groups. The four functional groups we examineare non-nitrogen fixing C3 forbs, non-nitrogen fixing C3 grasses, nitrogen fixing C3 legumes (referred to aslegumes), and non-nitrogen fixing C4 species.5.1.1 Response of species from different functional groups to elevated CO2Species with C3 photosynthesis have generally been found to be more responsive to eCO2 than species withC4 photosynthesis (Bowes, 1993; Lee et al., 2001, but see Reich et al. (2018)). Thus, when C3 forbs andgrasses are grown in con- and heterospecific competition with C3 and/or C4 species in ambient and elevatedCO2 the C3 forbs and grasses tended to have a larger increase in biomass in mixture with C4 species thanwhen grown only with other C3 forb and grass species (e.g. Arp et al., 1993; Carter and Peterson, 1983; Lauet al., 2010; Wray and Strain, 1987, but see Bazzaz and Garbutt (1988)). Similarly, C4 species have a largerresponse in mixture than in a community made up of only C4 species, but in the opposite direction to thatobserved in C3 forbs and grasses (e.g. Arp et al., 1993; Carter and Peterson, 1983; Lau et al., 2010; Wrayand Strain, 1987, but see Bazzaz and Garbutt (1988)) .When legumes were grown in monoculture in eCO2 biomass was increased by a large amount. Whereas,when legumes were grown in competition with C3 forbs and grasses, biomass was still increased but to alesser extent. In contrast, forbs and grasses increased in biomass in monoculture but showed no change inbiomass in competition with legumes in eCO2. This is likely to be a result of eCO2 increasing the flow oflabile carbon to the soil which stimulates free-living soil microbes to sequester nutrients, thereby reducing theavailability of nitrogen to plants (Lüscher et al., 1996). However, species that can fix nitrogen are less limitedby low nitrogen availability and therefore are stronger competitors than species that do not fix nitrogen(Lüscher et al., 1996; Patterson et al., 1984). Similar results were found for competition between legumesand C4 species (Patterson et al., 1984).If the selective pressures acting on plants are reflected in biomass changes then we predict that changes infunctional traits should be smallest in legumes and largest in C4 plants to eCO2. With this in mind we predictthat for most species we should find a significant interaction between abiotic change and plot diversity, withthe possible exception of legumes.5.1.2 Response of species from different functional groups to elevated nitrogenThe biomass response of plants to nitrogen has been shown to be mediated by the presence of other func-tional groups in the competitive community but whether these differences translate to changes in functionaltraits remains unknown. C3 forbs and grasses are typically stronger competitors in nitrogen rich environ-ments than C4 species (Wedin and Tilman, 1993, 1996). For example, Wedin and Tilman (1993) grew P.pratensis, Agropyron repens (C3 grasses), and Schizachyrium scoparium (C4 grass) in monoculture or withone other competitor species, with or without additional nitrogen. Without additional nitrogen, S. scoparium505.2. Materials and Methodsdisplaced or greatly reduced the biomass of the other two C3 species. However, when nitrogen was addedthe biomass of S. scoparium was reduced while that of the competitor species, P. pratensis or A. repens,was greatly enhanced relative to when grown in monoculture. Avolio et al. (2014) and Isbell et al. (2013)found similar results where the biomass of C4 grasses declined and the biomass of C3 grasses increasedwith additional nitrogen. Interestingly, Avolio et al. (2014) found relatively little change in the abundanceof nitrogen fixing forbs in response to added nitrogen, while Isbell et al. (2013) found that the abundanceof legumes declined with additional nitrogen. If selective forces are reflected in changes in biomass thenwe expect large trait responses of both C3 forbs and grasses and C4 species to elevated nitrogen in mixturebut in opposite directions. However, legumes that are not nitrogen limited may not alter traits in response toelevated nitrogen in any community setting and therefore should show no difference in trait values betweencommunities of one and four functional groups.We test these predictions by measuring the functional traits of seven different species grown for 16 yearsin experimental plots planted with one or four different functional groups in ambient and elevated CO2and nitrogen. Counter to our predictions we find no effect of the interaction between plot functional grouprichness and abiotic environment for any species. In contrast we find strong main effects of functional grouprichness and abiotic treatment suggesting that the effects of species richness many overwhelm those fromabiotic change, at least in the traits we measured.5.2 Materials and Methods5.2.1 BioCON experimentThis study was carried out in the Biodiversity, Carbon dioxide, Nitrogen addition field experiment (Bio-CON), which is a complete factorial manipulation of species richness, carbon dioxide and nitrogen locatedat the Cedar Creek Ecosystem Science Reserve in Minnesota (Reich et al., 2001a). BioCON was initiatedin 1997 and consists of six, 20m diameter rings. In three of the rings Free Air Carbon dioxide Enrichment(FACE) technology is used to elevate the atmospheric CO2 concentration to±180 p.p.m above ambient con-ditions, the other three rings are maintained at ambient conditions. Within each ring there are approximately66 plots, each 2 × 2m. Each plot was initially seeded with either 1, 4, 9 or 16 different perennial grasslandspecies by sowing 12 g.m-2 of seed divided equally among the species assigned to that plot. Invading speciesthat were not initially planted in a plot are removed via regular weeding. Furthermore, species that go ex-tinct in a plot are not reseeded. The 16 species grown in BioCON are four C3 grasses (Agropyron repens,Bromus inermis, Koeleria cristata, Poa pratensis), four C4 grasses (Andropogon gerardii, Bouteloua gra-cilis, Schizachyrium scoparium, Sorghastrum nutans), four nitrogen-fixing legumes (Amorpha canescens,Lespedeza capitata, Lupinus perennis, Petalostemum villosum) and four non-nitrogen fixing forbs (Achilleamillefolium, Anemone cylindrical, Asclepias tuberosa, Solidago rigida). In addition to manipulating speciesdiversity, functional group diversity was also manipulated in the four species plots by planting either fourspecies from a single functional group or combinations of species from different functional groups to createplots of two, three or four different functional groups (Reich et al., 2004). Lastly, half of the plots also receiveadditional nitrogen (4g N m-2 yr-1) applied over three dates each year (Reich et al., 2001a).515.2. Materials and MethodsIn this study, we focused on plots containing one, four and sixteen species with one or four functionalgroups that contained at least one A. gerardii, A. tuberosa, B. inermis, L. capitata, L. perennis, P. pratensis, orS. rigida individual. These seven species were our focal species because they are relatively well representedacross the diversity levels, and thus replicate trait measurements could be obtained in each treatment. Wesampled all monoculture plots, all four species one-functional group plots, all four species four-functionalgroup plots, and two randomly chosen 16-species plots that contained these species per abiotic treatment (seeTable D.1 for number of plots sampled per species and treatment). We obtained functional trait measurementsof plants growing in ambient CO2 and ambient nitrogen, elevated CO2 and ambient nitrogen, and ambientCO2 and elevated nitrogen for each diversity level. Measurements for elevated CO2 and elevated nitrogenwere not taken for logistical reasons. The measured traits were plant height, leaf size (one-sided leaf area),specific leaf area (one sided area divided by dry mass, SLA), leaf dry matter content (leaf dry mass dividedby its fresh mass, LDMC) and seed mass (the average mass of an individual seed). Due to low replication ofthe monoculture plots, four species one-functional group plots, and four species four-functional group plotswe lumped all four species one-functional group plots together with the monoculture plots to establish plotscontaining a single functional group, and we combined all four species four-functional group plots togetherwith the 16-species plots to establish plots containing four functional groups. In this way the effects of lowand high functional group diversity on trait responses of species to abiotic change could be investigated. Theeffect of species diversity (not considering functional group diversity) is also explored (Supp Mat D.1.1) but,due to the low replication of monoculture and four species plots this analysis lacks power. Nevertheless,results for this analysis are outlined in the Supp Mat. D.1.1 and D.4. Generally, results of the functionalgroup analyses and species richness analyses were qualitatively similar.5.2.2 Trait measurementsPlant height, leaf area, SLA, LDMC and seed mass were measured following the standardised methodsoutlined in Cornelissen et al. (2003) and Pérez-Harguindeguy et al. (2013). We measured these plant traitsbecause they are known to be involved in plant resource acquisition and use and are also likely to influencecompetitive ability (Kraft et al., 2015) under both elevated CO2 and nitrogen. For example, vegetativeheight is related to light acquisition and competitive ability (Gaudet and Keddy, 1988) where competitionfor light is likely to increase under both elevated CO2 and nitrogen. Similarly, changes in leaf traits such asSLA are related to photosynthetic rates and growth; high SLA values are typically associated with short leaflifespans and low investment into leaf structural defences and are associated with resource rich environments.Consequently, one might expect higher SLA values in elevated nitrogen but the opposite might be expectedin elevated CO2 if nitrogen then becomes limiting. LDMC is also correlated with growth and photosyntheticrates and is related to leaf resistance to physical stress. Leaf area is related to stress tolerance with smallerleaves being selected under drought stress (Cornelissen et al., 2003) or low nutrient availability (Pérez-Harguindeguy et al., 2013). Lastly, nutrient rich environments may allow plants to produce more or largerseeds which would improve plant fitness. In elevated CO2 seeds may be smaller or larger depending onwhether CO2 is effectively an additional resource or whether it leads to limited nitrogen. Overall, all thesetraits are likely influenced in different ways by the biotic and abiotic environment.525.2. Materials and MethodsWe measured plant height and sampled leaves at roughly the same time as sampling seeds, so that thesemeasurements coincided with the peak of a given species’ growing season (with the exception of S. rigida,which was sampled a month earlier than seed maturation). Plant height was measured, and leaves and seedswere sampled for: L. perennis on 19 - 21 June, P. pratensis on 1 - 2 July, A. tuberosa on 24 July, B. inermison 25 - 26 July, L. capitata on 5 - 8 August, S. rigida on 13 -14 August, and A. gerardii on 16 August.Although heights and leaves were sampled on different dates for different species, comparisons were madewithin species and not between species.To measure plant height, we first laid out a 40 × 40cm grid within each 2 × 2m plot. At each gridintersection, we searched for the closest mature plant of each species expected in the plot that was exposedto full sunlight and measured its standing height (sample sizes per plot and species richness is presented inTable D.2 for CO2 and in Table D.3 for nitrogen). For the grasses P. pratensis, B. inermis and A. gerardii wemeasured height from the ground to the tip of the longest leaf when fully extended upwards. For L. perennis, S. rigida, A. tuberosa and L. capitata the standing height of each plant was measured without extending theleaves upwards, i.e., the measurement was taken from the ground to the highest point with foliage that wasnot on a reproductive bolting stem.To measure leaf size, specific leaf area and leaf dry matter content we harvested a single leaf from sixseparate individuals in each plot. Some of the sixteen species plots did not contain six individuals of allfocal species; in this case a single leaf from all individuals present was collected, and more plots fromthe same treatment were sampled (exact number of leaves sampled per plot, biotic and abiotic treatmentare presented in Table D.2 and D.3). Individuals were selected that were not growing in the middle 0.5× 1 m quadrat, which is reserved for long-term percent cover estimates only. We selected young, fullyexpanded leaves growing in full sunlight. After cutting, the leaves were immediately wrapped in paper towelmoistened with deionized water. The leaves were then stored in a cool dark place for approximately 12 hoursbefore processing. Some preliminary trials showed that wrapping the leaves in moist paper towel resultedin more consistent rehydration than placing the ends of the leaves in vials of deionized water (the methodrecommended by Cornelissen et al. (2003)). After 12 hours of rehydration the leaves were removed from themoist paper towel, patted dry and weighed to obtain their wet mass. They were then immediately scannedand placed in a drying room held at 40◦C for a minimum of 7 days before weighing. Leaf area was calculatedusing image J software (http://rsbweb.nih.gov/ij/).We measured the average weight of an individual seed for legumes, and diaspores of the other species,note we use the term seed loosely to describe seeds and diaspores. We did not obtain any seeds for A. tuberosabecause many of the flower heads had been eaten by deer and thus obtaining seed pods with mature seedswas impossible for most plots. For the other species mature seeds were harvested from healthy adult plants(Pérez-Harguindeguy et al., 2013). For A. gerardii and S. rigida we obtained average seed mass by harvestingall seed heads originating from three randomly selected flower stalks per plot. For both P. pratensis and B.inermis all inflorescences originating from the same basal clump for three separate locations were taken and,for L. perennis we harvested three branches with mature but unopened pods from three separate individualsfrom each plot. To minimise sampling multiple seed heads or inflorescence from the same individual, wetook the closest seed head or inflorescence that was 50cm away (at a 45 degree angle) from the NE, NW and535.3. Statistical analysisSE corners. Once the seed heads were harvested they were dried at 40◦C for a minimum of 7 days. Thenthe seeds were separated from the rest of the inflorescence or pod and weighed in a manner that dependedon seed size. For A. gerardii, B. inermis, and P. pratesnsis, the mass of 30 mature seeds was determined toobtain an average individual seed mass. For S. rigida (very small seeds), the average individual mass of 50seeds weighed together was obtained. Lastly, for L. perennis and L. capitata, the seeds were dissected out ofthe pod, and all seeds were counted and weighed together to obtain an average individual seed weight. Seedmass was log transformed prior to analysis.5.3 Statistical analysisThere are two ways to test whether the functional group diversity of the community influences the responseof a species to elevated CO2 or elevated nitrogen. Firstly, we could run separate statistical analyses: one totest for differences between ambient and elevated CO2 in ambient nitrogen, and another to test for differencesbetween ambient and elevated nitrogen in ambient CO2. Alternatively, we could combine these tests in oneanalysis by including the interaction between functional group richness and abiotic treatment and performpost-hoc tests to determine which groups are significantly different to one another. We decided to analysethe data using the second method i.e. perform one analysis because this has fewer multiple comparisons.Analysing the data in the first way provides identical results (not shown).To test whether the response of plant traits to elevated CO2 or elevated nitrogen (eN) was affected bycommunity diversity, we performed a multivariate analysis on all the plant traits of each species separately.For this analysis, the unit of measurement was the plot level. We measured the height of many plants perplot (average = 16 per plot), but we were limited in how many leaves (average = 6) and flowers (average= 3) could be destructively sampled. Furthermore, the various traits were not all measured from the sameindividual plants. Thus, we consider the plot as our unit of sampling and calculated average values for eachtrait within a plot. The distributions of plot averages of each trait were then normalised to have a mean of zeroand a variance of one so that the scale of each trait was the same. These normalised values were then usedin PERMANOVA models, which are non-parametric multivariate analyses of variance using permutations totest significance (Anderson, 2001; McArdle and Anderson, 2001). Pseudo-F is calculated as the test statisticand has the same interpretation as the F-value of an ANOVA. We ran the PERMANOVA analyses in R withadonis2 in the vegan package, version 2.4-6 (Oksanen et al., 2018). Euclidean distance was used to calculatethe distance between points in trait space with Type III sums of squares, and 4999 permutations were used tocalculate the pseudo-F statistic. PERMANOVA tests the null hypothesis that there is no difference betweenthe multivariate centroids of the groups.Thus, to test whether plant traits respond more strongly to abiotic conditions in four functional groupplots than in single functional group plots, we tested the interaction between the abiotic treatment (CO2 ornitrogen) and functional richness. If the interaction was not significant, the interaction was removed, and themain effects of abiotic treatment and plot functional richness were tested. PERMANOVA models can alsoaccommodate blocking factors by constraining the permutations to occur within specified blocks. Althoughring is a blocking factor in this experiment, we did not include ring in the analysis for two reasons. Firstly,545.4. Resultssome rings only contain three plots and so there was limited scope for performing permutations using plots.Secondly, when the effect of ring was tested it was found to have no significant effect for any species or anytreatment. Lastly, to account for the inflation of Type I error rates due to multiple comparisons we performeda Bonferroni correction on the obtained P-values (Whitlock and Schluter, 2015).One assumption of PERMANOVA is that dispersion is equal among groups. We tested for homogeneityof group dispersions for each species and model term separately using the betadisper function in the packagevegan, version 2.4-6 (Oksanen et al., 2018). In other words, for each species we tested whether dispersionwas similar among environments (ambient versus eCO2 versus eN), between functional group numbers (oneversus four), or among the full set of environments and functional group numbers (allowing for interactions).Thus, in total seven species by three dispersion tests per species were performed. For all species and treat-ments, dispersion between groups after Bonferroni correction was not significant and this assumption wasmet in all cases as the smallest P-value obtained for any treatment and species combination was 0.47. Thebetadisper function was also used to visualise the PERMANOVA results by reducing the original distancesbetween points to principal coordinates (Ramette, 2007).To investigate how each trait individually responded to functional group richness, eCO2 and elevatednitrogen, we also performed linear mixed models on all species together for each trait. The normalizedtrait values were entered into models that included richness, CO2, nitrogen, the interaction between speciesrichness and CO2, and the interaction between species richness and nitrogen, as fixed factors. For each trait,we used z-scores by calculating the trait value as the number of standard deviations from the mean for thatspecies (across all treatments), which ensures that the data satisfy the assumption of mixed-effects modelsthat variances among groups (species) are equal. For each model, plot and plant species were also includedas random factors. We implemented all models using the restricted maximum-likelihood method (REML) toestimate variance components (Zuur et al., 2009). The significance of all fixed effects was evaluated usingType III estimable functions, and denominator degrees of freedom were determined using Satterthwaite’sapproximation (Satterthwaite, 1946). The fit of each final model was assessed through visual inspection ofthe fitted and residual values; all models were found to meet assumptions of normality and homogeneity(Zuur et al., 2009). Data analysis was carried out in R version 3.5.0 (R Core Team, 2018) using the lme4(Bates et al., 2015) and lmerTest (Kuznetsova et al., 2017) packages.5.4 ResultsWe found no significant interactions between abiotic treatment (eCO2 or elevated nitrogen) and functionalgroup richness for any of the species examined. This is in contradiction to our predictions for non-nitrogenfixing C3 and C4 species in elevated CO2 and nitrogen and for legumes in elevated CO2 but in accordancewith our predictions for legumes in elevated nitrogen (Table 5.1). In contrast, for most species we foundrelatively strong main effects of functional group richness and to a lesser extent abiotic treatment. Howspecies responded to these main effects is discussed below.We found trait responses to the functional group richness of the plot in five of the seven species examined(Table 5.1, Figure 5.1). Furthermore, the generalised linear mixed model analysis for each trait separately555.4. Results(but with all species simultaneously) revealed that across species, traits generally responded to functionalgroup richness, as indicated by the significant functional group richness term for all traits except seed weight(Table 5.2). In general, plants grew taller and produced leaves with higher SLA, lower LDMC and greaterleaf areas in more diverse plots (Figure 5.2). However, the nitrogen fixing legume L. capitata and the forbS. rigida were both exceptions to this trend with neither responding significantly to the functional groupdiversity of the plot (Table 5.1 and Figure 5.1).Three species, A. gerardii, L. perennis, and S. rigida responded significantly via trait changes to theabiotic environment while B. inermis had a marginally significant response (Figure 5.1). Post-hoc testsrevealed that for A. gerardii the significant response to the abiotic environment was driven by the differ-ence in multivariate trait space between ambient nitrogen and elevated nitrogen in ambient CO2 conditions(P=0.006). For L. perennis the significant trait response to the abiotic environment was driven by a signifi-cant difference between ambient and elevated CO2 (P=0.042). Similarly, the significant effect of the abioticenvironment on S. rigida was in part driven by the difference in trait values between ambient and elevatedCO2 (P=0.003), although the difference between trait values in elevated CO2 and elevated nitrogen was alsosignificant (P=0.003). Lastly, the marginally significant effect of abiotic environment on the trait values ofB. inermis was driven by a significant difference between elevated CO2 and elevated nitrogen (P=0.015).When all species were analysed together using a generalised linear mixed model to test for the effect ofthe abiotic environment on individual traits we found a significant effect of nitrogen on plant height, SLA,and LDMC but not for leaf area or seed weight (Table 5.2 and Figure 5.2). Overall, increased nitrogen causedspecies to grow taller and to produce leaves with higher SLA values and lower LDMC, as would be expectedfor resource rich environments (Figure 5.2, D.1 and D.2). In contrast, elevated CO2 resulted in changes inSLA and marginal changes in LDMC in most species (Table 5.2 and Figure 5.2). Plants tended to have lowerSLA values and higher LDMC values in eCO2 (Figure 5.2). The opposite trait responses to eN and eCO2 isconsistent with the hypothesis that increased growth under eCO2 lead to nitrogen limitation.Lastly, classifying plots by species richness instead of functional group richness did not quantitativelyalter the results for most species (Sup. Mat D.1.1). The response of species to functional group diversity wassimilar to that of species richness, given that similar responses were found for monoculture vs. four-species,one functional group plots and for four-species four functional group vs. 16-species four functional groupplots (Table D.4 and Figure 5.2).565.4. ResultsTable 5.1: PERMANOVA results based on Euclidean distances of normalised trait data for each speciesgrown in plots containing species of one versus four functional groups (FG) in ambient or elevated CO2or nitrogen. Bold text indicates significant differences (P<0.05, after Bonferroni correction) and † indicatesmarginally significant results (P<0.1 after Bonferroni correction)Species Functionalgroup mem-bershipAbiotic treatment FG richness Abiotic treatment × FG richnessR2 F-value R2 F-value R2 F-valueAndropogon gerardii C4 grass 0.14 3.49* 0.18 8.77** 0.07 1.67Bromis inermis C3 grass 0.13 3.09† 0.18 8.55** 0.03 0.60Poa pratensis C3 grass 0.09 2.46 0.29 16.46** 0.07 2.09Lupinus perennis Legume 0.15 3.95* 0.21 10.66** 0.02 0.56Lespedeza capitata Legume 0.09 1.61 0.08 2.83 0.03 0.59Asclepias tuberosa Forb 0.04 0.82 0.14 5.74* 0.05 0.97Solidago rigida Forb 0.20 4.94** 0.03 1.29 0.08 2.0ns P > 0.1; † P < 0.1; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001Table 5.2: Mixed effect model results for the response of plant height, SLA, LDMC, leaf area and seedweight of all species to the concentration of CO2 and nitrogen, and to the functional group (FG) richnessof the plot in which plants were grown. All data were normalised per species to a mean of zero and avariance of one. Linear mixed models were run separately for each trait and included the terms functionalgroup richness, CO2, and nitrogen as single factors, and CO2 × functional group richness, and nitrogen ×functional group richness as interactions. Plant species identity and plot were included as random effects.F-tests for fixed effects were constructed in R and denominator degrees of freedom (dfD) were obtained fromthe Satterthwaite approximation. Significance of random effects was determined by likelihood ratio tests.Plant height SLA LDMC Leaf area Seed weightdf dfD F-value dfD F-value dfD F-value dfD F-valuedfD F-valueCO2 1 114.3 0.54 101.49 6.51* 92.07 3.53† 82.15 2.29 110.30 2.38Nitrogen 1 114.9 9.25** 106.0 5.19* 95.04 4.06* 83.83 1.18 105.70 3.00†FG richness 1 115.5 40.44**** 104.94 27.97*** 94.01 13.48*** 82.96 8.68** 110.55 0.96CO2 × FG richness 1 114.45 1.04 101.49 2.66 92.07 0.32 82.15 0.22 110.30. 0.19Nitrogen× FG richness 1 114.88 1.60 106.0 0.02 95.04 0.001 83.83 0.30 105.70 0.15Random effects n P P P PPlot 86 <0.0001**** <0.0001**** <0.0001**** <0.0001**** <0.0001****Plant species 7 <0.0001**** <0.0001**** <0.0001**** <0.0001**** <0.0001****ns P > 0.1; † P < 0.1; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001575.4. ResultsH (44%)LA (37%)SLA (45%)LDMC (34%)−2−1012−2 −1 0 1 2PCoA1 (38%)PCoA2 (24%)Andropogon gerardii ALDMC (37%)SLA (35%)SW (22%)H (51%)LA (26%−2−1012−2 0 2 4PCoA1 (41%)PCoA2 (25%)Bromus inermis CSLA (26%)H (25%)LDMC (24%)LA (23%)SW (94%−2−1012−4 −2 0 2 4PCoA1 (61%)PCoA2 (19%)Poa pratensis E−2−1012−2 −1 0 1 2PCoA1PCoA2 B−2−1012−2 0 2 4PCoA1PCoA2 D−2−1012−4 −2 0 2 4PCoA1PCoA2 FFunctional group 1fg 4fg abiotic treat aC & aN eC & aN aC & eNFigure 5.1: Continued next page585.4. ResultsSW (33%)H (25%)LDMC (43)SLA (35%)−3−2−1012−2 0 2PCoA1 (35%)PCoA2 (32)Lespedeza captiata GSLA (33%)LDMC (32%)H (22%)LA (58%)SW (33%−2−1012−4 −2 0 2PCoA1 (49%)PCoA2 (24%)Lupinus perennis ISLA (35%)LDMC (35%)LA (25%)H (54%)LA (21%−2−1012−4 −2 0 2PCoA1 (50%)PCoA2 (33%)Asclepias tuberosa KSLA (39%)LDMC (32%)LA (63%)H (24%)−3−2−1012−2 0 2PCoA1 (39%)PCoA2 (24%)Solidago rigida M−3−2−1012−2 0 2PCoA1PCoA2 H−2−1012−4 −2 0 2PCoA1PCoA2 J−2−1012−4 −2 0 2PCoA1PCoA2 L−3−2−1012−2 0 2PCoA1PCoA2 NFunctional group 1fg 4fg Abiotic treat aC & aN eC & aN aC & eNFigure 5.1: Continued next page595.5. DiscussionFigure 5.1: Principal coordinate analysis of trait responses of Andropogon gerardii (C4 grass) (A, B), Bromusinermis (C3 grass) (C, D), and Poa pratensis (C3 grass) (E, F), Lespedeza capitata (legume) (G, H), Lupinusperennis (legume) (I, J), Asclepias tuberosa (forb) (K, L), and Solidago rigida (forb) (M, N) to the abioticenvironment (ambient and elevated CO2 and nitrogen) in plots with one or four functional groups. For eachspecies the data shown in each pair of panels is the same with panels A, C, E, G, I, K, and M showingthe data subset according to functional group diversity and panels B, D, F, H, J, L, and N showing the datasubset according to abiotic treatment. Figures with a grey background indicate significant results (P<0.05,after Bonferroni correction) from the PERMANOVA analysis, see Table 5.1 for more details. Ordinationswere generated with the output of PCoA, with the percent variation explained by each PCoA axis presentedin parentheses in the axis label for the first plot for functional group richness (Panels A, C, E, G, I, K, andM ). These panels also highlight the plant traits that contributed more than 20% to a PCoA axis. Black textindicates a positive association with that PCoA axis while grey text indicates a negative association. H =plant height; LA = leaf area; LDMC = leaf dry matter content; SLA = specific leaf area, and SW = seedweight5.5 DiscussionChanges in the biotic and abiotic environment are occurring due to climate and land-use change, nutrientdeposition, and species introductions (Canadell et al., 2007; Galloway et al., 2008; McGeoch et al., 2010;Urban, 2015; Vitousek et al., 1997). These changes result in different selection pressures in different contextsso that species adapt via changes in biomass and reproduction differently to abiotic conditions when grownin communities of low and high diversity (Kleynhans et al., 2016; Lau et al., 2014). However, whether theresponse of plants in their functional traits (such as changes in plant height or SLA) to the abiotic environmentis mediated by the biotic environment has not previously been documented under natural conditions. Herewe tested whether plant functional traits of species responded to abiotic change differently in communitiesconsisting of a single or four functional groups. In contrast to our predictions, we found no difference inhow species responded to abiotic conditions as a result of changes in the biotic community. Instead, for mostspecies we found strong plant trait responses to the main effects of functional group richness and to a lesserextent CO2 and nitrogen. Each of these points is discussed in further detail below.5.5.1 Community context has no detectable impact on how functional traits respond toabiotic changeOur hypothesis that the trait responses of plants to elevated CO2 and nitrogen would be mediated by commu-nity diversity and composition was not supported. For all species, the interaction between functional grouprichness and abiotic treatment was not significant. Furthermore, the interaction with species richness insteadof functional group richness was also not significant (Supp Mat D.1.1 and D.4). These results are in linewith those from various short-term pot experiments, where a focal species was grown with an con or het-erospecific competitor in ambient or elevated CO2 and functional traits such as plant height and SLA weremeasured (e.g. Carter and Peterson, 1983; Wray and Strain, 1987). Although these studies typically foundsignificant main effects of CO2 and competitor identity, they did not find a significant interaction between thetwo. In contrast, studies examining changes in biomass and reproduction, where short term plastic responses605.5. DiscussionaC & aN eC & aN aC & eN−101Plant heightAaC & aN eC & aN aC & eN−0.50−0.250.000.250.50SLACaC & aN eC & aN aC & eN−0.30.00.30.6LDMCEaC & aN eC & aN aC & eN−0.50.00.5Leaf areaGaC & aN eC & aN aC & eN1 4 1 4 1 42.252.302.35Functional group diversitySeed weightI−101B−0.50−0.250.000.250.50D−0.30.00.30.6F−0.50.00.5H2.252.302.35aC & aN eC & aN aC & eNAbiotic treatmentJFigure 5.2: The response of plant height, specific leaf area (SLA), leaf dry matter content (LDMC), leaf area,and logarithm of the seed weight to the number of species present in a plot (Panel A, C, E, G, I), or to the615.5. DiscussionFigure 5.2: cont, CO2 and nitrogen environments (Panel B, D, F, H, J). Generalised linear mixed modelswere performed including the single effects of CO2, nitrogen and species richness as fixed factors; and plantspecies and plot as random factors. Predicted values from the model output are plotted to better highlightthe general effect of species richness, CO2 and nitrogen on plant traits. The median of the predicted valuesis represented by the thick horizontal line, notches extend 1.58 × the inter-quartiles range / square root ofthe sample size. The notch corresponds roughly to the 95% confidence interval for comparing medians. Thehinge (upper and lower edges of the box) represent the 25th and 75th percentiles of the data with whiskersextending from the hinge to the smallest/largest value that is at most 1.5 × the inter-quartile range. Pointsbeyond the end of the whiskers are considered outliers.(Arp et al., 1993; Lüscher et al., 1996; Wedin and Tilman, 1993) and longer term genetic changes (Kleyn-hans et al., 2016; Lau et al., 2014) were observed, found significant abiotic treatment by species richnesseffects. Why our long-term study found no abiotic treatment by community diversity effect is not entirelyclear but there are several possibilities. Firstly, biomass is a holistic trait, with many underlying components,including plant height and leaf traits, thus the traits we measured could vary less across environments thanbiomass, given responses of other unmeasured traits such as number of leaves. Another possibility is thatin many long term experiments the responses in functional traits of species to elevated CO2 and nitrogenare largest in the first few years, and then return to values similar to those originally measured, presumablybecause plants acclimate or adapt to the local conditions (Strauss et al., 2008). Such a response has beenobserved in BioCON for a number of leaf traits including SLA, photosynthetic rate and water use efficiency(Crous et al., 2010; Lee et al., 2001). If plants acclimate or adapt to the local conditions (returning to nearoptimal functional trait values) then relative changes in trait values between ambient and elevated CO2 andnitrogen in different diversity treatments might be small and difficult to detect. It would be very interesting toexamine these traits at various intervals across the duration of BioCON to examine whether the effect sizeswere displaced at the start of the experiment and have re-equilibrated over time.Kleynhans et al. (2016) performed a transplant experiment, shifting individuals evolved in ambient andelevated CO2 in communities of low and high diversity into all combinations of CO2 and diversity, to in-vestigate the adaptation of plants to the biotic and abiotic conditions. Kleynhans et al. (2016) found thatthe biological community shaped the nature of selection so that P. pratensis adapted to eCO2 differentlyin different community contexts. More specifically, P. pratensis decline in biomass and reproduction whenplanted out of the treatment in which they had evolved, i.e., a reduction in performance due to maladapta-tion. The results from Kleynhans et al. (2016) however are not directly comparable with the results of thepresent study because individuals measured in the current study were not transplanted but rather measuredin situ where they had been growing for 16 years. Thus, a more direct comparison to make with the presentstudy is to look for an interaction between the biotic and abiotic environment in individuals transplanted intothe same environment as they had evolved in. An analysis of this subset of the data from Kleynhans et al.(2016) revealed no interaction between CO2 and diversity environments for aboveground (P=0.57), below-ground (P=0.92), or total biomass (P=0.77). This suggests that shifting environments may initially decreasefitness but over time adaptation likely increases fitness to pre-trasplantation levels. Similar results have pre-viously been found in BioCON. In a study conducted after five years of experimental treatments in BioCON625.5. DiscussionL. perennis was found to have adapted to ambient and elevated CO2 conditions, but when individuals fromboth treatments were planted into eCO2, those previously adapted to eCO2 increased in biomass by 29%while those with no prior exposure to eCO2 increased by 55% suggesting that adaptation had reduced theimmediate effect size of eCO2 (Strauss et al., 2008). If over time plant trait responses to elevated CO2 ornitrogen decline, due to presumably some physiological acclimation or adaptation, then this might providean explanation for why we did not find a significant interaction between the biotic and abiotic environmentfor these functional traits. Instead we found stronger main effects of diversity and the abiotic environment.5.5.2 Trait responses to functional group richnessAcross most species and contexts, we found strong trait responses to the functional diversity of the commu-nity. Communities of high diversity or functional group richness differ from those with a single species orlow functional group richness in large part because increased functional and species richness leads to denserneighbourhoods of plants (Marquard et al., 2009), which in turn leads to increased competition for light,space (Lorentzen et al., 2008) and nutrients (Oelmann et al., 2011). Furthermore, more diverse communitiesalso tend to have more interactions with other species such as herbivores, pollinators and pathogens (Scherberet al., 2010). Thus the growing environment for a plant is rather different between communities of low andhigh biodiversity. In terms of aboveground plant traits, previous work has shown that plants grown in morediverse communities typically are taller, have fewer branches and have leaves with higher SLA (Weisseret al., 2017). This is in agreement with our findings, where five out of the seven species significantly alteredtraits in response to plot diversity. These five species in general grew taller, and had higher SLA, lowerLDMC, and larger leaf areas in more functionally diverse plots than in single functional group plots. Twospecies, S. rigida and L. capitata, however, did not respond to functional group richness in the traits exam-ined. One potential reason for this is that in comparison to the other BioCON species, species-level averageheight of S. rigida and L. capitata is quite high (for L. capitata its over one meter). Because the traits wemeasured are most related to light acquisition, it is possible that these two species experienced no selectionto alter plant height or leaf traits to compensate for changes in the light environment. Similar results werefound in forbs by (Lipowsky et al., 2015) and legumes by (Thein et al., 2008): taller species grown in diverseplots did not respond in SLA or height as much as shorter species. Lastly, changes in biodiversity have beenfound to have at least as great an impact on productivity as other human-caused drivers such as drought, fire,herbivory, nitrogen, and CO2 (Tilman et al., 2012). In accordance with this we found changes in biodiversityto strongly impact functional traits of plants as much, if not more so, than changes in CO2 or nitrogen.5.5.3 Trait responses to the abiotic environmentWe found significant main effects of CO2 or nitrogen in three of the seven species studied. The legumeL. perennis and the forb species S. rigida responded to eCO2, while the C4 grass A. gerardii responded toelevated nitrogen. In general, legumes are expected to respond strongly to eCO2 because they are better ableto capitalise on increased CO2 given that they are less nitrogen limited than other species (Zanetti et al.,1996). However, why L. capitata did not respond to eCO2 while L. perennis did is not clear. Previousstudies have found species specific chemical compositions (Novotny et al., 2007) and nitrogen fixation rates635.5. Discussion(West et al., 2005) in legumes, however in general, it is L. capitata that responds more strongly to eCO2 inthese traits than L. perennis. S. rigida responded to eCO2 more strongly than the other treatments. A previousstudy found reduced leaf disease and consequently improved photosynthesis in eCO2 in S. rigida in BioCON(Strengbom and Reich, 2006), thus protection from disease under eCO2 may improve the performance of thisforb. Lastly, A. gerardii responded most strongly to elevated nitrogen. In nitrogen poor environments - thebaseline condition at the BioCON field site - mycorrhizae and A. gerardii have been shown to compete fornutrients. However, under nitrogen addition, colonisation of arbuscular mycorrhiza in the roots of A. gerardiiincreased and nutritional and growth benefits to the plant were observed (Püschel et al., 2016). Furthermore,Lee et al. (2011) found that A. gerardii had amongst the largest increases in net photosynthesis as a result ofgrowing in elevated nitrogen. Thus previous work has documented a significant effect of elevated nitrogenon A. gerardii.Averaged over all plant species, only SLA responded significantly to eCO2, with plants producing leaveswith lower SLA. A reduction in SLA under eCO2 is one response that is commonly observed (reviewed inBazzaz, 1990). Averaged across all plants, elevated nitrogen caused plants to generally grow taller and toproduce thinner larger leaves, again a commonly observed response of plants to nutrient rich environments.Overall, for most plant functional traits in most species, we found weaker trait responses to abioticenvironmental change than to functional group richness. However, we did not find that plants respond toelevated CO2 or nitrogen differently depending on the community in which they were living. As such, wehave no evidence that community context will strongly alter trait responses to elevated CO2 or nitrogen.64Chapter 6Conclusions6.1 OutlineIn this thesis I have addressed questions about how interactions between species might alter responses toabiotic environmental change. In the following sections I briefly recap the major conclusions of each chapter,mention some caveats and, highlight some interesting future directions. I conclude by highlighting where Ithink future research should be focused to move this field forward.6.2 Conclusions and future directions6.2.1 Chapter 2: Evolutionary rescue when resources change.In Chapter 2, I investigated the impact that intraspecific competition has on the shape of the G-matrix andhow this influences the probability of evolutionary rescue in two trait dimensions. Competition for resourcescreates diversifying selection so that different phenotypes exploit different portions of a resource, therebypotentially altering the shape of the G-matrix. Furthermore, empirical studies on the evolution of drugresistant pathogens, parasites, and cancer cells have all shown that intraspecific competition may play an im-portant role in the establishment of drug resistant types (Read et al., 2011; Huijben et al., 2013; Pena-Milleret al., 2013). Yet theoretical models on the topic (such as Chevin, 2013; Gomulkiewicz and Houle, 2009;Matuszewski et al., 2015) typically ignored intraspecific competition, assuming that populations nearingextinction experience little competition. In this chapter, I investigated the consequence of intraspecific com-petition by developing an individual-based multilocus model that tracked traits and competition for resourcesalong the same axes. For comparison, I also developed a model identical to Gomulkiewicz and Houle (2009)that excluded competition. I then simulated population growth when the resource optimum was shifted in asingle abrupt step.Overall I found that diversifying selection from intraspecific competition played a significant role inshaping the G-matrix. Instead of the G-matrix being shaped by mutation-selection balance, as previouslyfound (Arnold et al., 2008), it now rather reflected the availability of resources. This in turn impactedevolutionary rescue because instead of standing genetic variation being a good predictor of evolutionaryrescue in all trait directions, de novo mutation better explained rescue in directions that mutations couldexplore but where resource availability is low. We conclude that with respect to evolutionary rescue, the“genetic lines of least resistance” are sometimes shaped by the standing genetic variation but sometimes bythe distribution of de novo mutations in populations competing for limiting resources.656.2. Conclusions and future directionsCaveatsIn Chapter 2 the ecological and genetic models are unfortunately different from one another. For example,the fitness function in the ecological model is density dependent while in the genetic model it is not. Ide-ally I would have a model where I could vary the amount of diversifying selection while holding all otherparameters constant. This would have allowed a more elegant investigation of the impact of intraspecificcompetition on the shape of the G-matrix and evolutionary rescue. Indeed I attempted to do this by makingthe resource distribution very tall and narrow and the competition kernel very broad. Our intuition was thatindividuals with trait values different to the optimum would still be able to survive because they would stillhave access to the resources at the optimum. Thus the distribution of trait values in the population should bemore similar to that of a model with no competition. Then by varying the ratio of the width of the competi-tion kernel and resource kernel we could alter the amount of diversifying selection. However, simulations ofthe extreme scenario only resulted in all individuals being pilled up at the resource optimum which was notwhat we wanted. Thus in the end we turned back to examining these two separate models. Overall, howeverI think the conclusion that competition might be important in shaping standing genetic variation is likelyrobust, and should be considered when determining genetic lines of least resistance.Unlike the rest of my thesis, this chapter ignored interspecific interactions and rather only focused onintraspecific interactions. Nevertheless, I do not think that including other competing species would alterthe results substantially. The main expected change would be the shape of the resource landscape andconsequently, the shape of the G-matrix would be altered to match available resources. There is no reasonto expect that this would alter our general conclusion of standing genetic variation improving the chance ofevolutionary rescue in directions of low mutational variance but abundant resources and de novo mutationsaiding rescue in directions of larger mutational variation but few resources.Lastly, the results from this chapter will only apply to traits related to resource acquisition, as competitionfor resources drives diversifying selection. In contrast many studies investigating genetic variance covariancematrices examine traits related to life history or sexual selection. Under these circumstances results shouldbe more similar to those made by previous theoretical studies such as that found by Gomulkiewicz and Houle(2009)Future directionsThe predictive ability of the G-matrix in determining the trajectory of evolution depends, to a large degree,on its stability. Yet as Lande (1979) noted the G-matrix is bound to vary over time in a population of finitesize due to drift. This is especially true for small populations, that do not experience correlational selection orpleiotropic mutation, and when selection and mutation are the only factors influencing its stability (reviewedin Arnold et al., 2008). One observation I made while running my simulations was that diversifying selectionseemed to increase the stability of the G-matrix, even at small population sizes because selection favouredindividuals with extreme trait values exploiting under-utilised resources. If future studies indicate that thisis generally true, then this would lend support to the predictive ability of the G-matrix for traits related toresource acquisition.666.2. Conclusions and future directions6.2.2 Chapter 3: Asymmetric competition impacts evolutionary rescue in a changingenvironmentIn Chapter 3 we investigated how two competing species adapted to abiotic change when they differed ininitial population size or competitive ability. Previously, Johansson (2008) showed that when two speciescompete for a resource with a directionally shifting distribution, the species lagging behind the resourcepeak is the first to go extinct due to competitive exclusion. However, this work assumed symmetricallydistributed resources and competition. Asymmetries can generate differences between species in populationsizes, genetic variation, and trait means. Furthermore, species competing with one another typically do nothave perfectly symmetrical competitive abilities (Connell, 1983; Weiner, 1990) or population sizes (e.g Jonesand Barmuta, 1998), nor are the resources for which they are competing perfectly symmetrical (e.g. Boagand Grant, 1984; Kleynhans et al., 2011). Thus, in Chapter 3 we investigated the impact of asymmetries onadaptation to abiotic change. We developed an individual-based model and tracked the population dynamicsof two species adapting to a continuously shifting environment. Overall we showed that asymmetric resourceavailability or competition could facilitate coexistence and even occasionally caused the leading species to goextinct first. Surprisingly, we also found cases where traits evolved in the opposite direction to the changingenvironment because of a “vacuum of competitive release” created when the lagging species’ populationdeclined in size. Thus, the species exhibiting the slowest rate of trait evolution was not always the mostlikely to go extinct in a changing environment. Our results demonstrated that the extent to which speciesappeared to be tracking environmental change and the extent to which they were pre-adapted to that changemay not necessarily determine which species would be the winners and which would be the losers in arapidly changing world.Future directionsMathematical models are essentially quantitative thought experiments whose results need to be tested in em-pirical systems to determine their applicability. Thus, empirically investigating the possibility of a “vacuumof competitive release” i.e. the evolution of a species in a direction opposite to that predicted by environmen-tal change due to the rapid decline in abundance of a competitor species, is an important future research direc-tion. In Chapter 3 we highlight a few natural systems where one might expect to observe this phenomenon,but I think the most promising short-term research avenue would be to perform a laboratory experiment in amodel system with high level of control. For example, one could use the algae Asterionella and Cyclotella(studied by (Tilman, 1977)) that compete for different ratios of phosphorus and silicate and induce stablecoexistence of the two species with Asterionella being initially rare. Then one could initiate environmentalchange by reducing the concentration of phosphorus with the expectation that Asterionella numbers wouldincrease, because it is the stronger competitor under phosphate-limited conditions. However, its ability toutilise limited phosphorus would initially decrease due to competitive release from Cyclotella as Cyclotelladeclines in abundance.676.2. Conclusions and future directions6.2.3 Chapter 4: Adaptation to elevated CO2 in different biodiversity contextsIn Chapter 4 we experimentally investigated the impact of community diversity on adaptation to abioticchange. The diversity of a plant’s neighbourhood can influence many factors including its population size,competitive ability, genetic diversity and/or the direction and form of natural selection. But how this mightinfluence adaptation to an abiotic change, such as elevated CO2, has not previously been explored. We testedthis ideas in Chapter 4 through a full reciprocal transplant experiment performed in BioCON (BiodiversityCarbon dioxide, Nitrogen experiment) at the Cedar Creek Ecosystem Science reserve. In 2011 we sampledPoa pratensis seeds in monoculture and 16-species plots in ambient and elevated CO2. We then raised theseseeds in a common garden environment and sampled rhizomes from each individual. These rhizomes werethen planted back into BioCON in plots consisting of only Poa pratensis or into a diverse community ofspecies in ambient and elevated CO2. By assessing changes in biomass and flowering we could test each ofthe following hypotheses:1. Species diversity has no impact on adaptation to abiotic change. This would be expected if the effectof elevated CO2 was much larger than the effect of the plant neightbourhood on fitness.2. Species richness might decrease adaptation to abiotic change. This would be expected if competitionfor space or food meant that each species has a smaller population size and so less genetic diversity.3. Species richness might increase adaptation to abiotic change. This might occur if neighbourhooddiversity increased genetic diversity because individuals from the same population experience a wider varietyof micro-environmental conditions.4. Species diversity might alter the fitness landscape. This would be expected if species diversity modi-fied the selective environment created by the abiotic change.Overall, we found support for hyptothesis 4. Specifically, we found evidence for local adaptation toelevated CO2, but only for plants assayed in a community of similar diversity to the one experienced duringthe period of selection. These findings have important implications at both small and large spatial scales. Ona local scale, communities can vary a great deal due to slope, aspect, etc. Thus, evolution of the metapopula-tion might be hampered if the selective pressure from an abiotic change is experienced differently in differentmicro-sites and there is high gene flow. At a larger scale, even if species can migrate (due to climate change)to track their preferred abiotic conditions, they may still experience a decline in fitness as they will likelyfind themselves interacting with novel competitors and predators, that alter the nature of selection and lessenthe species adaptation to those abiotic conditions.CaveatsWe list a variety of caveats in the discussion of Chapter 4 but it is worth highlighting here the question ofwhether the results we observed are a result of transgenerational (maternal) effects or whether they are truegenetic adaptation. Although we tried to remove maternal effects by raising the seeds in the greenhouse andsampling rhizomes we cannot be 100% sure that these effects were fully removed. Whether our results aredue to maternal effects or genetic change, they are still fascinating and have important implications for ourunderstanding how species might adapt to climate change in different biodiversity contexts. Nevertheless,686.2. Conclusions and future directionsconfirming these experimental results in other contexts that breed for longer in a common environment wouldgreatly strengthen these resultsFuture directionsCO2 is unlike other resources for plants (e.g., nutrients, water, light) as the uptake of CO2 by one plantis unlikely to influence the uptake by its neighbours and therefore species are unlikely to directly competefor CO2. Instead, CO2 likely alters the competitive interactions between species. Therefore, one importantquestion is whether similar results are found for other abiotic drivers of global change such as nitrogen.In a similar vein we do not know whether these results hold across species. As outlined in the introductionto Chapter 5 species differ substantially in their response to CO2 (and other abiotic drivers) and thereforewhether the same response would be seen in other species, particularly ones with different photosyntheticpathways (such as C4) or able to fix nitrogen (such as legumes), is an important line of research to assess thegenerality of our conclusions.Deeper insights into underlying mechanisms could be generated by investigating the effective populationsizes and genetic diversity of the populations growing in each of the CO2 and diversity treatments. Althoughwe found the most support for the fitness landscape hypothesis, the other scenarios might also be occurringand knowing the effective population size and genetic diversity of different Poa pratensis populations wouldhelp elucidate this.We also have little understanding of the mechanism of evolutionary change. Did it occur via selectionamong a few genotypes planted at the start of the experiment or has mutation occurred to allow these popula-tions to adapt? Although we cannot answer this question, future long-term experiments of this nature couldpermit test of evolutionary mechanisms by sampling genetic material at various times across the duration ofthe experimentLastly, if most species do not shift their distribution under climate change then communities will beadapting to abiotic change in situ. However, when a migrant arrives in this community, the migrant willbe competing with species that are adapted to one another but are under selection to adapt to the abioticconditions. Thus, does the fact that community is adapting to abiotic change reduce the strength of selectionthat the community imposes on a migrant or does it alter the selective landscape again in a different way?6.2.4 Chapter 5: How plant diversity influences intraspecific trait responses to abioticenvironmental changeIn Chapter 5 we investigated whether community context and increases in atmospheric carbon dioxide andnitrogen concentrations influenced plant functional traits. Plant growth generally responds positively to el-evated nitrogen and CO2 but this response is not uniform across functional groups. For example, legumesusually do not respond to nitrogen additions, and C4 grasses usually do not respond to elevated CO2. Thesedifferential responses of species to abiotic change could alter the balance of species within a communityand their competitive interactions. In turn, these differences in competitive abilities might influence thefunctional trait responses of species in different community and abiotic contexts so that for most species wemight expect to find a significant biodiversity-by-abiotic change interaction. Thus in Chapter 5 we tested696.3. Future directions and general conclusionswhether plants show consistent trait responses to elevated concentrations of CO2 and nitrogen when grownin communities that differ in plant functional richness (1 or 4 functional groups). As for Chapter 4, our workwas carried out in BioCON, where plant diversity, CO2, and nitrogen concentrations have been manipulatedfor the past 17 years. We measured plant height, specific leaf area, total leaf area, leaf dry matter content andseed masses of seven different species from four functional groups growing in plots that differed in func-tional richness, CO2 and nitrogen concentrations. In contrast to expectations, we did not find a significantbiodiversity-by-abiotic change interaction for any species. Instead we found strong functional trait responsesin most species to the plot diversity. In particular, most species grew taller and produced larger, thinner leavesin more diverse communities. We also found responses to the abiotic environment in some species but notothers and this response was not consistent within functional groups. Overall, we have no evidence thatcommunity context will strongly influence trait responses of species to abiotic change. Instead, communitycontext, on its own, seems to be a more important factor influencing the trait responses of species.CaveatsOne major caveat of this study stems from the fact that BioCON included few plots of some treatments inwhich any given species was planted: only two monoculture plots per species, two 4-species one functionalgroup plots, zero to three 4-species four functional group plots, and twelve 16-species plots per abiotictreatment. Thus, plots of low species richnesses suffered from low sample sizes. As such, we focused onresponses of species to functional group diversity. This comparison permits larger sample sizes per treatmentbut still not as large as one would include in an experiment designed specifically for our questions. In ouranalyses species diversity and functional group diversity were confounded because one functional groupplots contain either one or four species and four functional group plots contain either four or 16-species.That said, when the analysis was conducted on species richness only, the results were very similar.Future directionsOne intriguing possible explanation for the lack of a significant diversity-abiotic interaction is that plantsacclimated or adapted to the conditions so that over time their trait values returned to an original species-specific optimum. This possibility could be tested in future experiments by measuring functional traits ofspecies across the duration of an experiment like BioCON.6.3 Future directions and general conclusionsIn the introduction I outline the holes my thesis research attempts to fill. Thus, here I provide insights intowhat I think should be investigated next. I think one important and outstanding question is whether ratesof evolution to abiotic change might differ as a consequence of different community contexts. On the onehand rates of evolution might be reduced with more species interactions because population sizes are smaller(e.g. Johansson, 2008). On the other hand, rates of evolution might be increased if interactions provide aselective push (Chapter 3 and Osmond and de Mazancourt (2013); Osmond et al. (2017)). Then again, theeffects of species interactions on rates of adaptation might cancel out because of opposing selective forces.706.3. Future directions and general conclusionsFor example, a plants fitness may be impaired by a herbivore consuming its leaves, but augmented by myc-orrhizae increasing its tissue production (Haloin and Strauss, 2008). Alternatively competition may reducegrowth while elevated CO2 may enhance growth. Teasing out under what circumstances rates of evolu-tion are enhanced, reduced or unchanged would greatly inform questions related to evolutionary rescue andadaptation of species to abiotic environmental change. This research would not only have implications forunderstanding adaptation to climate change but also for the evolution of unwanted species, such as antibioticresistant bacteria or crop pests. Furthermore, with a proper understanding of what factors influence rates ofevolution we could target management practices or conditions to bolster or reduce rates of evolution (Legerand Espeland, 2010) in response to an abiotic change.Overall, my thesis has demonstrated that species interactions can potentially play a large role in influ-encing adaptation to abiotic environmental change. Not only do species interactions influence the amount ofstanding genetic variation that a population has (Chapter 2) but they also alter the trajectory of evolution inresponse to abiotic change over the short- and longer-term (Chapter 3 and Chapter 4).71BibliographyAdair, E. C., P. B. Reich, J. J. Trost, and S. E. Hobbie. 2011. Elevated CO2 stimulates grassland soilrespiration by increasing carbon inputs rather than by enhancing soil moisture. Global Change Biology17:3546–3563.Akerberg, E. 1939. Apomictic and sexual seed formation in Poa pratensis. Hereditas pages 359–370.Alexander, H. K., and S. Bonhoeffer. 2012. Pre-existence and emergence of drug resistance in a generalizedmodel of intra-host viral dynamics. Epidemics 4:187–202.Alexander, J. M., J. M. Diez, and J. M. Levine. 2015. Novel competitors shape species’ responses to climatechange. Nature 525:515–518.Anderson, M. J. 2001. A new method for non-parametric multivariate analysis of variance. Austral Ecology26:32–46.Antonovics, J. 1971. The effects of a heterogeneous environment on the genetics of natural populations.American Scientist 59:593–599.Arnold, S. J. 1992. Constraints on phenotypic evolution. The American Naturalist 140:S85–s107.Arnold, S. J., R. Bürger, P. A. Hohenlohe, B. C. Ajie, and A. G. Jones. 2008. Understanding the evolutionand stability of the g-matrix. Evolution 62:2451–2461.Arp, W. J., B. G. Drake, W. T. Pockman, C. P. S, and W. D. F. 1993. Interactions between C3 and C4 saltmarsh plant species during four years of exposure to elevated atmospheric CO2. Vegetatio 104/105:133–143.Ashmole, N. P. 1968. Body size, prey size, and ecological segregation in five sympatric tropical terns (aves:Laridae). Systematic Zoology 17:292–304.Avolio, M. L., S. E. Koerner, K. J. La Pierre, K. R. Wilcox, G. W. T. Wilson, M. D. Smith, and S. L. Collins.2014. Changes in plant community composition, not diversity, during a decade of nitrogen and phosphorusadditions drive above-ground productivity in a tallgrass prairie. Journal of Ecology 102:1649–1660.Bailey, S. F., J. R. Dettman, P. B. Rainey, and R. Kassen. 2013. Competition both drives and impedesdiversification in a model adaptive radiation. Proceedings of the Royal Society B. 280:1–8.72BibliographyBaptestini, E. M., M. A. de Aguiar, D. I. Bolnick, and M. S. Araujo. 2009. The shape of the competitionand carrying capacity kernels affects the likelihood of disruptive selection. Journal of Theoretical Biology259:5–11.Barraclough, T. G. 2015. How do species interactions affect evolutionary dynamics across whole communi-ties? Annual Review of Ecology, Evolution, and Systematics 46:25–48.Barrett, R. D., and D. Schluter. 2008. Adaptation from standing genetic variation. Trends in Ecology andEvolution 23:38–44.Bates, D., M. Mächler, B. Bolker, and S. Walker. 2015. Fitting linear mixed-effects models using lme4.Journal of Statistical Software 67:1–48.Bates, D., M. Maechler, B. M. Bolker, and S. Walker. 2014. lme4: Linear mixed-effects models using Eigenand S4.Bazzaz, F. A. 1990. The response of natural ecosystems to the rising global CO2 levels. Annual Review ofEcology and Systematics 21:167–196.Bazzaz, F. A., and K. Garbutt. 1988. The response of annuals in competitive neighbourhoods: effects ofelevated CO2. Ecology 69:937–946.Begin, M., and D. A. Roff. 2004. From micro- to macroevolution through quantitative genetic variation:positive evidence from field crickets. Evolution 58:2287–2304.Bell, G. 2013. Evolutionary rescue and the limits of adaptation. Philosophical Transactions of the RoyalSociety B: Biological Sciences 368:1–6.Bell, G., and A. Gonzalez. 2009. Evolutionary rescue can prevent extinction following environmentalchange. Ecology Letters 12:942–948.Benkman, C. W., T. L. Parchman, A. Favis, and A. M. Siepielski. 2003. Reciprocal selection causes acoevolutionary arms race between crossbills and lodgepole pine. The American Naturalist 162:182–194.Berner, D., D. C. Adams, A. C. Grandchamp, and A. P. Hendry. 2008. Natural selection drives patterns oflake-stream divergence in stickleback foraging morphology. J. Evol. Biol 21:1653–1665.Berner, D., W. E. Stutz, and D. I. Bolnick. 2010. Foraging trait (co)variances in stickleback evolve deter-ministically and do not predict trajectories of adaptive diversification. Evolution 64:2265–2277.Björklund, M., A. Husby, and L. Gustafsson. 2013. Rapid and unpredictable changes of the g-matrix in anatural bird population over 25 years. Journal of Evolutionary Biology 26:1–13.Blows, M. W., and A. A. Hoffmann. 2005. A reassessment of genetic limits to evolutionary change. Ecology86:1371–1384.73BibliographyBoag, P. T., and P. R. Grant. 1984. Darwin’s finches (geospiza) on isla daphne major, galapagos: Breedingand feeding ecology in a climatically variable environment. Ecological Monographs 54:463–489.Bocedi, G., K. E. Atkins, J. Liao, R. C. Henry, J. M. J. Travis, and J. J. Hellmann. 2013. Effects of localadaptation and interspecific competition on species’ responses to climate change. Annals of the New YorkAcademy of Sciences 1297:83–97.Bolnick, D. I. 2001. Intraspecific competition favours niche width expansion in drosophila melanogaster.Nature 410:463–466.Bonduriansky, R., A. J. Crean, and T. Day. 2012. The implications of nongenetic inheritance for evolutionin changing environments. Evolutionary Applications 5:192–201.Bonin, M. C., L. Boström-Einarsson, P. L. Munday, and G. P. Jones. 2015. The prevalence and importance ofcompetition among coral reef fishes. Annual Review of Ecology, Evolution, and Systematics 46:169–190.Booth, R. E., and J. P. Grime. 2003. Effects of genetic impoverishment on plant community diversity. Journalof Ecology 91:721–730.Bowes, G. 1993. Facing the inevitable: Plants and increasing atmospheric co2. Annu. Rev. Plant Physiol.Plant Mol. Biol 44:309–332.Bradshaw, A. D. 1991. Genostasis and the limits to evolution. Philosophical Transactions of the RoyalSociety B: Biological Sciences 333:289–305.Bradshaw, W. E., and C. M. Holzapfel. 2001. Genetic shift in photoperiodic response correlated with globalwarming. Proceedings of the National Academy of Sciences of the United States of America 98:14509–14511.———. 2006. Evolutionary response to rapid climate change. Science 312.Brooker, R. W. 2006. Plant-plant interactions and environmental change. New Phytologist 171:271–284.Brown, J. H., and G. A. Lieberman. 1973. Resource utilization and coexistence of seed-eating desert rodentsin sand dune habitats. Ecology 54:788–797.Burrows, M. T., D. S. Schoeman, A. J. Richardson, J. G. Molinos, A. Hoffmann, L. B. Buckley, P. J. Moore,C. J. Brown, J. F. Bruno, C. M. Duarte, et al. 2014. Geographical limits to species-range shifts aresuggested by climate velocity. Nature 507:492–495.Canadell, J. G., C. Le Quéré, M. R. Raupach, C. B. Field, E. T. Buitenhuis, P. Ciais, T. J. Conway, N. P.Gillett, R. a. Houghton, and G. Marland. 2007. Contributions to accelerating atmospheric CO2 growthfrom economic activity, carbon intensity, and efficiency of natural sinks. Proceedings of the NationalAcademy of Sciences of the United States of America 104:18866–18870.Carlson, S. M., C. J. Cunningham, and P. A. H. Westley. 2014. Evolutionary rescue in a changing world.Trends in Ecology and Evolution 29:521–530.74BibliographyCarter, D., and K. Peterson. 1983. Effects of co2-enriched atmosphere on the growth and competitive inter-action of a c3 and a c4 grass. Oecologia 58:188–193.Chen, I.-C., J. K. Hill, R. Ohlemuller, D. B. Roy, and C. D. Thomas. 2011. Rapid range shifts of speciesassociated with high levels of climate warming. Science 333.Chevin, L.-M. 2010. Adaptation, plasticity, and extinction in a changing environment: towards a predictivetheory. PLoS Biology 8:e1000357.———. 2013. Genetic constraints on adaptation to a changing environment. Evolution 67:708–721.Clutton-Brock, T., S. Albon, R. Gibson, and F. Guinness. 1979. The logical stag: Adaptive aspects of fightingin red deer (Cervus elaphus L.). Animal Behavior 27:211–225.Collins, S. 2011. Competition limits adaptation and productivity in a photosynthetic alga at elevated CO2.Proceedings of the Royal Society of London B: Biological Sciences 278:247–255.Connell, J. 1983. On the prevalence and relative importance of interspecific competition: evidence from fieldexperiments. American Naturalist 122:661–696.Cornelissen, J. H. C., S. Lavorel, E. Garnier, S. Díaz, N. Buchmann, D. E. Gurvich, P. B. Reich, H. T. Steege,H. D. Morgan, M. G. a. V. D. Heijden, J. G. Pausas, and H. Poorter. 2003. A handbook of protocols forstandardised and easy measurement of plant functional traits worldwide. Australian Journal of Botany51:335.Crous, K. Y., P. B. Reich, M. D. Hunter, and D. S. Ellsworth. 2010. Maintenance of leaf N controls thephotosynthetic CO2 response of grassland species exposed to 9 years of free-air CO2 enrichment. GlobalChange Biology 16:2076–2088.Davis, M. A., P. B. Reich, M. J. B. Knoll, L. Dooley, M. Hundtoft, and I. Attleson. 2007. Elevated atmo-spheric CO2: a nurse plant substitute for oak seedlings establishing in old fields. Global Change Biology13:2308–2316.de Mazancourt, C., E. Johnson, and T. G. Barraclough. 2008. Biodiversity inhibits species’ evolutionaryresponses to changing environments. Ecology Letters 11:380–388.Dhondt, A. A. 1977. Interspecific competition between great and blue tit. Nature 268:521–523.Dhondt, A. A., and F. Adriaensen. 1999. Experiments on competition between great and blue tit: Effects onblue tit reproductive success and population processes. Ostrich 70:39–48.Díaz, S., and M. Cabido. 2001. Vive la difference: plant functional diversity matters to ecosystem processes.Trends in Ecology & Evolution 16:646–655.Dieckmann, U., and M. Doebeli. 1999. On the origin of species by sympatric speciation. Nature 400:354–257.75BibliographyDieckmann, U., and R. Law. 1996. The dynamical theory of coevolution: a derivation from stochasticecological processes. Journal of mathematical biology 34:579–612.Doebeli, M. 2011. Adaptive Diversification. Princeton University Press.Doebeli, M., and U. Dieckmann. 2000. Evolutionary branching and sympatric speciation caused by differenttypes of ecological interactions. The American Naturalist 156:S77–S101.Doney, S. C., V. J. Fabry, R. A. Feely, and J. A. Kleypas. 2009. Ocean acidification: the other CO2 problem.Marine Science 1:169–192.Draghi, J. A., and M. C. Whitlock. 2012. Phenotypic plasticity facilitates mutational variance, geneticvariance, and evolvability along the major axis of environmental variation. Evolution 66:2891–2902.Drossel, B., and A. McKane. 1999. Ecological character displacement in quantitative genetic models. Journalof Theoretical Biology 196:363–376.Dunnell, K. L., and S. E. Travers. 2011. Shifts in the flowering phenology of the northern great plains:patterns over 100 years. American Journal of Botany 98:935–945.Durand, E., M. I. Tenaillon, C. Ridel, D. Coubriche, P. Jamin, S. Jouanne, A. Ressayre, A. Charcosset, andC. Dillmann. 2010. Standing variation and new mutations both contribute to a fast response to selectionfor flowering time in maize inbreds. BMC Evolutionary Biology 10:1–21.Early, R., B. A. Bradley, J. S. Dukes, J. J. Lawler, J. D. Olden, D. M. Blumenthal, P. Gonzalez, E. D.Grosholz, I. Ibañez, L. P. Miller, C. J. B. Sorte, and A. J. Tatem. 2016. Global threats from invasive alienspecies in the twenty-first century and national response capacities. Nature Communications 7:12485.Eisenhauer, N., S. Cesarz, R. Koller, K. Worm, and P. B. Reich. 2012. Global change belowground: impactsof elevated CO2, nitrogen, and summer drought on soil food webs and biodiversity. Global ChangeBiology 18:435–447.Eisenhauer, N., T. Dobies, S. Cesarz, S. E. Hobbie, R. J. Meyer, K. Worm, and P. B. Reich. 2013. Plantdiversity effects on soil food webs are stronger than those of elevated CO2 and N deposition in a long-term grassland experiment. Proceedings of the National Academy of Sciences of the United States ofAmerica 110:6889–94.Eroukhmanoff, F., and E. I. Svensson. 2011. Evolution and stability of the g-matrix during the colonizationof a novel environment. Journal of Evolutionary Biology 24:1363–1373.Falconer, D. S., and T. F. C. Mackay. 1996. Introduction to quantitative genetics. 4th ed. Longman GroupLtd.Falster, D. S., and M. Westoby. 2003. Plant height and evolutionary games. Trends in Ecology & Evolution18:337–343.76BibliographyFenchel, T., and L. Kofoed. 1976. Evidence for exploitative interspecific competition in mud snails (hydro-biidae). Oikos 27:367–376.Frankham, R. 1996. Relationship of genetic variation to population size in wildlife. Conservation Biology10:1500–1508.Futuyma, D. J. 2009. Evolution. 2nd ed. Sinauer Associates Inc., Massachusetts, USA.Futuyma, D. J., and M. Slatkin, eds. 1983. Coevolution. Sinauer Associates Inc., Massachusetts, USA.Galloway, J. N., A. R. Townsend, J. W. Erisman, M. Bekunda, Z. Cai, J. R. Freney, L. A. Martinelli, S. P.Seitzinger, and M. A. Sutton. 2008. Transformation of the Nitrogen Cycle: Recent trends, questions, andpotential solutions. Science 320:889–892.Gatenby, R. A. 2009. A change of strategy in the war on cancer. Nature 459:508–509.Gaudet, C. L., and P. A. Keddy. 1988. A comparative approach to predicting competitive ability from planttraits. Nature 334:242–243.Gause, G. 1934. The struggle for existence. The Williams and Wilkins Co.Geyer, C. J., C. E. Ridley, R. G. Latta, J. R. Etterson, and R. G. Shaw. 2013. Local adaptation and geneticeffects on fitness: Calculations for exponential family models with random effects. Annals of AppliedStatistics 7:1778–1795.Gomulkiewicz, R., and R. D. Holt. 1995. When does evolution by natural selection prevent extinction?Evolution 49:201–207.Gomulkiewicz, R., and D. Houle. 2009. Demographic and genetic constraints on evolution. The AmericanNaturalist 174:E218–E229.Gomulkiewicz, R., and R. G. Shaw. 2012. Evolutionary rescue beyond the models. Philosophical Transac-tions of the Royal Society B: Biological Sciences 368:20120093.Gonzalez, A., O. Ronce, R. Ferriere, and M. E. Hochberg. 2013. Evolutionary rescue: an emerging focusat the intersection between ecology and evolution. Philosophical Transactions of the Royal Society B:Biological Sciences 368:20120404.Gould, F., G. Kennedy, and M. Johnson. 1991. Effects of natural enemies on the rate of herbivore adaptationto resistant host plants. Entomologia Experimentalis et Applicata 58:1–14.Grant, P. R., and B. R. Grant. 2006. Evolution of character displacement in Darwin’s finches. Science313:224–226.Gubsch, M., N. Buchmann, B. Schmid, E.-D. Schulze, and A. Lipowsky. 2011. Differential effects of plantdiversity on functional trait variation of grass species. Annals of Botany 107:157–169.77BibliographyGuillaume, F., and M. C. Whitlock. 2007. Effects of migration on the genetic covariance matrix. Evolution61:2398–2409.Haloin, J. R., and S. Y. Strauss. 2008. Interplay between ecological communities and evolution. review offeedbacks from microevolutionary to macroevolutionary scales. Annals of the New York Academy ofSciences 1133:87–125.Hendry, A. P., and M. T. Kinnison. 1999. Perspective: The pace of modern life: Measuring rates of contem-porary microevolution. Evolution 53:1637–1653.Hermisson, J., and P. S. Pennings. 2005. Soft sweeps: Molecular population genetics of adaptation fromstanding genetic variation. Genetics 169:2335–2352.Holt, R. D., and R. Gomulkiewicz. 1997. How does immigration influence local adaptation? a reexaminationof a familiar paradigm. The American Naturalist 149:563–572.Hooper, U. D., M. Solan, A. Symstad, S. Díaz, M. O. Gessner, N. Buchmann, V. Degrange, P. Grime, F. Hu-lot, F. Mermillod-Blondin, J. Roy, E. Spehn, and L. van Peer. 2002. Biodiversity and Ecosystem Function-ing Synthesis and Perspectives, chap. Species diversity, functional diversity, and ecosystem functioning,pages 195–208. 17. Oxford University Press.Huey, R. B., and J. G. Kingsolver. 1993. Evolution of resistance to high temperature in ectotherms. AmericanNaturalist 142:S21–S46.Hufbauer, R. A., M. Szücs, E. Kasyon, C. Youngberg, M. J. Koontz, C. Richards, T. Tuff, and B. A. Mel-bourne. 2015. Three types of rescue can avert extinction in a changing environment. Proceedings of theNational Academy of Sciences of the United States of America 112:10557–10562.Huijben, S., A. S. Bell, D. G. Sim, D. Tomasello, N. Mideo, T. Day, and A. F. Read. 2013. Aggressivechemotherapy and the selection of drug resistant pathogens. PLos Pathog 9:e1003578.Isbell, F., P. B. Reich, D. Tilman, S. E. Hobbie, S. Polasky, and S. Binder. 2013. Nutrient enrichment, biodi-versity loss, and consequent declines in ecosystem productivity. Proceedings of the National Academy ofSciences of the United States of America 110:11911–11916.Johansson, J. 2008. Evolutionary responses to environmental changes: How does competition affect adapta-tion? Evolution 62:421–435.Jones, A. G. 2008. A theoretical quantitative genetic study of negative ecological interactions and extinctiontimes in changing environments. BMC Evolutionary Biology 8:119.Jones, A. G., S. J. Arnold, and R. Burger. 2003. Stability of the g-matrix in a population experiencingpleiotropic mutation, stabilizing selection, and genetic drift. Evolution 57:1747–1760.Jones, M. E., and L. A. Barmuta. 1998. Diet overlap and relative abundance of sympatric dasyurid carnivores:A hypothesis of competition. Journal of Animal Ecology 67:410–421.78BibliographyJump, A. S., and J. Penuelas. 2005. Running to stand still: adaptation and the response of plants to rapidclimate change. Ecology Letters 8:1010–1020.Kaplan, I., and R. F. Denno. 2007. Interspecific interactions in phytophagous insects revisited: a quantitativeassessment of competition theory. Ecology Letters 10:977–994.Kimball, S., A. L. Angert, T. E. Huxman, and D. Lawrence Venable. 2010. Contemporary climate change inthe sonoran desert favors cold-adapted species. Global Change Biology 16:1555–1565.Kisdi, E. 1999. Evolutionary branching under asymmetric competition. Journal of Theoretical Biology197:149–162.Kisdi, E., F. Jacobs, and S. Geritz. 2002. Red queen evolution by cycles of evolutionary branching andextinction. Selection 2:161–176.Klanderud, K. 2005. Climate change effects on species interactions in an alpine plant community. Journalof Ecology 93:127–137.Kleynhans, E. J., A. E. Jolles, M. R. E. Bos, and H. Olff. 2011. Resource partitioning along multiple nichedimensions in differently sized African savanna grazers. Oikos 120:591–600.Kleynhans, E. J., S. Otto, P. Reich, and M. Vellend. 2016. Adaptation to elevated CO2 in different biodiversitycontexts. Nature Communications 7:12358.Knops, J. M. H., S. Naeem, and P. B. Reich. 2007. The impact of elevated CO2, increased nitrogen availabil-ity and biodiversity on plant tissue quality and decomposition. Global Change Biology 13:1960–1971.Kokko, H., A. López-Sepulcre, and L. J. Morrell. 2006. From hawks and doves to self-consistent games ofterritorial behavior. The American Naturalist 167:901–912.Kopp, M., and S. Matuszewski. 2014. Rapid evolution of quantitative traits: theoretical perspectives. Evo-lutionary Applications 7:169–191.Kraft, N. J. B., O. Godoy, and J. M. Levine. 2015. Plant functional traits and the multidimensional nature ofspecies coexistence. Proceedings of the National Academy of Sciences of the United States of America112:797–802.Kuznetsova, A., P. B. Brockhoff, and R. H. B. Christensen. 2014. lmerTest: Tests for random and fixedeffects for linear mixed effect models (lmer objects of lme4 package).———. 2017. lmerTest package: Tests in linear mixed effects models. Journal of Statistical Software82:1–26.Lande, R. 1979. Quantitative genetic analysis of multivariate evolution, applied to brain-body size allometry.Evolution 33:402–416.79Bibliography———. 1993. Risks of population extinction from demographic and environmental stochasticity and randomcatastrophes. The American Naturalist 142:911–927.Lanfear, R., H. Kokko, and A. Eyre-Walker. 2014. Population size and the rate of evolution. Trends inEcology and Evolution 29:33–41.Lau, J. A., R. G. Shaw, P. B. Reich, and P. Tiffin. 2010. Species interactions in a changing environment:elevated CO2 alters the ecological and potential evolutionary consequences of competition. EvolutionaryEcology Research 12:435–455.———. 2014. Indirect effects drive evolutionary responses to global change. The New Phytologist 201:335–343.Lavorel, S., and E. Garnier. 2002. Predicting changes in community composition and ecosystem functioningfrom plant traits: revisiting the holy grail. Functional Ecology 16:545–556.Lawrence, D., F. Fiegna, V. Behrends, J. G. Bundy, A. B. Phillimore, T. Bell, and T. G. Barraclough. 2012.Species interactions alter evolutionary responses to a novel environment. PLoS-Biology 10:941.Lawton, J., and M. Hassell. 1981. Asymmetrical competition in insects. Nature 289:793–795.Lee, T. D., S. H. Barrott, and P. B. Reich. 2011. Photosynthetic responses of 13 grassland species across11 years of free-air CO2 enrichment is modest, consistent and independent of N supply. Global ChangeBiology 17:2893–2904.Lee, T. D., P. B. Reich, and M. G. Tjoelker. 2003. Legume presence increases photosynthesis and N concen-trations of co-occurring non-fixers but does not modulate their responsiveness to carbon dioxide enrich-ment. Oecologia 137:22–31.Lee, T. D., M. G. Tjoelker, D. S. Ellsworth, and P. B. Reich. 2001. Leaf gas exchange responses of 13 prairiegrassland species to elevated CO2 and increased nitrogen supply. New Phytologist 150:405–418.Leger, E. A., and E. K. Espeland. 2010. Coevolution between native and invasive plant competitors: impli-cations for invasive species management. Evolutionary Applications 3:169–178.Lehikoinen, A., P. Byholm, E. Ranta, P. Saurola, J. Valkama, E. Korpimäki, H. Pietiäinen, and H. Henttonen.2009. Reproduction of the common buzzard at its northern range margin under climatic change. Oikos118:829–836.Lehikoinen, A., A. Lindén, P. Byholm, E. Ranta, P. Saurola, J. Valkama, V. Kaitala, and H. Lindén. 2013.Impact of climate change and prey abundance on nesting success of a top predator, the goshawk. Oecologia171:283–293.Lenski, R. E., and A. F. Bennett. 1993. Evolutionary response of Escherichia coli to thermal stress. AmericanNaturalist 142:S47–S64.80BibliographyLinhart, Y. B., and M. C. Grant. 1996. Evolutionary significance of local genetic differentiation in plants.Annual Review of Ecology and Systematics 27:237–277.Lipowsky, A., C. Roscher, J. Schumacher, S. G. Michalski, M. Gubsch, N. Buchmann, E.-D. Schulze, andB. Schmid. 2015. Plasticity of functional traits of forb species in response to biodiversity. Perspectives inPlant Ecology, Evolution, and Systematics 17:66–77.Liu, X., M. Chen, H. L. Collins, D. W. Onstad, R. T. Roush, Q. Zhang, E. D. Earle, and A. M. Shelton. 2014.Natural enemies delay insect resistance to Bt crops. PLoS ONE 9:e90366.Lorentzen, S., C. Roscher, J. Schumacher, E.-D. Schulze, and B. Schmid. 2008. Species richness and identityaffect the use of aboveground space in experimental grasslands. Perspectives in Plant Ecology, Evolutionand Systematics 10:73 – 87.Lovric, M., ed. 2011. International Encyclopedia of Statistical Science, chap. Skew-Normal Distribution.Springer.Lüscher, A., T. Hebeisen, S. Zanetti, U. A. Hartwig, H. Blum, G. R. Hendry, and J. Nösberger. 1996. Carbondioxide, populations, and communities, chap. Differences between legumes and nonlegumes of perma-nent grassland in their responses to free-air Carbon dioxide enrichment: Its effect on competition in amultispecies mixture, pages 273–286. 18. Academic Press.MacArthur, R., and E. Wilson. 1967. Island biogeography. Princeton University Press, Princeton, N.J.Marquard, E., A. Weigelt, C. Roscher, M. Gubsch, A. Lipowsky, and B. Schmid. 2009. Positive biodiversity-productivity relationship due to increased plant density. Journal of Ecology 97:696–704.Martin, G., R. Aguile ÌA˛e, J. Ramsayer, O. Kaltz, and O. Ronce. 2012. The probability of evolution-ary rescue: towards a quantitative comparison between theory and evolution experiments. PhilosophicalTransactions of the Royal Society B: Biological Sciences 368:(doi:10.1098/rstb.2012.0088).Matsuda, H., and P. A. Abrams. 1994. Runaway evolution to self-extinction under asymmetrical competition.Evolution 48:1764–1772.Matuszewski, S., J. Hermisson, and M. Kopp. 2015. Catch me if you can: Adaptation from standing geneticvariation to a moving phenotypic optimum. Genetics 200:1255–1274.Mazzucato, A., M. Falcinelli, and F. Veronesi. 1996. Evolution and adaptedness in a facultatively apomicticgrass, Poa pratensis L. Euphytica 92:13–19.McArdle, B. H., and M. J. Anderson. 2001. Fitting multivariate models to community data: a comment ondistance-based redundancy analysis. Ecology 82:290–297.McGeoch, M. A., S. H. M. Butchart, D. Spear, E. Marais, E. J. Kleynhans, A. Symes, J. Chanson, andM. Hoffmann. 2010. Global indicators of biological invasion: species numbers, biodiversity impact andpolicy responses. Diversity and Distributions 16:95–108.81BibliographyMellard, J. P., C. de Mazancourt, and M. Loreau. 2015. Evolutionary responses to environmental change:trophic interactions affect adaptation and persistence. Proceedings of the Royal Society B. 282:1–8.Mikola, J., R. Bardgett, and K. Hedlund. 2002. Biodiversity, ecosystem functioning and soil decomposerfood webs. Pages 169–180 in M. Loreau, S. Naeem, and P. Inchausti, eds. Biodiversity and EcosystemFunctioning Synthesis and Perspectives. Oxford University Press, Oxford.Mwangi, M. M., S. W. Wu, Y. Zhou, K. Sieradzki, H. de Lencastre, P. Richardson, D. Bruce, E. Rubin,E. Myers, E. D. Siggia, and A. Tomasz. 2007. Tracking the in vivo evolution of multidrug resistance instaphylococcus aureus by whole-genome sequencing. Proceedings of the National Academy of Sciencesof the United States of America 104:9451–9456.Navas, M.-L., E. Garnier, M. P. Austin, and R. M. Gifford. 1999. Effect of competition on the responses ofgrasses and legumes to elevated atmospheric CO2 along a nitrogen gradient: Differences between isolatedplants, monocultures and multi-species mixtures. New Phytologist 143:323–331.Navas, M.-L., and C. Violle. 2009. Plant traits related to competition: how do they shape the functionaldiversity of communities? Community Ecology 10:131–137.Nicotra, A., O. Atkin, S. Bonser, A. Davidson, E. Finnegan, M. U., P. Poot, M. Purugganan, C. Richards,F. Valladares, and M. van Kleunen. 2010. Plant phenotypic plasticity in a changing climate. Trends inPlant Science 15:684–692.Norberg, J., M. C. Urban, M. Vellend, C. A. Klausmeier, and N. Loeuille. 2012. Eco-evolutionary responsesof biodiversity to climate change. Nature Climate Change 2:747–751.Northfield, T. D., and A. R. Ives. 2013. Coevolution and the effects of climate change on interacting species.PLoS Biology 11:e1001685.Novotny, A. M., J. D. Schade, S. E. Hobbie, A. D. Kay, M. Kyle, P. B. Reich, and J. J. Elser. 2007. Stoichio-metric response of nitrogen-fixing and non-fixing dicots to manipulations of CO2, nitrogen, and diversity.Oecologia 151:687–696.Nussey, D. H., E. Postma, P. Gienapp, and M. Visser. 2005. Selection on heritable phenotypic plasticity in awild bird population. Science 310:304–306.Oelmann, Y., N. Buchmann, G. Gleixner, M. Habekost, C. Roscher, S. Rosenkranz, E.-D. Schulze, S. Stein-beiss, V. M. Temperton, A. Weigelt, W. W. Weisser, and W. Wilcke. 2011. Plant diversity effects onaboveground and belowground n pools in temperate grassland ecosystems: Development in the first 5years after establishment. Global Biogeochemical Cycles 25.Oksanen, J., F. G. Blanchet, M. Friendly, R. Kindt, P. Legendre, D. McGlinn, P. R. Minchin, R. B. O’Hara,G. L. Simpson, P. Solymos, M. H. H. Stevens, E. Szoecs, and H. Wagner. 2018. vegan: CommunityEcology Package. R package version 2.4-6.82BibliographyOrr, H. A. 2001. The genetics of species differences. Trends in Ecology & Evolution 16:343–350.Orr, H. A., and R. L. Unckless. 2008. Population extinction and the genetics of adaptation. The AmericanNaturalist 172:160–169.———. 2014. The population genetics of evolutionary rescue. PLoS Genetics 10:e1004551.doi:10.1371/journal.pgen.1004551.Osmond, M. M., and C. de Mazancourt. 2013. How competition affects evolutionary rescue. PhilosophicalTransactions of the Royal Society B: Biological Sciences .Osmond, M. M., S. P. Otto, and C. A. Klausmeier. 2017. When predators help prey adapt and persist in achanging environment. The American Naturalist 190:83–98. PMID: 28617633.Otto, S. P. 2018. Adaptation, speciation and extinction in the anthropocene. Proceedings of the Royal SocietyB. 285:20182047.Pachauri, R., and L. Meyer. 2014. Climate change 2014: Synthesis report. contribution of working groups i,ii and iii to the fifth assessment report of the integovernmental panel on climate change. Tech. rep., IPCC,Geneva, Switzerland.Pakanen, V.-M., M. Orell, E. Vatka, S. Rytkönen, and J. Broggi. 2016. Different ultimate factors definetiming of breeding in two related species. PLoS ONE 11:e0162643.Parmesan, C., and G. Yohe. 2003. A globally coherent fingerprint of climate change impacts across naturalsystems. Nature 421:37–42.Partridge, L., B. Barrie, N. H. Barton, K. Fowler, and V. French. 1995. Rapid laboratory evolution of adultlife-history traits in Drosophila melanogaster in response to temperature. Evolution 49:538–544.Patterson, D. T., E. P. Flint, and J. L. Beyers. 1984. Effects of co2 enrichment on competition between a c4weed and a c3 crop. Weed Science 32:101–105.Pena-Miller, R., D. Laehnemann, G. Jansen, A. Fuentes-Hernandez, P. Rosenstiel, H. Schulenburg, andR. Beardmore. 2013. When the most potent combination of antibiotics selects for the greatest bacterialload: The smile-frown transition. PLoS Biology 11:e1001540.Pérez-Harguindeguy, N., S. Díaz, E. Garnier, S. Lavorel, H. Poorter, P. Jaureguiberry, M. S. Bret-Harte,W. Cornwell, J. M. Craine, D. E. Gurvich, C. Urcelay, E. J. Veneklaas, P. B. Reich, L. Poorter, I. J.Wright, P. Ray, L. Enrico, J. G. Pausas, A. de Vos, N. Buchmann, G. Funes, F. Quetier, J. G. Hodg-son, K. Thompson, H. Morgan, H. ter Steege, M. van der Heijden, L. Sack, B. Blonder, P. Poschlod,M. Vaieretti, G. Conti, A. Staver, S. Aquino, and J. H. C. Cornelissen. 2013. New handbook for measure-ment of plant functional traits worldwide. Australian Journal of Botany 61:167–234.Persson, L. 1983. Effects of intra- and interspecific competition on dynamics and size structure of a perchperca fluviatilis and roach rutilus rutilus population. Oikos 41:126–132.83BibliographyPetchey, O. L., and K. J. Gaston. 2002. Functional diversity (fd), species richness and community composi-tion. Ecology Letters 5:402–411.Poorter, H., and M.-L. Navas. 2003. Plant growth and competition at elevated CO2: on winners, losers andfunctional groups. New Phytologist 157:175–198.Price, T. D., and M. Kirkpatrick. 2009. Evolutionarily stable range limits set by interspecific competition.Proceedings of the Royal Society B. 276:1429–1434.Püschel, D., M. Janouskova, M. Hujslova, R. Slavikova, H. Gryndlerova, and J. Jansa. 2016. Plant-funguscompetition for nitrogen erases mycorrhizal growth benefits of Andropogon gerardii under limited nitro-gen supply. Ecology and Evolution 6:4332–4346.R Core Team. 2018. R: A Language and Environment for Statistical Computing. R Foundation for StatisticalComputing, Vienna, Austria.Ramette, A. 2007. Multivariate analyses in microbial ecology. FEMS Microbiol Ecol 62:142–160.Ramsayer, J., O. Kaltz, and M. E. Hochberg. 2012. Evolutionary rescue in populations of Pseudomonasfluorescens across and antibiotic gradient. Evolutionary Applications 6:608–616.RCoreTeam. 2013. R: A language and environment for statistical computing.Read, A. F., T. Day, and S. Huijben. 2011. The evolution of drug resistance and the curious orthodoxyof aggressive chemotherapy. Proceedings of the National Academy of Sciences of the United States ofAmerica 108:10871–10877.Reich, P. B., S. E. Hobbie, T. D. Lee, and M. A. Pastore. 2018. Unexpected reversal of C3 versus C4 grassresponse to elevated CO2 during a 20-year field experiment. Science 360:317–320.Reich, P. B., J. Knops, D. Tilman, J. Craine, D. Ellsworth, M. G. Tjoelker, T. Lee, D. Wedin, S. Naeem,D. Bahauddin, G. Hendrey, S. Jose, K. Wrage, J. Goth, and W. Bengston. 2001a. Plant diversity enhancesecosystem responses to elevated CO2 and nitrogen deposition. Nature 410:809–812.Reich, P. B., D. Tilman, J. Craine, D. Ellsworth, M. G. Tjoelker, J. Knops, D. Wedin, S. Naeem, D. Bahaud-din, J. Goth, W. Bengtson, and T. D. Lee. 2001b. Do species and functional groups differ in acquisitionand use of C, N and water under varying atmospheric CO2 and N availability regimes? A field test with16 grassland species. New Phytologist 150:435–448.Reich, P. B., D. Tilman, F. Isbell, K. Mueller, S. E. Hobbie, D. F. B. Flynn, and N. Eisenhauer. 2012. Impactsof biodiversity loss escalate through time as redundancy fades. Science 336:589–92.Reich, P. B., D. Tilman, S. Naeem, D. Ellsworth, J. M. H. Knops, J. Craine, D. Wedin, and J. J. Trost. 2004.Species and functional group diversity independently influence biomass accumulation and its response toCO2 and N. Proceedings of the National Academy of Sciences of the United States of America 101:10101–10106.84BibliographyRenaud, S., J.-C. Auffray, and J. Michaux. 2006. Conserved phenotypic variation patterns, evolution alonglines of least resistance, and departure due to selection in fossil rodents. Evolution 60:1701–1717.Reznick, D. N., and C. K. Ghalambor. 2001. The population ecology of contemporary adaptations: whatempirical studies reveal about the conditions that promote adaptive evolution. Genetica 112-113:183–98.Richardson, J. L., M. C. Urban, D. I. Bolnick, and D. K. Skelly. 2014. Microgeographic adaptation and thespatial scale of evolution. Trends in Ecology and Evolution 29:165–76.Roscher, C., B. Schmid, N. Buchmann, A. Weigelt, and E.-D. Schulze. 2011. Legumes species differ in theresponses of their functional traits to plant diversity. Oecologia 165:437–452.Rosenzweig, M. L. 1995. Species diversity in space and time. Cambridge University Press, Cambridge.Roughgarden, J. 1979. Theory of population genetics and evolutionary ecology: an introduction. Macmillan.Rummel, J. D., and J. Roughgarden. 1983. Some difference between invasion-structured and coevolution-structured competitive communities: A preliminary theoretical analysis. Oikos 41:477–486.Satterthwaite, F. E. 1946. An approximate distribution of estimates of variance components. BiometricsBulletin 2:110–114.Scherber, C., N. Eisenhauer, W. W. Weisser, B. Schmid, W. Voigt, M. Fischer, E.-D. Schulze, C. Roscher,A. Weigelt, E. Allan, H. Beßler, M. Bonkowski, N. Buchmann, F. Buscot, L. W. Clement, A. Ebeling,C. Engels, S. Halle, I. Kertscher, A.-M. Klein, R. Koller, S. König, E. Kowalski, V. Kummer, A. Kuu,M. Lange, D. Lauterbach, C. Middelhoff, V. D. Migunova, A. Milcu, R. Müller, S. Partsch, J. S. Peter-mann, C. Renker, T. Rottstock, A. Sabais, S. Scheu, J. Schumacher, V. M. Temperton, and T. Tscharntke.2010. Bottom-up effects of plant diversity on multitrophic interactions in a biodiversity experiment. Na-ture 468:553–556.Schlichting, C. 1986. The evolution of phenotypic plasticity in plants. Annual Review of Ecology andSystematics 17:667–693.Schloss, C. A., T. A. Nuñez, and J. J. Lawler. 2012. Dispersal will limit ability of mammals to track climatechange in the western hemisphere. Proceedings of the National Academy of Sciences of the United Statesof America 109:8606–8611.Schluter, D. 1996. Adaptive radiation along genetic lines of least resistance. Evolution 50:1766–1774.Schöb, C., R. W. Brooker, and D. Zuppinger-Dingley. 2018. Evolution of facilitation requires diverse com-munities. Nature Ecology & Evolution https://doi.org/10.1038/:S41559–018–0623–2.Schoener, T. W., D. A. Spiller, and J. B. Losos. 2001. Predators increase the risk of catastrophic extinctionof prey populations. Nature 412:183–186.Schwinning, S., and J. Weiner. 1998. Mechanisms determining the degree of size asymmetry in competitionamong plants. Oecologia 113:447–455.85BibliographyShaw, F. H., R. G. Shaw, G. S. Wilkinson, and M. Turelli. 1995. Changes in genetic variances and covari-ances: G whiz! Evolution 49:1260–1267.Shaw, R. G., C. J. Geyer, S. Wagenius, H. H. Hangelbroek, and J. R. Etterson. 2008. Unifying life-historyanalyses for inference of fitness and population growth. The American Naturalist 172:E35–47.Slatkin, M. 1980. Ecological character displacement. Ecology 61:163–177.Song, Y., S. Endepols, N. Klemann, D. Richter, F.-R. Matuschka, C.-H. Shih, M. W. Nachman, and M. H.Kohn. 2011. Adaptive introgression of anticoagulant rodent poison resistance by hybridization betweenold world mice. Current Biology 21:1296–1301.Stearns, S. C. 1989. The evolutionary significance of phenotypic plasticity. BioScience 39:436–445.Steinger, T., A. Stephan, and B. Schmid. 2007. Predicting adaptive evolution under elevated atmosphericCO2 in the perennial grass Bromus erectus. Global Change Biology 13:1028–1039.Strauss, S. Y., J. A. Lau, T. W. Schoener, and P. Tiffin. 2008. Evolution in ecological field experiments:implications for effect size. Ecology Letters 11:199–207.Strengbom, J., and P. B. Reich. 2006. Elevated [CO2] and increased N supply reduce leaf disease and relatedphotosynthetic impacts on Solidago rigida. Oecologia 149:519–525.Sunday, J. M., P. Calosi, S. Dupont, P. L. Munday, J. H. Stillman, and T. B. Reusch. 2014. Evolution in anacidifying ocean. Trends in Ecology & Evolution 29:117–125.Thein, S., C. Roscher, and E.-D. Schulze. 2008. Effects of trait plasticity on aboveground biomass productiondepend on species identity in experimental grasslands. Basic and Applied Ecology 9:475–484.Tilman, D. 1977. Resource competition between plankton algae: an experimental and theoretical approach.Ecology 58:338–348.Tilman, D., J. M. H. Knops, D. Wedin, P. B. Reich, R. Mark, and S. Evan. 1997. The influence of functionaldiversity and composition on ecosystem processes. Science 227:1300–1302.Tilman, D., P. B. Reich, and F. Isbell. 2012. Biodiversity impacts ecosystem productivity as much as re-sources, disturbance, or herbivory. Proceedings of the National Academy of Sciences of the United Statesof America 109:10394–10397.Tilman, D., P. B. Reich, J. M. H. Knops, D. Wedin, M. Troy, and L. Clarence. 2001. Diversity and produc-tivity in a long-term grassland experiment. Science 294:843–845.Tomiolo, S., W. H. Van der Putten, and K. Tielbörger. 2015. Separating the role of biotic interactions andclimate in determining adaptive response of plants to climate change. Ecology 96:1298–1308.Tseng, M., and M. I. O’Connor. 2015. Predators modify the evolutionary response of prey to temperaturechange. Biology Letters 11.86BibliographyTurkington, R., and J. L. Harper. 1979. The growth, distribution and neighbour relationships of Trifoliumrepens in a permanent pasture: IV. Fine-scale biotic differentiation. Journal of Ecology 67:245–254.Uecker, H., S. P. Otto, and J. Hermisson. 2014. Evolutionary rescue in structured populations. The AmericanNaturalist 183:E17–E35.Urban, M. C. 2015. Accelerating extinction risk from climate change. Science 348:571–573.USDA, and NRCS. 2015. Kentucky bluegrass (Poa pratensis L.).Vamosi, S. M. 2005. On the role of enemies in divergence and diversification of prey: review and synthesis.Canadian Journal of Zoolology 83:894–910.Van Den Elzen, C. L., E. J. Kleynhans, and S. P. Otto. 2017. Asymmetric competition impacts evolutionaryrescue in a changing environment. Proc. R. Soc. B 284:20170374.van Doorslaer, W., R. Stoks, C. Duvivier, A. Bednarska, and L. de Meester. 2009. Population dynamicsdetermine genetic adaptation to temperature in daphnia. Evolution 63:1867–1878.Van Doorslaer, W., R. Stoks, I. Swillen, H. Feuchtmayr, D. Atkinson, B. Moss, and L. De Meester. 2010.Experimental thermal microevolution in community-embedded Daphnia populations. Climate Research43:81–89.Vellend, M., L. Baeten, A. Becker-Scarpitta, V. Boucher-Lalonde, J. L. McCune, J. Messier, I. H. Myers-Smith, and D. F. Sax. 2017. Plant biodiversity change across scales during the anthropocene. Annu. Rev.Plant Biol. 68:563–586.Vellend, M., and M. A. Geber. 2005. Connections between species diversity and genetic diversity. EcologyLetters 8:767–781.Visser, M., A. van Noordwijk, J. Tinbergen, and C. Lessells. 1998. Warmer springs lead to mistimed repro-duction in great tits (Parus major). Proceedings of the Royal Society B. 265:1867–1870.Vitousek, P. M., J. D. Aber, R. W. Howarth, G. E. Likens, P. A. Matson, D. W. Schindler, W. H. Schlesinger,and D. G. Tilman. 1997. Human alteration of the global nitrogen cycle: sources and consequences.Ecological Applications 7:737–750.Wagner, L. K. 1989. Size dependent reproduction in Poa annua and P. pratensis. Pages 273–284 in J. H.Bock and Y. B. Linhart, eds. The Evolutionary Ecology of Plants. Westview Press, Boulder.Walther, G.-R., E. Post, P. Convey, A. Menzel, C. Parmesan, T. J. C. Beebee, J.-m. Fromentin, O. Hoegh-Guldberg, and F. Bairlein. 2002. Ecological responses to recent climate change. Nature 416:389–395.Wedin, D., and D. Tilman. 1993. Competition among grasses along a nitrogen gradient: initial conditionsand mechanisms of competition. Ecological Monographs 63:199–229.87Bibliography———. 1996. Influence of nitrogen loading and species composition on the carbon balance of grasslands.Science 274:1720–1723.Weiner, J. 1990. Asymmetric competition in plant populations. Trends in Ecology & Evolution 5:360–364.Weisser, W. W., C. Roscher, S. T. Meyer, A. Ebeling, G. Luo, E. Allan, H. Besser, R. L. Barnard, N. Buch-mann, F. Buscot, C. Engels, C. Fischer, M. Fischer, A. Gessler, G. Gleixner, S. Halle, A. Hildebrandt,H. Hillebrand, H. de Kroon, M. Lange, S. Leimer, X. Le Roux, A. Milcu, L. Mommer, P. A. Niklaus,Y. Oelmann, R. Proulx, J. Roy, C. Scherber, M. Scherer-Lorenzen, S. Scheu, T. Tscharntke, M. Wachen-dorf, C. Wagg, A. Weigelt, W. Wilcke, C. Wirth, E.-D. Schulze, B. Schmid, and N. Eisenhauer. 2017.Biodiversity effects on ecosystem functioning in a 15-year grassland experiment: Patterns, mechanisms,and open questions. BASIC AND APPLIED ECOLOGY 23:1–73.West, J. B., J. HilleRisLambers, T. D. Lee, S. E. Hobbie, and P. B. Reich. 2005. Legume species identityand soil nitrogen supply determine symbiotic nitrogen-fixation responses to elevated atmospheric [CO2].New Phytologist 167:523–530.Whitlock, M. C., and D. Schluter. 2015. The analysis of biological data. 2nd ed. Roberts and CompanyPublishers.Wilkinson, G. S., K. Fowler, and L. Partridge. 1990. Resistance of genetic correlation structure to directionalselection in Drosophila melanogaster. Evolution 44:1990–2003.Willi, Y., and A. A. Hoffmann. 2009. Demographic factors and genetic variation influence population per-sistence under environmental change. Journal of Evolutionary Biology 22:124–133.Wolinska, J., and K. C. King. 2009. Environment can alter selection in host-parasite interactions. Trends inParasitology 25:236–44.Wray, S. M., and B. D. Strain. 1987. Competition in old-field perennials under co2 enrichment. Ecology69:1116–1120.Wright, A., S. A. Schnitzer, I. A. Dickie, A. R. Gunderson, G. A. Pinter, S. A. Mangan, and P. B. Reich.2013. Complex facilitation and competition in a temperate grassland: Loss of plant diversity and elevatedCO2 have divergent and opposite effects on oak establishment. Oecologia 171:449–458.Wright, A., S. A. Schnitzer, and P. B. Reich. 2014. Living close to your neighbours: The importance of bothcompetition and facilitation in plant communities. Ecology 95:2213–2223.Zanetti, S., U. A. Hartwig, A. Lüscher, T. Hebeisen, F. M, F. B. U., H. G. R., H. Blum, and J. Nösberger.1996. Stimulation of symbiotic N2 fixation in Trifolium repens L. under elevated atmospheric CO2 in agrassland ecosystem. Plant Physiology 112.Zuppinger-Dingley, D., B. Schmid, J. S. Peterman, V. Yadav, G. B. De Deyn, and D. F. B. Flynn. 2014.Selection for niche differentiation in plant communities increases biodiversity effects. Nature 515:108–111.88BibliographyZuur, A. E., E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith. 2009. Mixed effects modeling fornested data. Chap. Chapter 5, pages 101–142 in Mixed Effects Models and Extensions in Ecology withR. Springer Science+Business Media, New York.89Appendix90Appendix AAppendix for Chapter 2: Evolutionaryrescue when resources change.A.1 Supplementary resultsWe investigated the probability of rescue as a function of distance from the burn-in optimum when thestandard deviation of the selection surface (~σw) and resource surface (~σk) were not equal in the x and ydirections (Figure A.2). Rescue, in the genetic model, from standing genetic variation or standing geneticvariation and de novo mutation was still biased in the direction of greatest mutational variance althoughless so than if the selection surface is equal in all directions (Figure A.2 A and C). However, rescue fromde novo mutation alone was unexpectedly shifted towards the direction of greatest variance in the selectionsurface (Figure A.2 B). This is likely because evolution occurs in directions of least resistance. Because theselection surface is so shallow in the x direction, evolution occurs more easily in that direction even throughmutational variance is greatest in the y direction. Furthermore, mutations of large size are relatively rare,even in the direction of greatest mutational variance (y direction) thus although some populations are rescuedin the y direction at distance further out than in the x direction the number of populations rescued in thatdirection is not large enough to pull the line up to reflect the mutational variance. In contrast with standinggenetic variation there has been enough time for mutations to accumulate in the y direction and thereforethe probability of rescue reflects the mutational variance more than the selection surface, as is expected forthe genetic model. Further evidence for this comes from the fact that if the mutation rate is increased to0.1, while holding all other parameters constant, the slope of the line for de novo mutation shifts up to 1.46suggesting that it is a lack of mutations in the y direction that causes the unexpected result.In contrast to the genetic model, the ecological model behaves as expected. Rescue from standing geneticvariation and both standing genetic variation and mutation more closely mimics the shape of the resourcedistribution than the M-matrix as a result of diversifying selection from competition (Figure A.2 D and F).De novo mutation more closely resembles the shape of the M-matrix and is biased in the direction of greatestmutational variance (Figure A.2 E).In figure A.3, we show how the probability of rescue falls off as the optimum shifts in the horizontal(first column) or vertical (second column) directions, in the case of an asymmetrical mutation matrix and anasymmetrical selection surface (FigureA.3 A and B) or asymmetrical resource surface (Figure A.3 panelsC and D). As before, in the genetic model, the relative contribution of standing and de novo variance issimilar in both directions (Figure A.3 A and B). Similarly, in the ecological model, the probability of rescuechanges according to whether mutations are coming from standing genetic variation or de novo mutation. If91A.2. Supporting Figuresthe mutational effect size is less than the width of the resource kernel (σν<σk) rescue from standing geneticvariation allows populations to adapt to larger changes in the environment than if adaptation were to comefrom de novo mutation alone (Figure A.3 C (P<0.0001)). In contrast, if σν>σk then de novo mutation playsa larger role allowing rescue to occur further out than if it came from standing genetic variation (compareblue and red curves Figure A.3 D (P<0.0001)).A.2 Supporting Figures-5 5-55-5 5-55-5 5-55-5 5-55-5 5-55-10 -5 5 10-10-5510-10 -5 0 5 10-10-50510-5 5-55-5 5-55-5 5-55-5 5-55-5 5-55-10 -5 5 10-10-5510-10 -5 0 5 10-10-50510No	CompetitionCompetition1000Adaptive	landscapeM-matrix 2000 3000 4000 5000Figure A.1: Representation of the g-matrix every 1000 generations up to 5000 generations during burn-in toshow that the population has reached mutation selection balance for both the genetic and ecological models.The shape of the adaptive landscape is indicated by a black circle, the m-matrix is shown in red, blue pointsare individuals and the blue circle shows the area containing 95% of the individuals.92A.2. Supporting Figures!"1	=	1!"2 =	4!w1 =	30!w2 =	15Slope	=	1.24 Slope	=	0.57 Slope	=	1.26D E FSlope	=	0.44 Slope	=	1.97 Slope	=0.61!"1	=	1!"2 =	4!k1 =	3!k2 =	1.5Distance	to	new	environmental	resource	optimum	in	! " 2directionDistance	to	new	environmental	resource	optimum	in	!"1 directionPrinted by Mathematica for Students1.000.20.40.60.8Proportion	rescued0 10 20010200 10 20010200 10 2001020-10 -5 5 10-10-5510A B CGenetic	modelEcological	model0 10 20010200 10 2001020-10 -5 5 10-10-55100 10 2001020Trait	value	1Trait	value	2Trait	value	1Trait	value	2Figure A.2: Probability of rescue as a function of the new environmental optimum in two dimensions forthe genetic (A, B, and C) and ecological (D, E, and F) models when the fitness or resource distributionvariances are unequal. Grey scale indicates the number of rescue events out of ten independent simulations.Columns show the results from standing genetic variation (A and D), de novo mutation (B and E), and both(C and F). The squares with dots indicated the areas where fitness is greater than one even for individuals ofphenotype {0,0} so rescue is not needed. The inset figures in panels A and D are an example for the resultsobtained from one burn-in after 5000 generations. The blue ellipse represents the area within which 95% ofthe population is present (G-matrix), with each dot showing the trait combinations of an individual; the redellipse indicates the area within which 95% of mutations occur (M-matrix); the black ellipse indicates wherethe fitness of 100 individuals in the population of phenotype {0,0} equals one, representing the adaptivelandscape. The standard deviation of the mutational effect size (~σν), the standard deviation of the selectionsurface (~σw), and the standard deviation of the resource surface (~σk) are indicated in the left most column ofthe figure, all other parameters were set to their default values (Table 2.1).93A.2. Supporting Figures!"1 =	1 !"2 =	4Probability	of	rescueDistance	of	fitness	peak	away	from	burnin optimumA BC DDe	novo	mutationStanding	genetic	variationGenetic	model!w1 =	30!w2 =	15Ecological	model!k1 =	3!k2 =	1.55 10 15 20 250.00.20.40.60.81.05 10 15 20 250.00.20.40.60.81.05 10 15 20 250.00.20.40.60.81.05 10 15 20 250.00.20.40.60.81.0Figure A.3: Probability of rescue from standing genetic variation and de novo mutation in the genetic andecological models. Points represent the proportion of populations rescued over 10 independent replicatesimulation runs, solid lines are the logistic model fits, and shaded areas are 95% confidence intervals forthe logistic model. The standard deviation of the mutational effect size (~σν), the standard deviation of theselection surface (~σw), and the standard deviation of the resource surface (~σk) are indicated on the figure.All other parameters were set to their default values (Table 3.1).94Appendix BAppendix for Chapter 3: Asymmetriccompetition impacts evolutionary rescue ina changing environmentB.1 Supplementary MethodsWhen the environment changed rapidly, the leading species persisted for longer than a single-species com-munity. As discussed in the text, however, this has two potential causes: the existence of more resources inthe two-species simulations than the one-species simulations (Km in Table B.1), so that the average popu-lation size of each species was 2500, and the effects of competition on the mean trait, causing the leadingspecies to have a head-start when the environment changed, relative to the single species simulations.To disentangle these factors, we manipulated the two-species simulations right after the extinction of thelagging species in Figure 3.2 d, f. The population of the remaining (leading) species was then reconstructedto resemble the single species community at that same time point, either in population size or in trait mean.Specifically, the remaining population was reinitialized with either (A) the trait mean of the single speciesbut the larger population size of the leading species (with a higher Km), or (B) the smaller population sizeof the single species (with a lower Km) but the trait mean of the leading species. This was accomplished byresampling with replacement from either the leading species population (for a leading trait mean) or fromthe single species population (for a single species trait mean) until there were enough individuals to make upthe required population size.When the leading species had the larger initial population size (Figure 3.2 d), its extended persistencewas largely due to its size. As a consequence of its larger population, the leading species had more geneticvariation and evolved faster phenotypically, exhibiting a mean trait change per generation of 2.6 × 10−5for the leading species, compared to 1.7 × 10−5 for the single species, over the first 500 generations afterthe environment began to change. By restarting the simulations when the lagging species went extinct, wedetermined that the greater population size of the leading species accounted for 78% of the difference inpersistence between the leading species and the single species (median extinction times for single species:10200, reconstructed population type (A) with the population size of the leading species but the trait meanof the single species: 10900, leading species: 11100 generations). By contrast, allowing for the head start intrait mean but reducing the population size of the leading species to that of the single species did not extendpersistence (median extinction times for reconstructed population type (B): 10200).When the leading species initially had the smaller population size (Figure 3.2 f ), its extended persistence95B.1. Supplementary Methodswas largely due to its head start. In this example, the ESS trait mean for a single species was only -0.564(the peak of the resource distribution) but was 0.424 for the leading species (Table B.1) in the two-speciescase. As a consequence of this head start, the mean trait value of the leading species remained substantiallyahead of that for the single species over the full period that the single species persisted (compare Figure 3.2fto Figure B.2 f ). Restarting the simulations at the point in time that the lagging species went extinct revealedthat the head start explained 78% of the longer persistence time of the leading species (median extinctiontimes for single species: 4200, reconstructed population type (B) with the trait mean of the leading speciesbut the population size of the single species: 6000, leading species: 6500 generations), whereas keepingthe large population size of the leading species but eliminating the head start accounted for only 17% of theextended persistence (median extinction times for reconstructed population type (A): 4500).96B.2. Supporting TablesB.2 Supporting TablesTable B.1: Equilibrium (initial) population configurations. The values of Km, population sizes, and traitvalues were determined by numerically solving for when W (ui) = 1 and when individuals carrying smallmutations could not invade the system (dW (um)/dum = 0 for um near an ESS trait value u∗), as detailed inthe supplementary Mathematica package. Remaining parameters were set to their default values (Table 3.1).All data has been deposited in Dryad.One-species Community Two-species Communityκ β Km n1 u1 Km nleading nlagging uleading ulagging0 0 6267 2500 0 11828 2500 2500 0.293 -0.2934 0 5669 2500 -0.564 10756 621 4378 0.424 -0.602-4 0 5669 2500 0.564 10756 4378 621 0.602 -0.4240 0.6 7816 2500 0.665 14492 4344 656 0.672 -0.4440 -0.6 7816 2500 -0.665 14492 656 4344 0.444 -0.67297B.2. Supporting TablesTable B.2: Extinction time, in generations, with higher mutation rate (ν) or mutational effect sizes (σν).Populations were recorded every 100 generations for extinction or persistence. If no replicate went extinctby 15000 generations, an “∗” is reported. In all other cases, the mean extinction time is recorded, with SEMin parentheses. Rate of environmental change was set to Vm = 0.00005 (slow) or Vm = 0.00005 (fast),except where noted. Remaining parameters were set to their default values (Table 3.1). All data has beendeposited in Dryad.Slow environmental change Fast environmental changeκ β Leading Lagging Leading LaggingDefault parameters (ν = 0.02, σν = 0.0015)4 0 ∗ 14840 (160) 6480 (37) 1380 (20)-4 0 ∗ 7700 (396) 11100 (45) 520 (20)Higher mutation rate (ν = 0.2)4 0 ∗; 28400† 41625 (1556); ∗‡ * 1620 (49)[4; 1 rep] [4; 1 rep]4 0 Intermediate rate (Vm = 0.00015): * 6490 (139)§4 0 Intermediate rate (Vm = 0.0001): * 11980 (446)¶-4 0 ∗ ∗ * 920 (86)Larger mutation effect size (σν = 0.015)4 0 ∗ ∗ ∗; 4967 (338) 3535 (184); ∗[17; 3 reps||] [17; 3 reps]-4 0 ∗ ∗ * 3200 (339)†Average values include only those replicates that were run to extinction. Leading species went extinct in one replicate that wasextended until extinction.‡Average values include only those replicates that were run to extinction. Lagging species went extinct in four replicates thatwere extended until extinction.§Two replicates per burn-in population were run with an intermediate rate of environmental change (Vm = 0.00015) to allowgreater trait convergence before extinction.¶Two replicates per burn-in population were run with an intermediate rate of environmental change (Vm = 0.0001) to allowgreater trait convergence before extinction.||A total of four replicates were performed per burn-in population. Exclusion of the leading species was observed in one rep fromeach of burn-in populations: 1, 2, 4.98B.3. Supporting FiguresB.3 Supporting Figures–4 –2 2 4(a) αij0.20.40.60.81.01.2Phenotype, ui100020003000400050000–2 –1 1 20Phenotype, uiK(ui )(b)(c)100020003000400050000Figure B.1: Symmetric and asymmetric competition coefficients and their impact on population size andtrait value. (a) The competitive impact (Equation (3.7)) of an individual of phenotype uj = 0 exerted uponindividuals of phenotype ui is shown for β = 0 (solid: symmetrical), β = 0.6 (dashed: right-skewed),and β = −0.6 (dotted: left-skewed). Right-hand panels show the resource distributions (black: solid anddashed curves), the initial ESS for each species (thin red and blue vertical lines), the post-burn-in speciesdistributions (in 0.01 bins, narrow red and blue histograms, doubled in height for clarity), and the competitionexerted by those two populations on individuals of phenotype ui (red and blue curves, respectively) for (b)the symmetrical case (κ = β = 0) and (c) the case with left-skewed competitive coefficients (κ = 0,β = −0.6), where the species with a negative phenotype (red) has a larger population size than the specieswith a positive phenotype (blue).99B.3. Supporting FiguresTrait ValueGeneration0 5000 10000 15000-5000**(b)(a)(d)(c)(f)(e)-0.50.00.51.0*0 5000 10000 15000-50001.5-0.50.00.51.01.5-0.50.00.51.01.5Slow Environmental Change Fast Environmental ChangeResource peakPop sizeResource peakResource peakPop sizePop sizePop sizePop sizePop sizeFigure B.2: Asymmetrical resource distributions and the evolution of a one-species community in a changingenvironment. Identical to Figure 3.2, except that there is only one species and y-axis ranges from 0 to 3500for the inset population size graphs. All data has been deposited in Dryad.100B.3. Supporting FigureseulaV tiarTGeneration0 5000 10000 15000-5000 0 5000 10000 15000-5000*0.00.51.01.5(b)(a)(c) (d)(f)(e)**-0.50.01.01.5-0.50.5Lagging (smaller N)Leading (equal N)Lagging (equal N)Leading (larger N)Lagging (larger N)Leading (smaller N) Resource peakResource peakResource peakPop sizePop sizePop sizeSlow Environmental Change Fast Environmental ChangePop sizePop sizePop sizeFigure B.3: The evolution of two-species communities with asymmetrical competition coefficients in theface of environmental change. Symbols as in Figure 3.2, except that the resource distribution was symmetric(κ = 0) with the dashed line illustrating the trajectory of the resource peak (equal to the mean), while theshape of the competition function was varied. Specifically, the impact of competition was either symmetric(β = 0, panels a,b), right skewed (β = 0.6, panels c,d), or left skewed (β = −0.6, panels e,f ), with theshaded distributions along the y-axis showing the effective competition witnessed by individuals of a giventrait value induced by the two species at their initial ESS values. Remaining parameters were set to theirdefault values (Table 3.1). All data has been deposited in Dryad.101B.3. Supporting FiguresTrait ValueGeneration**A BC DE F*0 5000 10000 15000-5000-0.50.00.51.01.5-0.50.00.51.01.5-0.50.00.51.01.5Slow Environmental Change Fast Environmental ChangeResource peakPop sizeResource peakResource peakPop sizePop sizePop sizePop sizePop sizeFigure B.4: The evolution of a one-species community with asymmetrical competition coefficients in theface of environmental change. Identical to Figure B.3, except that there is only one species and the insetpopulation size graphs have a y-axis range of 0 to 3500. All data has been deposited in Dryad.102B.3. Supporting FigureseulaV tiarTGeneration0 5000 10000 15000-5000 0 5000 10000 15000-50000.00.51.01.5(a) (b)(d)(c)-0.5 Lagging (smaller N)Leading (larger N)Lagging (larger N)Leading (smaller N)Resource peakResource peakPop sizePop sizeSlow Environmental Change Fast Environmental Change0.00.51.01.5-0.5Pop sizePop sizeFigure B.5: Evolutionary response to a changing environment with a higher mutation rate (ν = 0.2). Theresource distribution is asymmetrical and otherwise identical to Figure 3.2 (panels c-f ), but with a broadery-axis range. Note that populations better track the changing environment and extinctions are delayed withthe higher mutation rate (see Table B.2). All data has been deposited in Dryad.103B.3. Supporting FiguresTrait ValueGeneration0 5000 10000 15000-5000 0 5000 10000 15000-50000123(a) (b)(d)(c)-1Lagging (smaller N)Leading (larger N)Lagging (larger N)Leading (smaller N)Resource peakResource peakPop sizePop sizeSlow Environmental Change Fast Environmental Change0123-1Pop sizePop sizeFigure B.6: Evolutionary response to a changing environment when mutation effect sizes are drawn froma broader distribution (σν = 0.015). The resource distribution is asymmetrical and otherwise identical toFigure 3.2 (panels c-f ), but with a broader y-axis range. Note that populations better track the changing en-vironment (large population (red) falls very near the resource peak given by the dashed line) and extinctionsare delayed when mutations are larger (see Table B.2). In one replicate in panel d, the lagging species (red)persisted, while the leading species (blue) went extinct; this replicate is separated from the remaining fourreplicates and drawn as a line. All data has been deposited in Dryad.104Appendix CAppendix for Chapter 4: Adaptation toelevated CO2 in different biodiversitycontextsC.1 Supplementary MethodsC.1.1 Analysis with assay environment instead of change in environment.Here we present a more in depth analysis of our results and re-analyse our data using the assay environments(CO2ass and divass) (not the change in environments) as fixed factors.Single effects of diversity and CO2 environments: P. pratensis plants were found to respond to the assayenvironment in which they were growing. Plants assayed in species-poor communities produced 91% moreaboveground, 70% more belowground, and 79% more total biomass than plants assayed in species-rich plots(aboveground biomass: F1, 51.7 = 20.2, P < 0.0001; belowground biomass: F1, 66.5 = 44.9, P<0.0001; totalbiomass F1, 73.5 = 51.4, P<0.0001) (Table C.4). Assay plot richness also strongly influenced flowering: 41%of individuals planted into species-poor plots flowered while only 9% of individuals planted into species-richplots flowered (Supplementary Figure C.3d and Table C.6, divcur P<0.0001). These results likely reflect theincreased intensity of competition in the multi-species plots (Supplementary Figure C.3). In our analysiswith ∆div this single effect manifested itself as a highly significant divsel × ∆div interaction (Table C.3,C.5, P<0.0001).P. pratensis also responded to the assay CO2 environment. Plants assayed in eCO2 produced 37%more belowground biomass (F1, 6.1 = 14.5, P=0.009) and 34% more total biomass (F1, 6.0 = 10.8, P=0.02)than plants assayed in aCO2. However, this response was only marginally significant for abovegroundbiomass (F1, 5.7 = 5.0, P=0.07) (Supplementary Table C.4) and not significant for flowering (Supplemen-tary Table C.6). Again in our analysis with ∆CO2 this single effect manifested itself as a significant CO2sel×∆CO2 interaction (Supplementary Table C.3).In our analysis using selection and assay environments (instead of change in the environment), we alsofound a significant interaction between CO2ass × divsel × divass, for aboveground biomass (F1,685.4 = 6.0,P=0.015), belowground biomass(F1,686.5 = 4.3, P=0.038) and total biomass (F1,680.4=4.43, P=0.036) (Sup-plementary Table C.4) and a marginally significant interaction for inflorescence production (P=0.06; Sup-plementary Table C.6). As illustrated in Supplementary Figure C.3a-c, plants that experienced selectionin species-poor plots and were assayed in species-poor plots showed the greatest biomass, but this advan-tage was substantial only when also assayed in eCO2. This suggests that eCO2 serves to amplify or reveal105C.1. Supplementary Methodsadaptations to communities with low diversity.C.1.2 Correlation between aboveground biomass and percent cover of species in the assayplot.To assess whether the degree of adaptation observed in the high diversity plots was driven by diversity perse or by the abundance of a particular species, we tested, using a GLM, whether aboveground biomassproduction was correlated with the percent cover of any of the species in the assay plot when plants wereselected and assayed in the same environment. We did not find that there was any one species that stronglypredicted the evolutionary change in aboveground biomass in P. pratensis, although Andropogon gerardii(a dominant C4 grass) had a marginally negative influence in both ambient and elevated CO2 environments(Supplementary Table C.7a and C.7b). We performed the same test but using percent cover of functionalgroups, instead of species identity, and again found that no single functional group strongly predicted theresponse in aboveground biomass of P. pratensis, although C3 and C4 grasses had a marginally negativeinfluence in aCO2 (Supplementary Tables C.8a and C.8b).C.1.3 History and construction of experimental assay plotsDue to the disturbance that planting and watering a substantial number of ramets back into the originalBioCON plots would have created, we used supplementary plots that were created in the BioCON ringsin 1999 but that were not in use at the time of our study. These supplementary plots (1.5m × 2m, twelveper BioCON ring), were created within the FACE rings on the edge of the main BioCON plots. Of thesupplementary plots, half had received additional nitrogen and were not used in the current study, leavingsix plots in six rings (36 plots in total) (from now on referred to as assay plots). Briefly, the history of theassay plots is as follows: in 1999 eight pin oak (Quercus ellipsoidalis) acorns were planted per plot and anyvegetation that managed to invade from the surrounding old field or from BioCON was allowed to establish(Davis et al., 2007). In 2007 all 36 assay plots were tilled. The six assay plots on the northern side of each ringwere then planted with 12 trees and kept free of any invading vegetation through regular weeding until 2011when all trees were harvested and vegetation from the surroundings was allowed to re-establish. In contrast,the six assay plots on the southern side of each ring were seeded with 6 prairie species (Andropogon gerardii,Bromis inermis, Lespedeza capitata, Lupinus perennis, Elymus canadensis and Panicum vigatum), the firstfour of which are also present in the BioCON experiment. In 2008 the southern assay plots were weeded ofall species not initially seeded, and in 2009 they were weeded of any flowering non-seeded species. After2009 invading vegetation was allowed to establish.As of 2012 (i.e., the initiation of our experiment), the dominant species of the previously treed assay plotswas P. pratensis, such that these plots were very close ecological analogues of the P. pratensis monocultureplots in BioCON (e.g., they were comparable in abundance of P. pratensis, Supplementary Table C.9 andSupplementary Figure C.4a-b). We weeded out all other species and used these assay plots as the species-poor treatment. The three southern assay plots were comparable in species composition and abundance tothe 16-species BioCON plots and were thus used as our species-rich treatment (Supplementary Table C.9,Supplementary Figure C.4c-d). Although our assay plots were not the same as the plots in which selection106C.2. Supporting Tablestook place, we still found a significant CO2sel × CO2cur × divsel × divcur interaction for aboveground andtotal biomass (Figure 4.3a, c Supplementary Table C.4) suggesting that plants grew best when planted backinto the same environment they were selected in (’home’ plot advantage). This supports our view that theassay plots were similar to their matched BioCON selection plots.C.2 Supporting Tables107C.2.SupportingTablesTable C.1: Model output for analysis of aboveground, belowground, total biomass and survival and inflorescence data, when diversity environmentis held constant i.e. plants selected in low diversity were assayed in low diversity. Biomass data was log transformed and analysed using linearmixed effects models, with F-tests for fixed effects constructed in R and denominator degrees of freedom (dfD) obtained from the Satterthwaiteapproximation. Significance of random effects was determined by likelihood ratio tests. Survival and number of inflorescences was determined usingaster models to account for dependence among fitness estimates. The significance of each factor was evaluated using a likelihood ratio test. Notethat the significant CO2sel × ∆CO2 term is a consequence of significant CO2ass effects (see Supplementary Table C.4) and are measures of a plasticresponse to the assay environment.Aboveground biomass (g) Belowground biomass (g) Total biomass (g) Survival & number of inflorescencesdf dfD F value dfD F value dfD F value Test df Test DevianceCO2sel 1 11.1 0.1 6.0 1.6 55.7 1.0 1 0.1∆CO2 1 306.6 7.8** 301.8 0.5 302.9 4.1* 1 0.4div 1 12.2 23.9** 6.9 46.8*** 42.0 46.0**** 1 14.6***∆CO2 × div 1 306.4 0.1 301.6 2.0 302.7 1.1 1 0.9CO2sel ×∆CO2 1 3.7 4.7 3.6 12.2* 3.7 7.6 1 0.2CO2sel × div 1 11.1 3.2 6.1 1.0 55.7 2.3 1 7.6**Random effects n p p p pSelection plot 20 ns ns ns 0.02Mother nested in selection plot 64 0.0006*** 0.01* 0.0001*** 0.005**Assay ring 6 ns ns ns -Assay plot nested in ring 35 ns ns ns 0.0006***ns P > 0.1; † P < 0.1; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001108C.2.SupportingTablesTable C.2: Model output for analysis of aboveground, belowground, total biomass and survival and inflorescence data, when average over the diversityassay environment. Biomass data was log transformed and analysed using linear mixed effects models, with F-tests for fixed effects constructed inR and denominator degrees of freedom (dfD) obtained from the Satterthwaite approximation. Significance of random effects was determined bylikelihood ratio tests. Survival and number of inflorescences was determined using aster models to account for dependence among fitness estimatesand the significance of each factor was evaluated using a likelihood ratio test. Note that the significant CO2sel × ∆CO2 term is a consequence ofsignificant CO2ass effects (see Supplementary Table C.4) and are measures of a plastic response to the assay environment.Aboveground biomass (g) Belowground biomass (g) Total biomass (g) Survival & number of inflorescencesdf dfD F value dfD F value dfD F value Test df Test DevianceCO2sel 1 59.6 0.2 59.1 2.6 60.3 1.7 1 0.9∆CO2 1 682.6 2.5 667.6 0.05 671.0 1.1 1 2.5divsel 1 59.9 0.9 59.5 0.7 60.7 0.7 1 2.2∆CO2 × divsel 1 681.8 0.003 666.1 0.3 670.3 0.2 1 0.9CO2sel ×∆CO2 1 3.5 3.6 32.7 10.7** 32.9 7.8** 1 0.7CO2sel × divsel 1 59.6 3.0† 59.1 0.5 60.3 1.4 1 2.4Random effects n p p p pSelection plot 20 ns ns ns 0.02Mother nested in selection plot 64 <0.0001**** <0.0001**** <0.0001**** <0.0001****Assay ring 6 ns ns ns -Assay plot nested in ring 35 <0.0001**** <0.0001**** <0.0001**** <0.0001****ns P > 0.1; † P < 0.1; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001109C.2. Supporting TablesTable C.3: Linear mixed effects model analysis of log-transformed aboveground, belowground and totalbiomass data. F-tests for fixed effects were constructed in R, with denominator degrees of freedom (dfD)obtained from the Satterthwaite approximation. Significance of random effects was determined by likelihoodratio tests. Note that the significant divsel ×∆div and CO2sel ×∆CO2 terms are a consequence of significantdivass and CO2ass effects (see Supplementary Table C.4) and are measures of a plastic response to the assayenvironment.Aboveground biomass (g) Belowground biomass (g) Total biomass (g)df dfD F value dfD F value dfD F valueCO2sel 1 59.8 0.2 59.4 2.3 60.4 1.6divsel 1 60.3 0.9 59.8 1.0 60.7 0.9∆CO2 1 680.0 2.8 675.1 0.01 671.9 1.3∆div 1 686.9 0.01 687.3 0.9 681.3 0.002CO2sel ×∆CO2 1 3.7 3.7 3.9 13.7* 3.8 9.4*divsel ×∆div 1 24.4 32.4**** 20.3 99.5**** 24.1 100.4****CO2sel × divsel 1 59.9 3.0 † 59.5 0.3 60.5 1.4∆CO2 × divsel 1 680.0 0.03 675.0 0.5 672.3 0.1CO2sel ×∆div 1 683.0 0.01 679.7 0.1 675.1 0.01∆CO2 ×∆div 1 683.7 4.8* 681.0 0.5 676.6 3.3†CO2sel ×∆CO2 ×∆div 1 689.1 5.8* 690.3 4.2* 684.3 4.2*CO2sel ×∆CO2 × divsel 1 687.5 1.4 686.0 1.2 681.3 1.6Random effects n p p pSelection plot 20 ns ns nsMother nested in selection plot 64 p<0.0001**** p<0.0001**** p<0.0001****Assay ring 6 ns ns nsAssay plot nested in ring 35 0.0001*** ns 0.07†ns P > 0.1; † P < 0.1; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001110C.2. Supporting TablesTable C.4: Linear mixed effects model analysis of log-transformed aboveground, belowground and total(aboveground + belowground) biomass data. F-tests for fixed effects were constructed in R, with denomina-tor degrees of freedom (dfD) obtained from the Satterthwaite approximation. Significance of random effectswas determined by likelihood ratio tests. Selection and assay environments are indicated as “ sel” and “ ass”respectively.Aboveground biomass (g) Belowground biomass (g) Total biomass (g)df dfD F value dfD F value dfD F valueCO2sel 1 62.3 1.0 62.3 0.4 62.8 0divsel 1 60.2 0.9 59.8 1.0 60.7 0.8CO2ass 1 5.7 5.0† 6.1 14.5** 6.0 10.8*divass 1 51.7 20.2**** 66.5 44.9**** 73.5 51.4****CO2sel × CO2ass 1 676.6 1.4 674.1 0.4 670.3 0.2divsel × divass 1 684.5 0.0 685.2 0.8 678.8 0.02CO2sel × divsel 1 59.9 3.0 † 59.6 0.3 60.5 1.4CO2ass × divsel 1 684.4 1.4 683.6 1.2 678.4 1.6CO2sel × divass 1 679.1 0.5 678.2 0.4 672.1 0.4CO2ass × divass 1 52.1 2.8 67.1 3.2† 74.6 3.2†CO2sel × CO2ass × divass 1 681.9 1.1 680.9 0.2 670.0 0.2CO2sel × divsel × divass 1 680.3 0.02 677.8 0.1 672.8 0.03CO2sel × CO2ass × divsel 1 677.5 0.03 673.1 0.5 676.1 0.2CO2ass × divsel × divass 1 685.4 6.0* 686.5 4.3* 680.4 4.4*CO2sel × CO2ass × divsel × divass 1 681.1 4.5* 679.0 0.4 674.3 2.9†Random effects n p p pSelection plot 20 ns ns nsMother nested in selection plot 64 p<0.0001**** p<0.0001**** p<0.0001****Assay ring 6 ns ns nsAssay plot nested in ring 35 0.0001*** ns 0.07†ns P > 0.1; † P < 0.1; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001111C.2.SupportingTablesTable C.5: Summary of Aster model comparisons on P. pratensis survival and inflorescence production. Simpler models were tested against fullermodels, with each term added independently (i.e., single terms were tested against the Null model, two-way interactions were tested against the ΣMain effects model etc.). Analysis of deviance (− 2 log likelihood) and χ 2 test P-values are presented. Note that the significant divsel ×∆div term isa consequence of the significant divass effect (see Supplementary Table C.6) and is a measures of a plastic response to the assay environment.Model ForumlaNull: Response = survival at preceding stage + No. inflorescences + random termsΣ Main effects: Response = survival at preceding stage + No. inflorescences × (CO2sel + ∆CO2 + divsel + ∆div + random terms)Σ Two-way: Response = survival at preceding stage + No. inflorescences × (CO2sel + ∆CO2 + divsel + ∆div + all two-way interactions + random terms)Σ Three-way: Response = survival at preceding stage + No. inflorescences × (CO2sel + ∆CO2 + divsel + ∆div + all two-way interactions + all three-way interactions + random terms)Term Residual d.f. Test d.f. Model deviance Test deviance PNull 5 -1935CO2sel 6 1 -1934 0.94 0.33divsel 6 1 -1932 2.21 0.14∆CO2 6 1 -1931 3.14 0.08∆div 6 1 -1935 0.01 0.94Σ Main effects 9 -1927CO2sel × divsel 10 1 -1925 2.18 0.14CO2sel ×∆CO2 10 1 -1927 0.60 0.44∆CO2 × divsel 10 1 -1926 0.69 0.41CO2sel ×∆div 10 1 -1927 0.03 0.86divsel ×∆div 10 1 -1891 36.28 p<0.0001****∆CO2 ×∆div 10 1 -1924 2.95 0.09†Σ Two-way 15 -1880CO2sel ×∆CO2 × divsel 16 1 -1880 0.50 0.48CO2sel × divsel ×∆div 16 1 -1880 0.55 0.46CO2sel ×∆CO2 ×∆div 16 1 -1880 0.32 0.57∆CO2 × divsel ×∆div 16 1 -1880 0.14 0.71Σ Three-way 19 -1876CO2sel ×∆CO2 × divsel ×∆div 20 1 -1874 2.11 0.15Random TermsNull 5 -1935Selection plot 5 -1935 0.04 0.42Mother 5 -1959 24.20 P<0.0001****Assay plot 5 -2075 140.15 P<0.0001****ns P > 0.1; † P < 0.1; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001112C.2.SupportingTablesTable C.6: Summary of Aster model comparisons to test for differences in P. pratensis survival and inflorescence production as a result of theselection and assay CO2 and diversity environments. Simpler models were tested against fuller models with each term added independently. Analysisof deviance (− 2 log likelihood) and χ 2 test P-values are presented.Model ForumlaNull: Response = survival at preceding stage + No. inflorescences + random termsΣ Main effects: Response = survival at preceding stage + No. inflorescences × (CO2sel + ∆CO2 + divsel + ∆div + random terms)Σ Two-way: Response = survival at preceding stage + No. inflorescences × (CO2sel + ∆CO2 + divsel + ∆div + all two-way interactions + random terms)Σ Three-way: Response = survival at preceding stage + No. inflorescences × (CO2sel + ∆CO2 + divsel + ∆div + all two-way interactions + all three-way interactions + random terms)Term Residual d.f. Test d.f. Model deviance Test deviance PNull 5 -1935CO2sel 6 1 -1934 0.94 0.33divsel 6 1 -1932 2.21 0.14CO2ass 6 1 -1934 0.58 0.45divass 6 1 -1899 35.15 P<0.0001****Σ Main effects 8 -1895CO2sel × divsel 9 1 -1893 2.25 0.13CO2sel × CO2ass 9 1 -1892 2.60 0.11CO2ass × divsel 9 1 -1895 0.31 0.58CO2sel × divass 9 1 -1892 2.93 0.09†divsel × divass 9 1 -1893 2.42 0.12CO2ass × divass 9 1 -1895 0.10 0.75Σ Two-way 15 -1885CO2sel × CO2ass × divsel 16 1 -1885 0.27 0.60CO2sel × divsel × divass 16 1 -1883 1.54 0.22CO2sel × CO2ass × divass 16 1 -1884 1.27 0.26CO2ass × divsel × divass 16 1 -1881 3.66 0.06†Σ Three-way 19 -1879CO2sel × CO2ass × divsel × divass 20 1 -1874 5.04 0.02*Random TermsNull 5 -1935Selection plot 5 -1935 0.04 0.42Mother 5 -1959 24.20 P<0.0001****Assay plot 5 -2075 140.15 P<0.0001****ns P > 0.1; † P < 0.1; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001113C.2. Supporting TablesTable C.7: GLM of aboveground biomass versus percent cover of species. Only species with an averagepercent cover greater than five percent in the assay plots were analyzed.a) Selection in eCO2 and species-rich and assayed in eCO2 and species-rich plotsCoefficients Estimate SE F value Pintercept -3.98 3.11Andropogon gerardii -0.63 0.37 2.92 0.09Bromis inermis -0.00 0.79 0.00 0.99Leptoloma congnatum -0.71 0.49 2.11 0.15Lupinus perennis 0.05 0.72 0.00 0.94Panicum virgatum -0.19 0.31 0.38 0.54Poa pratensis 0.02 0.66 0.00 0.97a) Selection in aCO2 and species-rich and assayed in aCO2 and species-rich plotsCoefficients Estimate SE F value Pintercept -12.07 5.87Andropogon gerardii -1.89 0.97 3.77 0.06Bromis inermis -0.74 0.59 1.58 0.22Leptoloma congnatum 0.04 0.07 0.37 0.55Lupinus perennis -2.71 2.08 1.70 0.20Panicum virgatum -1.95 1.26 -2.40 0.13Poa pratensis 1.99 1.84 1.16 0.29Table C.8: GLM analysis of aboveground biomass and percent cover per functional group. All species wereincluded in the analysis.a) Selection in eCO2 and species-rich and assayed in eCO2 and species-rich plotsCoefficients Estimate SE F value Pintercept 0.089 1.122C4 grass -0.414 0.316 1.72 0.2C3 grass 0.396 0.396 1.0 0.32Forb 0.155 0.114 1.85 0.18Legume 0.244 0.750 0.11 0.74a) Selection in aCO2 and species-rich and assayed in aCO2 and species-rich plotsCoefficients Estimate SE F value Pintercept -2.56 0.51C4 grass -0.77 0.40 3.76 0.06C3 grass -1.04 0.54 3.71 0.06Forb 0.00 0.04 0.01 0.91Legume 0.00 0.36 0.00 0.99114C.2. Supporting TablesTable C.9: Percent cover of species grown in the assay and in the selection (BioCON) plots for both species-poor and species-rich communities.Plot diversity Assay plots Selection plotsaCO2 eCO2 aCO2 eCO2Poa pratensis Poor 46 49 43 42Poa pratensis Rich 18.9 24.8 9 10Achillea millefolium Rich 0.2 0.5 0 0.01Agropyron repense Rich 0.1 0.1 1 0.6Amorpha canescens Rich 0.1 0.4 6 4Andropogon geradii Rich 13.5 14.4 16 25Anemone cylindrica Rich 1.1 0.5 0 0.01Asclepias tuberosa Rich 0.4 0.3 0.2 0.2Bouteloua gracilis Rich 0 0 0.07 0.03Bromis inermis Rich 13.5 10.5 7 6Koeleria cristata Rich 0 0 0.08 0.1Lespedeza capitata Rich 2.9 4.4 15 8Lupinus perennis Rich 21.7 20.8 26 22Petalostemum villosum Rich 0 0 0.2 0Schizachyrium scoparium Rich 1.6 3.1 0.3 0.4Solidago rigida Rich 1.1 2.8 3 5Sorghastrum nutans Rich 0 0 0.07 0.001Leptoloma cognatum* Rich 8.9 11.7 - -Panicum vigatum* Rich 6.9 7.9 - -* Indicates species not planted into the BioCON plots. In BioCON these species are actively weeded out ofany plots in which they appear.115C.3. Supporting FiguresC.3 Supporting FiguresResponse	to	selection(eCO2sel	 -aCO2sel)Assay	CO2 environmentaCO2 eCO2 aCO2 eCO2 aCO2 eCO2 aCO2 eCO2Diversity	selection	and	assay	environments	Species-poor Species-rich Species-poor Species-rich-0.4-0.20.00.20.4 Aboveground biomass (g)      -0.4-0.20.00.20.4 Below-ground biomass (g)      -0.6-0.20.20.6 Total biomass (g)     -2-1012Number of inforescences     -0.4-0.20.00.20.4 Aboveground biomass (g)      -0.4-0.20.00.20.4 Below-ground biomass (g)      -0.6-0.20.20.6 Total biomass (g)     -2-1012Number of inforescences     a bc dFigure C.1: Local adaptation of P. pratensis to elevated CO2 when holding diversity environment constantfor a, aboveground biomass, b, belowground biomass, c, total biomass and d, number of inflorescences. Foreach CO2 assay environment we calculated the difference in biomass (g) or number of inflorescences (± 1standard error of the mean (s.e.m) produced by plants that had previously experienced selection in eCO2 andin aCO2 from the raw data. Note the data points in this figure are the same as in figure 4.3 except that pointsfor individuals that were selected and assayed in different diversity environments have been removed.116C.3. Supporting Figures-0.4-0.20.00.20.4 Aboveground biomass (g)      -0.4-0.20.00.20.4 Below-ground biomass (g)      -0.4-0.20.00.20.4 Total biomass (g)     -1.0-0.50.00.51.0 Number of inforescences     Response	to	selection(eCO2sel	 -aCO2sel)Assay	CO2 environmentaCO2 eCO2 aCO2 eCO2 aCO2 eCO2 aCO2 eCO2Selection	diversity	environment	Species-poor Species-rich Species-poor Species-rich-0.4-0.20.00.20.4 Aboveground biomass (g)      -0.4-0.20.00.20.4 Below-ground biomass (g)      -0.4-0.20.00.20.4 Total biomass (g)     -1.0-0.50.00.51.0 Number of inforescences     a bc dFigure C.2: Adaptation of P. pratensis to elevated CO2 when selected in communities of low and highspecies richness averaged across diversity assay environments for a, aboveground biomass, b, belowgroundbiomass, c, total biomass and d, number of inflorescences. For each CO2assay environment we calculatedthe difference in biomass (g) or number of inflorescences (± 1 s.e.m) produced by plants that had previouslyexperienced selection in eCO2 and in aCO2 from the raw data. Both aboveground and total biomass showedpatterns consistent with adaptation to eCO2, where there was greater biomass in the presence of eCO2 forplants that had previously experience eCO2 with the same holding for plants exposed to aCO2 (positivedashed slopes). The trends were not significant, however, when ignoring the current community, as donehere.117C.3. Supporting Figures0.00.51.01.52.02.5 Aboveground biomass (g)0.00.51.01.52.02.5 Belowground biomass (g)01234Total biomass (g)01234Number of inflorescencespoor	 rich	 poor	 rich	Divsel	poor	 poor	 rich	 rich	Divcur	a	c	b	d	poor	 rich	 poor	 rich	poor	 poor	 rich	 rich	aCO2sel		->	aCO2ass	eCO2sel		->	aCO2ass	aCO2sel		->	eCO2ass	eCO2sel		->	eCO2ass	CO2	environment	0.00.51.01.52.02.5 Aboveground biomass (g)0.00.51.01.52.02.5 Belowground biomass (g)01234Total biomass (g)01234Number of inflorescences0.00.51.01.52.02.5 Aboveground biomass (g)0.00.51.01.52.02.5 Belowground biomass (g)01234Total biomass (g)01234Number of inflorescencesO2 environmentO2  CO2O2  O2O2 -  O2O2  O20.00.51.01.52.02.5 Aboveground biomass (g)0.00.51.01.52.02.5 Belowground biomass (g)01234Total biomass (g)01234Number of inflorescencesCO2 environmentaCO  -> aCO2 -> aCO2 -> eCO22 -> eCO2Figure C.3: Average a, aboveground biomass, b, belowground biomass, c, total biomass and d, number ofinflorescences produced by plants originating from each of the selection environments (CO2 and diversity)and grown in each of the assay environments (CO2 and diversity). Open circles represent the mean of plantssampled from the same CO2 and diversity environment and grown in the same assay plot. The black filledsymbols are the mean (± 1 s.e.m.) biomass and number of inflorescences averaged across all the plots in atreatment from the raw data118C.3. Supporting FiguresSelection	(BioCON)	plots Assay	plotsa bdcFigure C.4: Photographs of the original selection plots (BioCON plots) (left) and the assay plots (transplantplots) used in the experiment (right) with either P. pratensis dominated species-poor a, b, or species-richplots c, d. (Photos taken by Elizabeth J. Kleynhans)119Appendix DAppendix for Chapter 5: How plantdiversity influences intraspecific traitresponses to abiotic environmental changeD.1 Supplementary AnalysisD.1.1 Analysis with species richness instead of functional group diversity.To test whether we obtained similar results when the data was analysed according to species richness insteadof by the functional group diversity of the plot we reanalysed the results including species richness of the plotin the PERMANOVA analysis. The statistical analysis was carried out exactly as described in the methodssection of Chapter 5 except in this analysis species richness of the plot was included in the analysis insteadof functional group diversity. Performing the analysis in this way has the major disadvantage of reducingpower as BioCON only has two monoculture plots per abiotic treatment (see Table D.1 for number of plotsper treatment type).Overall, we found no abiotic treatment × species richness interaction for any species (Table D.4). Theseare identical results to those for the analysis with functional group diversity and abiotic treatment. Similarly,the main effect of abiotic treatment was significant for A. gerardii, L. perennis and S.rigida and matched theresults for the analysis with functional group diversity. Lastly, the main effect of species richness was inlinewith those of functional group richness with the exception of L. capitata which showed significant differencesfor species diversity but not for functional group diversity (Table D.4). This is because monoculture andfour-species plots were significantly different to one another in the analysis with species richness but in theanalysis with functional group richness the four species plots get separated into one and four functionalgroups and thus this difference disappears (Supplementary Figure D.3 G) .D.2 Supporting Tables120D.2.SupportingTablesTable D.1: Number of plots sampled per biotic (species richness and functional group diversity) and abiotic (CO2 and nitrogen) treatment for eachspecies studied. The following abbreviations are used: sp. refers to species richness and fg. refers to functional group diversity.Speciesand func-tionalgroupTreatment AndropogongerardiiBromusinermisPoapratensisLespedezacapitataLupinusperennisAsclepiastuberosaSolidagorigidaCO2 treatment1sp. 1fg.ambient 2 2 2 2 2 2 2elevated 2 2 2 2 2 2 24sp. 1fg.ambient 3 2 3 3 3 3 3elevated 3 3 3 3 3 2 34sp. 4fg.ambient 1 2 2 1 1 0 2elevated 1 0 1 2 0 4 216sp. 4fg.ambient 6 6 6 6 6 6 6elevated 6 6 6 6 6 6 10Nitrogen treatment1sp. 1fg.ambient 2 2 2 2 2 2 2elevated 2 2 2 2 2 2 24sp. 1fg.ambient 3 2 3 3 3 3 3elevated 3 3 3 3 3 3 34sp. 4fg.ambient 1 2 2 1 1 0 2elevated 2 0 2 0 1 2 016sp. 4fg.ambient 6 6 6 6 6 6 6elevated 6 6 6 6 6 6 7121D.2.SupportingTablesTable D.2: The number of individuals measured to determine plant height and the number of leaves sampled to determine SLA, LDMC and leaf areafor each BioCON plot and species in ambient and elevated CO2 Values separated by a "," indicate the number of individuals sampled in separate plots.Species Functionalgroup1sp aCO2 1sp eCO2 4sp aCO2 4sp eCO2 16sp aCO2 16sp eCO2Number of plant height measurements obtained per plotAndropogon ger-ardiiC4 grass 6, 16 16, 15 10, 13, 14, 16 16, 16, 13 15, 16, 15, 16,16, 1615, 14, 16, 16,16, 16Bromus inermis C3 grass 16,16 16, 16 16, 16, 16, 16 16, 16, 16 16, 16, 16, 16,16, 1616, 16, 16, 11,11, 9Poa pratensis C3 grass 16, 16 16, 16 16, 16, 16 16, 16, 16, 16,16, 1616, 16, 16,16,16, 1616, 16, 16, 16,16, 16Lespedeza capitata Legume 15, 9 16, 16 12, 16, 15, 10 15, 13, 9 13, 7, 10, 9, 8,912, 12, 8, 7 6, 8Lupinus perennis Legume 15, 16 16, 16 16, 14, 16, 16 11, 16, 16, 10,1615, 16, 16, 15,14, 1613, 10, 16, 16,14, 16Asclepias tuberosa Forb 8, 16 14, 10 4, 1, 3, 4, 2 2, 6, 2 1, 2, 1, 4, 3 2, 1, 4, 8, 1, 1,1Solidato rigida Forb 16, 16 16, 16 16, 12, 13, 12,1214, 16, 15, 11,163, 4, 6, 4, 2, 2,143, 6, 1, 1, 2, 5,2, 5, 3, 2, 2Number of leaves sampled per plotAndropogon ger-ardiiC4 grass 6, 6 6, 6 6, 6, 5, 6 6, 4, 6, 6 6, 6, 6, 6, 6, 6 6, 5, 6, 6, 6, 6Bromus inermis C3 grass 6, 6 6, 6 6, 6, 6, 6, 6 6, 6, 6, 5 6, 6, 6, 6, 6, 6,66, 6, 6, 6, 6Poa pratensis C3 grass 6, 6 6, 6 5, 6, 6 4, 6, 6, 5, 6, 6,65, 5, 6, 6, 6, 6 6, 6, 6, 6, 6, 6Lespedeza capitata Legume 6, 6 6, 2 6, 6, 6, 5 6, 6, 6 6, 6, 6, 6, 6, 6 6, 6, 6, 6, 6, 6Lupinus perennis Legume 6, 6 6, 6 6, 6, 6, 6 5, 6, 6, 6, 6 6, 6, 6, 6, 6, 6 6, 5, 6, 6, 6, 6,6Asclepias tuberosa Forb 5, 5 6, 3 1, 3, 4, 2 2, 5, 2 3, 4, 2 1, 1, 4, 6, 1, 1Solidato rigida Forb 6, 6 6, 6 6, 6, 6, 6, 6 6, 6, 6, 6, 6 1, 6, 3, 2, 2, 1 2, 4, 1, 1, 4, 15, 3, 1, 4122D.2.SupportingTablesTable D.3: The number of individuals measured to determine plant height and the number of leaves sampled to determine SLA, LDMC and leaf areafor each BioCON plot and species in ambient and elevated nitrogen Values separated by a "," indicate the number of individuals sampled in separateplots.Species Functionalgroup1sp aN 1sp eN 4sp aN 4sp eN 16sp aN 16sp eNNumber of plant height measurements obtained per plotAndropogon ger-ardiiC4 grass 6, 16 16, 16 10, 13, 14, 16 13, 16, 14, 16,1615, 16, 15, 16,16, 1612, 13, 15, 16,15, 16Bromus inermis C3 grass 16, 16 16, 16 16, 16, 16, 16,1616, 16, 15, 16,1616, 16, 16, 16,16, 1616, 16, 16, 16,16, 16Poa pratensis C3 grass 16, 16 16, 16 16, 16, 16 16, 16, 11, 16,1616, 16, 16, 16,16, 1616, 16, 16, 16,16, 16Lespedeza capitata Legume 15, 9 16, 16 12, 16, 15, 10 12, 5, 6, 8 13, 7, 10 ,9, 8,95, 2, 2, 3, 3, 5Lupinus perennis Legume 15,16 16, 16 16, 14, 16, 16 16, 16, 16 15, 16, 16, 15,14, 1614, 5, 14, 11, 9,16Asclepias tuberosa Forb 8, 16 12, 8 4, 1, 3, 4, 2 7, 4, 4 1, 2, 1, 4, 3 1, 1, 1, 2, 2, 4,1, 2Solidato rigida Forb 16, 16 14, 16 16, 12, 13, 12,1215, 16, 15 3, 4, 6, 4, 2, 2,141, 2, 1, 1, 1, 1,1, 1, 1, 3Number of leaves sampled per plotAndropogon ger-ardiiC4 grass 6, 6 6, 6 6, 6, 5, 6 6, 6, 6, 6, 6 6, 6, 6, 6, 6, 6 5, 6, 6, 6, 6, 6Bromus inermis C3 grass 6, 6 6, 6 6, 6, 6, 6, 6 6, 6, 6, 6, 6 6, 6, 6, 6, 6, 6 6, 6, 5, 6, 7, 6Poa pratensis C3 grass 6, 6 6, 6 5, 6, 6 6, 6, 6, 6, 6 5, 5, 6, 6, 6, 6 6, 6, 6, 6, 6, 6Lespedeza capitata Legume 6, 6 6, 6 6, 6, 6, 5 6, 4, 6, 5 6, 6, 6, 6, 6, 6 4, 2, 2, 2Lupinus perennis Legume 6, 6 5, 6 6, 6, 6, 6 6, 6, 6 6, 6, 6, 6, 6, 6,66, 6, 6, 6, 5, 6Asclepias tuberosa Forb 5, 5 6, 4 1, 3, 4, 2 6, 2, 3 3, 4, 2 1, 1, 1, 2, 1, 4,1, 1Solidato rigida Forb 6, 6 6, 6 6, 6, 6, 6, 6 6, 6, 6 1, 6, 3, 2, 2, 1 1, 1, 2, 1, 1, 1,2123D.2. Supporting TablesTable D.4: PERMANOVA results based on Euclidean distances of normalised trait data for each speciesgrown in one, four or 16 species plots and in ambient or elevated CO2 and nitrogen. Bold text indicatessignificant differences (P<0.05, after Bonferroni correction) and † indicates marginally significant results(P<0.1 after Bonferroni correction)Species Functionalgroupmember-shipAbiotic treatment Species richness Abiotic treat-ment × SpeciesrichnessR2 F-value R2 F-value R2 F-valueAndropogon gerardii C4 grass 0.16 4.19** 0.26 7.02** 0.10 1.39Bromis inermis C3 grass 0.13 2.78 0.14 3.11† 0.07 0.74Poa pratensis C3 grass 0.08 2.29 0.37 10.84** 0.05 0.77Lupinus perennis Legume 0.15 3.41* 0.18 4.25* 0.07 0.82Lespedeza capitata Legume 0.08 1.74 0.19 3.95** 0.06 0.58Asclepias tuberosa Forb 0.05 1.01 0.20 3.74* 0.10 0.97Solidago rigida Forb 0.20 4.88** 0.03 0.81 0.13 1.64ns P > 0.1; † P < 0.1; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001124D.3. Supporting FiguresD.3 Supporting Figures125D.3. Supporting FiguresAndropogon gerardii Bromus inermis Poa pratensis300400500600300400500500600700800Plant heightAndropogon gerardii Bromus inermis Poa pratensis1214161718192021222316171819SLAAndropogon gerardii Bromus inermis Poa pratensis350375400425450360380400340350360370380LDMCAndropogon gerardii Bromus inermis Poa pratensis40060080010005006007001000150020002500Leaf areaAndropogon gerardii Bromus inermis Poa pratensis1 4 1 4 1 40.010.020.030.040.050.060.00100.00150.00200.00250.00300.00250.00500.0075Functional group diversitySeed weightAbiotic treatement aC & aN eC & aN aC & eNFigure D.1126D.3. Supporting FiguresFigure D.1: The response in plant height (mm), SLA (mm2.mg-1), LDMC (mg.g-1), one sided leaf area(mm2) and seed mass (g) to ambient CO2 and ambient nitrogen, elevated CO22 and ambient nitrogen, andambient CO2 and elevated nitrogen when grown in plots containing one for four functional groups for An-dropogon gerardii, Bromus inermis and Poa pratensis. Black points are the average values of a specific traitobtained in a plot. The box plot represents the median value, with the lower and upper hinges correspondingto the first and third quartiles. The upper and lower whiskers extend from the hinge to the largest/smallestvalue no further than 1.5 × the inter-quartile range. Data beyond the whiskers are considered outliers.127D.3. Supporting FiguresLespedeza capitata Lupinus perennis Asclepias tuberosa Solidago rigida200300400400500600240280320400500600700800Plant heightLespedeza capitata Lupinus perennis Asclepias tuberosa Solidago rigida89101112162020251011121314SLALespedeza capitata Lupinus perennis Asclepias tuberosa Solidago rigida340360380200250300120130140150160340360380400420LDMCLespedeza capitata Lupinus perennis Asclepias tuberosa Solidago rigida400600800100050010001500200024002800400500600700Leaf areaLespedeza capitata Lupinus perennis Asclepias tuberosa Solidago rigida1 4 1 4 1 4 1 42e−043e−044e−045e−040.40.50.60.0200.0250.0300.0010.0020.0030.004Functional group diversitySeed weightAbiotic treatement aC & aN eC & aN aC & eNFigure D.2128D.3. Supporting FiguresFigure D.2: The average response in plant height (mm), SLA (mm2.mg-1), LDMC (mg.g-1), one sided leafarea (mm2) and seed mass (g) to ambient CO2 and ambient nitrogen, elevated CO22 and ambient nitrogen,and ambient CO2 and elevated nitrogen when grown in plots containing one for four functional groupsfor Lespedeza capitata, Lupinus perennis, Asclepias tuberosa and Solidago rigida. Seed mass data for A.tuberosa is excluded due to very few pods being produced by any plants in any plots. Black points are theaverage values of a specific trait obtained in a plot. The box plot represents the median value, with the lowerand upper hinges corresponding to the first and third quartiles. The upper and lower whiskers extend from thehinge to the largest/smallest value no further than 1.5 × the inter-quartile range. Data beyond the whiskersare considered outliers.129D.3. Supporting FiguresH (44%)LA (37%)SLA (45%)LDMC (34%)−2−1012−2 −1 0 1 2PCoA1 (38%)PCoA2 (24%)Andropogon gerardii ALDMC (37%)SLA (35%)SW (22%)H (51%)LA (26%−2−1012−2 0 2 4PCoA1 (41%)PCoA2 (25%)Bromus inermis CSLA (26%)H (25%)LDMC (24%)LA (23%)SW (94%−2−1012−4 −2 0 2 4PCoA1 (61%)PCoA2 (19%)Poa pratensis E−2−1012−2 −1 0 1 2PCoA1PCoA2 B−2−1012−2 0 2 4PCoA1PCoA2 D−2−1012−4 −2 0 2 4PCoA1PCoA2 FSpp. richness 1 4 16 abiotic treat aC & aN eC & aN aC & eNFigure D.3: Continued next page 130D.3. Supporting FiguresSW (33%)H (25%)LDMC (43)SLA (35%)−3−2−1012−2 0 2PCoA1 (35%)PCoA2 (32)Lespedeza capitata GSLA (33%)LDMC (32%)H (22%)LA (58%)SW (33%−2−1012−2.5 0.0 2.5PCoA1 (49%)PCoA2 (24%)Lupinus perennis ISLA (35%)LDMC (35%)LA (25%)H (54%)LA (21%−2−1012−4 −2 0 2PCoA1 (50%)PCoA2 (33%)Asclepias tuberosa KSLA (39%)LDMC (32%)LA (63%)H (24%)−3−2−1012−2 0 2PCoA1 (39%)PCoA2 (24%)Solidago rigida M−3−2−1012−2 0 2PCoA1PCoA2 H−2−1012−2.5 0.0 2.5PCoA1PCoA2 J−2−1012−4 −2 0 2PCoA1PCoA2 L−3−2−1012−2 0 2PCoA1PCoA2 NSpp. richness 1 4 16 Abiotic treat aC & aN eC & aN aC & eNFigure D.3: Continued next page131D.3. Supporting FiguresFigure D.3: Principal coordinate analysis of trait responses of Andropogon gerardii (C4 grass) (A, B, C),Bromus inermis (C3 grass) (D, E, F), and Poa pratensis (C3 grass) (G, H, I), Lespedeza capitata (legume)(J, K, L), Lupinus perennis (legume) (M, N, O), Asclepias tuberosa (forb) (P, Q, R), and Solidago rigida(forb) (S, T, U) to ambient and elevated CO2 and nitrogen conditions when grown in communities of one,four or 16 species. For each species the data shown in both panel is the same, it is just subset in differentways to show the pattern according to that treatment. Panels A, C, E, G, I, K, and M show the data subsetaccording to species richness. Panels B, D, F, H, J, L, and N is the data subset according to abioitic treatment.Figures with a grey background indicate significant results (P<0.05, after Bonferroni correction) from thePERMANOVA analysis, also see Table D.4 for more details. Plant traits that explained more than 20% alongan axis are indicated on the plot for species richness (Panels A, C, E, G, I, K, and M) and applies to all otherpanels. Black text indicates a positive association with that PCoA axis while grey text indicates a negativeassociation. H = plant height; LA = leaf area; LDMC = leaf dry matter content; SLA = specific leaf area,and SW = seed weight132

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0375772/manifest

Comment

Related Items